LANDSLIDES AND ENGINEERED SLOPES

PROCEEDINGS OF THE TENTH INTERNATIONAL SYMPOSIUM ON LANDSLIDES

AND ENGINEERED SLOPES, 30 JUNE–4 JULY, 2008, XI’AN, CHINA

Landslides and Engineered Slopes

From the Past to the Future

Editors

Zuyu Chen
China Institute of Water Resources and Hydropower Research, Beijing, China

Jianmin Zhang
Department of Hydraulic Engineering, Tsinghua University, Beijing, China

Zhongkui Li
Department of Hydraulic Engineering, Tsinghua University, Beijing, China

Faquan Wu
Institute of Geology and Geophysics, Chinese Academy of Sciences, Beijing, China

Ken Ho
Civil Engineering and Development Department, Hong Kong SAR, China

VOLUME 1
CRC Press/Balkema is an imprint of the Taylor & Francis Group, an informa business

© 2008 Taylor & Francis Group, London, UK

Typeset by Vikatan Publishing Solutions (P) Ltd., Chennai, India

Printed and bound in Great Britain by Antony Rowe (A CPI-group Company), Chippenham, Wiltshire.

All rights reserved. No part of this publication or the information contained herein may be reproduced, stored
in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, by photocopying,
recording or otherwise, without written prior permission from the publisher.

Although all care is taken to ensure integrity and the quality of this publication and the information herein, no
responsibility is assumed by the publishers nor the author for any damage to the property or persons as a result
of operation or use of this publication and/or the information contained herein.

Published by: CRC Press/Balkema

P.O. Box 447, 2300 AK Leiden, The Netherlands
e-mail: Pub.NL@taylorandfrancis.com
www.crcpress.com – www.taylorandfrancis.co.uk – www.balkema.nl

ISBN set: 978-0-415-41196-7 (2 volumes + CD)

ISBN Vol.1: 978-0-415-41194-3 (hbk)
ISBN Vol.2: 978-0-415-41195-0 (hbk)
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Table of Contents

Preface XXIII

VOLUME 1

Keynote lectures
Landslides: Seeing the ground 3
N.R. Morgenstern & C.D. Martin
Limit equilibrium and finite element analysis – A perspective of recent advances 25
Z. Chen & K. Ugai
Improving the interpretation of slope monitoring and early warning data through better
understanding of complex deep-seated landslide failure mechanisms 39
E. Eberhardt, A.D. Watson & S. Loew
Effects of earthquakes on slopes 53
I. Towhata, T. Shimomura & M. Mizuhashi
Monitoring and modeling of slope response to climate changes 67
H. Rahardjo, R.B. Rezaur, E.C. Leong, E.E. Alonso, A. Lloret & A. Gens
Soil nailing and subsurface drainage for slope stabilization 85
W.K. Pun & G. Urciuoli

Special lectures
Loess in China and landslides in loess slopes 129
Z.G. Lin, Z.J. Xu & M.S. Zhang
Advances in landslide continuum dynamic modelling 145
S. McDougall, M. Pirulli, O. Hungr & C. Scavia
Deformation and failure mechanisms of loose and dense fill slopes with and without soil nails 159
C.W.W. Ng
Capturing landslide dynamics and hydrologic triggers using near-real-time monitoring 179
M.E. Reid, R.L. Baum, R.G. LaHusen & W.L. Ellis
The effects of earthquake on landslides – A case study of Chi-Chi earthquake, 1999 193
M.L. Lin, K.L. Wang & T.C. Kao
The role of suction and its changes on stability of steep slopes in unsaturated granular soils 203
L. Olivares & P. Tommasi
Prediction of landslide movements caused by climate change: Modelling the behaviour
of a mean elevation large slide in the Alps and assessing its uncertainties 217
Ch. Bonnard, L. Tacher & M. Beniston

Geology, geotechnical properties and site characterization

Geotechnical appraisal of the Sonapur landslide area, Jainita hills, Meghalya, India 231
R.C. Bhandari, P. Srinivasa Gopalan & V.V.R.S. Krishna Murty

V
The viscous component in slow moving landslides: A practical case 237
D.A. González, A. Ledesma & J. Corominas
The systematic landslide investigation programme in Hong Kong 243
K.K.S. Ho & T.M.F. Lau
General digital camera-based experiments for large-scale landslide physical
model measurement 249
X.W. Hu, H.M. Tang & J.S. Li
Shear strength of boundaries between soils and rocks in Korea 257
S.G. Lee, B.S. Kim & S.H. Jung
Cracks in saturated sand 263
X.B. Lu, S.Y. Wang & P. Cui
Some geomorphological techniques used in constraining the likelihood
of landsliding – Selected Australian examples 267
A.S. Miner, P. Flentje, C. Mazengarb, J.M. Selkirk-Bell & P.G. Dahlhaus
Rock failures in karst 275
M. Parise
Geotechnical study at Sirwani landslide site, India 281
V.K. Singh
Inferences from morphological differences in deposits of similar large rockslides 285
A.L. Strom
Movements of a large urban slope in the town of Santa Cruz do Sul (RGS), Brazil 293
L.A. Bressani, R.J.B. Pinheiro, A.V.D. Bica, C.N. Eisenberger & J.M.D. Soares
Geotechnical analysis of a complex slope movement in sedimentary successions of the
southern Apennines (Molise, Italy) 299
D. Calcaterra, D. Di Martire, M. Ramondini, F. Calò & M. Parise
Application of surface wave and micro-tremor survey in landslide investigation
in the Three Gorges reservoir area 307
A. Che, X. Luo, S. Feng & O. Yoshiya
A case study for the landslide-induced catastrophic hazards in Taiwan Tuchang Tribute 313
C.Y. Chen & W.C. Lee
Pir3D, an easy to use three dimensional block fall simulator 319
Y. Cottaz & R.M. Faure
Characterization of the fracture pattern on cliff sites combining geophysical imaging
and laser scanning 323
J. Deparis, D. Jongmans, B. Fricout, T. Villemin, O. Meric, A. Mathy & L. Effendiantz
In situ characterization of the geomechanical properties of an unstable
fractured rock slope 331
C. Dünner, P. Bigarré, F. Cappa, Y. Guglielmi & C. Clément
Properties of peat relating to instability of blanket bogs 339
A.P. Dykes
Stability problems in slopes of Arenós reservoir (Castellón, Spain) 347
J. Estaire, J.A. Díez & C. Olalla
The 22 August, 2006, anomalous rock fall along the Gran Sasso NE wall
(Central Apennines, Italy) 355
G.B. Fasani, C. Esposito, G.S. Mugnozza, L. Stedile & M. Pecci

VI
New formulae to assess soil permeability through laboratory identification and flow
coming out of vertical drains 361
J.C. Gress
Structure-controlled earth flows in the Campania Apennines (Southern Italy) 365
F.M. Guadagno, P. Revellino, G. Grelle, G. Lupo & M. Bencardino
Geotechnical and mineralogical characterization of fine grained soils affected by soil slips 373
G. Gullà, L. Aceto, S. Critelli & F. Perri
Vulnerability of structures impacted by debris flow 381
E.D. Haugen & A.M. Kaynia
Engineering geological study on a large-scale toppling deformation
at Xiaowan Hydropower Station 389
R. Huang, G. Yang, M. Yan & M. Liu
Characterization of the Avignonet landslide (French Alps) with seismic techniques 395
D. Jongmans, F. Renalier, U. Kniess, S. Schwartz, E. Pathier, Y. Orengo,
G. Bièvre, T. Villemin & C. Delacourt
Deformation characteristics and treatment measures of spillway
slope at a reservoir in China 403
N. Ju, J. Zhao & R. Huang
Sliding in weathered banded gneiss due to gullying in southern Brazil 409
W.A. Lacerda, A.P. Fonseca & A.L. Coelho Netto
Experimental and three-dimensional numerical investigations of the impact of dry
granular flow on a barrier 415
R.P.H. Law, G.D. Zhou, C.W.W. Ng & W.H. Tang
Temporal survey of fluids by 2D electrical tomography: The ‘‘Vence’’ landslide observatory
site (Alpes-Maritimes, SE France) 421
T. Lebourg, S. El Bedoui, M. Hernandez & H. Jomard
Characteristics of landslides related to various rock types in Korea 427
S.G. Lee, K.S. Lee, D.C. Park & S. Hencher
Two approaches to identifying the slip zones of loess landslides and related issues 435
T. Li & X. Lin
Testing study on the strength and deformation characteristics of soil in loess landslides 443
H.J. Liao, L.J. Su, Z.D. Li, Y.B. Pan & H. Fukuoka
Failure mechanism of slipping zone soil of the Qiangjiangping landslide in the Three
Gorges reservoir area: A study based on Dead Load test 449
X. Luo, A. Che, L. Cao & Y. Lang
Post-failure movements of a large slow rock slide in schist near Pos Selim, Malaysia 457
A.W. Malone, A. Hansen, S.R. Hencher & C.J.N. Fletcher
Characteristics of rock failure in metamorphic rock areas, Korea 463
W. Park, Y. Han, S. Jeon & B. Roh
Shape and size effects of gravel grains on the shear behavior of sandy soils 469
S.N. Salimi, V. Yazdanjou & A. Hamidi
Nonlinear failure envelope of a nonplastic compacted silty-sand 475
D.D.B. Seely & A.C. Trandafir
An investigation of a structurally-controlled rock cut instability at a metro station shaft
in Esfahan, Iran 481
A. Taheri

VII
Yield acceleration of soil slopes with nonlinear strength envelope 487
A.C. Trandafir & M.E. Popescu
Evaluation of rockfall hazards along part of Karaj-Chaloos road, Iran 491
A. Uromeihy, N. Ghazipoor & I. Entezam
Coupled effect of pluviometric regime and soil properties on hydraulic boundary
conditions and on slope stability 495
R. Vassallo, C. Di Maio & M. Calvello
Mechanical characters of relaxing zone of slopes due to excavation 501
H. Wang & X.P. Liao
Deformation characteristics and stability evaluation of Ganhaizi landslide in the Dadu River 507
Y. Wang, Y. Sun, O. Su, Y. Luo, J. Zhang, C. Zhou & S. Zhang
Landslide-prone towns in Daunia (Italy): PS interferometry-based investigation 513
J. Wasowski, D. Casarano, F. Bovenga, A. Refice, R. Nutricato & D.O. Nitti
Basic types and active characteristics of loess landslide in China 519
W. Wu, D. Wang, X. Su & N. Wang
Investigation of a landslide using borehole shear test and ring shear test 525
H. Yang, V.R. Schaefer & D.J. White
The importance of geological and geotechnical investigations of landslides occurred
at dam reservoirs: Case studies from the Havuzlu and Demirkent landslides
(Artvin Dam, Turkey) 531
A.B. Yener, S. Durmaz & B.M. Demir
An innovative approach combining geological mapping and drilling process
monitoring for quantitative assessment of natural terrain hazards 535
Z.Q. Yue, J. Chen & W. Gao
Types of cutslope failures along Shiyan-Manchuanguan expressway through
the Liangyun fracture, Hubei Province 543
H. Zhao, R. Wang, J. Fan & W. Lin

Advances in analytical methods, modeling and prediction of slope behavior

Probability limit equilibrium and distinct element modeling of jointed rock slope at northern
abutment of Gotvand dam, Iran 553
M. Aminpoor, A. Noorzad & A.R. Mahboubi
Rock block sliding analysis of a highway slope in Portugal 561
P.G.C. Santarém Andrade & A.L. Almeida Saraiva
Contribution to the safety evaluation of slopes using long term observation results 567
J. Barradas
Delimitation of safety zones by finite element analysis 573
J. Bojorque, G. De Roeck & J. Maertens
Laboratory and numerical modelling of the lateral spreading process involving
the Orvieto hill (Italy) 579
F. Bozzano, S. Martino, A. Prestininzi & A. Bretschneider
Albano Lake coastal rock slide (Roma, Italy): Geological constraints and numerical modelling 585
F. Bozzano, C. Esposito, S. Martino, P. Mazzanti & G. Diano
Superposition principle for stability analysis of reinforced slopes and its FE validation 593
F. Cai & K. Ugai
Soil suction modelling in weathered gneiss affected by landsliding 599
M. Calvello, L. Cascini, G. Sorbino & G. Gullà

VIII
Modelling the transient groundwater regime for the displacements analysis of slow-moving
active landslides 607
L. Cascini, M. Calvello & G.M. Grimaldi
Numerical modelling of the thermo-mechanical behaviour of soils in catastrophic landslides 615
F. Cecinato, A. Zervos, E. Veveakis & I. Vardoulakis
Some notes on the upper-bound and Sarma’s methods with inclined slices for stability analysis 623
Z.Y. Chen
Slope stability analysis using graphic acquisitions and spreadsheets 631
L.H. Chen, Z.Y. Chen & P. Sun
Efficient evaluation of slope stability reliability subject to soil parameter uncertainties
using importance sampling 639
J. Ching, K.K. Phoon & Y.G. Hu
Prediction of the flow-like movements of Tessina landslide by SPH model 647
S. Cola, N. Calabrò & M. Pastor
Applications of the strength reduction finite element method to a gravity dam stability analysis 655
Q.W. Duan, Z.Y. Chen, Y.J. Wang, J. Yang & Y. Shao
Study on deformation parameter reduction technique for the strength reduction finite
element method 663
Q.W. Duan, Y.J. Wang & P.W. Zhang
Stability and movement analyses of slopes using Generalized Limit Equilibrium Method 671
M. Enoki & B.X. Luong
Long-term deformation prediction of Tianhuangpin ‘‘3.29’’ landslide based on neural
network with annealing simulation method 679
F. Zhang, C. Xian, J. Song, B. Guo & Z. Kuai
New models linking piezometric levels and displacements in a landslide 687
R.M. Faure, S. Burlon, J.C. Gress & F. Rojat
3D slope stability analysis of Rockfill dam in U-shape valley 693
X.Y. Feng, M.T. Luan & Z.P. Xu
3-D finite element analysis of landslide prevention piles 697
K. Fujisawa, M. Tohei, Y. Ishii, Y. Nakashima & S. Kuraoka
Integrated intelligent method for displacement predication of landslide 705
W. Gao
A new approach to in situ characterization of rock slope discontinuities: the ‘‘High-Pulse
Poroelasticity Protocol’’ (HPPP) 711
Y. Guglielmi, F. Cappa, S. Gaffet, T. Monfret, J. Virieux, J. Rutqvist & C.F. Tsang
Fuzzy prediction and analysis of landslides 719
Y. He, B. Liu, W.J. Liu, F.Q. Liu & Y.J. Luan
LPC methodology as a tool to create real time cartography of the gravitational hazard: Application
in the municipality of Menton (Maritimes Alps, France) 725
M. Hernandez, T. Lebourg, E. Tric, M. Hernandez & V. Risser
Back-analyses of a large-scale slope model failure caused by a sudden drawdown of water level 731
G.W. Jia, T.L.T. Zhan & Y.M. Chen
Effect of Guangxi Longtan reservoir on the stability of landslide at Badu station of Nankun railway 737
R. Jiang, R. Meng, A. Bai & Y. He
Application of SSRM in stability analysis of subgrade embankments over sloped weak
ground with FLAC3D 741
X. Jiang, Y. Qiu, Y. Wei & J. Ling

IX
Strength parameters from back analysis of slips in two-layer slopes 747
J.-C. Jiang & T. Yamagami
Development characteristics and mechanism of the Lianhua Temple landslide,
Huaxian county, China 755
J.-Y. Wang, M.-S. Zhang, C.-Y. Sun & Z. Rui
Modeling landslide triggering in layered soils 761
R. Keersmaekers, J. Maertens, D. Van Gemert & K. Haelterman
Numerical modeling of debris flow kinematics using discrete element method combined with GIS 769
H. Lan, C.D. Martin & C.H. Zhou
Three dimensional simulation of landslide motion and the determination of geotechnical
parameters 777
Y. Lang, X. Luo & H. Nakamura
Stability analysis and stabilized works of dip bedded rock slopes 783
J.Y. Leng, Z.D. Jing & X.P. Liao
A GIS-supported logistic regression model applied in regional slope stability evaluation 789
X. Li, H. Tang & S. Chen
The stability analysis for FaNai landslide in Lubuge hydropower station 795
K. Li, J. Zhang, S. Zhang & S. He
Numerical analysis of slope stability influenced by varying water conditions in the reservoir
area of the Three Gorges, China 803
S. Li, X. Feng & J.A. Knappett
A numerical study of interaction between rock bolt and rock mass 809
X.P. Li & S.M. He
Macroscopic effects of rock slopes before and after grouting of joint planes 815
H. Lin, P. Cao, J.T. Li & X.L. Jiang
Two- and three-dimensional analysis of a fossil landslide with FLAC 821
X.L. Liu & J.H. Deng
Application of the coupled thin-layer element in forecasting the behaviors of landslide
with weak intercalated layers 827
Y.L. Luo & H. Peng
Numerical modelling of a rock avalanche laboratory experiment in the framework
of the ‘‘Rockslidetec’’ alpine project 835
I. Manzella, M. Pirulli, M. Naaim, J.F. Serratrice & V. Labiouse
Three-dimensional slope stability analysis by means of limit equilibrium method 843
S. Morimasa & K. Miura
Embankment basal stability analysis using shear strength reduction finite element method 851
A. Nakamura, F. Cai & K. Ugai
Back analysis based on SOM-RST system 857
H. Owladeghaffari & H. Aghababaei
Temporal prediction in landslides – Understanding the Saito effect 865
D.N. Petley, D.J. Petley & R.J. Allison
3D landslide run out modelling using the Particle Flow Code PFC3D 873
R. Poisel & A. Preh
Double-row anti-sliding piles: Analysis based on a spatial framework structure 881
T. Qian & H. Tang

X
Centrifuge modeling of rainfall-induced failure process of soil slope 887
J.Y. Qian, A.X. Wang, G. Zhang & J.-M. Zhang
A GIS-based method for predicting the location, magnitude and occurrence time of
landslides using a three-dimensional deterministic model 893
C. Qiu, T. Esaki, Y. Mitani & M. Xie
Application of a rockfall hazard rating system in rock slope cuts along a mountain road of
South Western Saudi Arabia 901
B.H. Sadagah
Model tests of collapse of unsaturated slopes in rainfall 907
N. Sakai & S. Sakajo
Calibration of a rheological model for debris flow hazard mitigation in the Campania region 913
A. Scotto di Santolo & A. Evangelista
Optical fiber sensing technology used in landslide monitoring 921
Y.X. Shi, Q. Zhang & X.W. Meng
Finite element analysis of flow failure of Tailings dam and embankments 927
R. Singh, D. Mitra & D. Roy
Landslide model test to investigate the spreading range of debris according to rainfall intensity 933
Y.S. Song, B.G. Chae, Y.C. Cho & Y.S. Seo
Occurrence mechanism of rockslide at the time of the Chuetsu earthquake in 2004 – A dynamic
response analysis by using a simple cyclic loading model 939
N. Tanaka, S. Abe, A. Wakai, H. Kawabata, M. Genda & H. Yoshimatsu
Analysis for progressive failure of the Senise landslide based on Cosserat continuum model 945
H.X. Tang
Large-scale deformation of the La Clapière landslide and its numerical modelling
(S.-E. de Tinée, France). 951
E. Tric, T. Lebourg & H. Jomard
A novel complex valued neuron model for landslide assessment 957
K. Tyagi, V. Jindal & V. Kumar
Prediction of slope behavior for deforming railway embankments 963
V.V. Vinogradov, Y.K. Frolovsky, A. Al. Zaitsev & I.V. Ivanchenko
Finite element simulation for the collapse of a dip slope during 2004 Mid Niigata Prefecture
earthquake in Japan 971
A. Wakai, K. Ugai, A. Onoue, K. Higuchi & S. Kuroda
Sensitivity of stability parameters for soil slopes: An analysis based on the shear
strength reduction method 979
R. Wang, X.Z. Wang, Q.S. Meng & B. Hu
Back analysis of unsaturated parameters and numerical seepage simulation of the Shuping
landslide in Three Gorges reservoir area 985
S. Wang, H. Zhang, Y. Zhang & J. Zheng
Slope failure criterion: A modification based on strength reduction technique 991
Y.G. Wang, R. Jing, W.Z. Ren & Z.C. Wang
Unsaturated seepage analysis for a reservoir landslide during impounding 999
J.B. Wei, J.H. Deng, L.G. Tham & C.F. Lee
A simple compaction control method for slope construction 1005
L.D. Wesley

XI
Numerical analysis of soil-arch effect of anti-slide piles 1011
Y. Xia, X. Zheng & R. Rui
Determination of the critical slip surface based on stress distributions from FEM 1017
D. Xiao, C. Wu & H. Yang
Effect of drainage facilities using 3D seepage flow analysis reflecting hydro-geological
structure with aspect cracks in a landslide – Example of analysis in OODAIRA Landslide area 1023
M. Yamada & K. Ugai
Back analysis of soil parameters: A case study on monitored displacement of foundation pits 1031
B. Yan, X.T. Peng & X.S. Xu
3D finite element analysis on progressive failure of slope due to rainfall 1035
G.L. Ye, F. Zhang & A. Yashima
Block-group method for rock slope stability analysis 1043
Z. Zhang, Y. Xu & H. Wu
Quantitative study on the classification of unloading zones of high slope 1051
D. Zheng & R.Q. Huang
Investigations on the accuracy of the simplified Bishop method 1055
D.Y. Zhu
Author index 1059

VOLUME 2

Landslide mechanism, monitoring and warning

GIS-based landslide susceptibility mapping in the Three Gorges area – Comparisons of mapping
results obtained by two methods: Analytical hierarchy process and logistic regression 1067
S. Bai, J. Wang, G. Lu, P. Zhou, S. Hou & F. Zhang
Importance of study of creep sliding mechanism to prevention and treatment
of reservoir landslide 1071
J. Bai, S. Lu, J. Han
Stability prediction of landsides before and after impoundment for Lijiaxia hydropower station 1077
J. Bai, S. Lu & J. Han
The technical concept within the Integrative Landslide Early Warning System (ILEWS) 1083
R. Bell, B. Thiebes, T. Glade, R. Becker, H. Kuhlmann, W. Schauerte, S. Burghaus, H. Krummel,
M. Janik & H. Paulsen
The Åknes rockslide: Monitoring, threshold values and early-warning 1089
L.H. Blikra
DInSAR techniques for monitoring slow-moving landslides 1095
D. Calcaterra, M. Ramondini, F. Calò, V. Longobardi, M. Parise, C.M. Galzerano & C. Terranova
Multitemporal DInSAR data and damages to facilities as indicators for the activity of
slow-moving landslides 1103
L. Cascini, S. Ferlisi, D. Peduto, G. Pisciotta, S. Di Nocera & G. Fornaro
The Serre La Voute Landslide (North-West Italy): Results from ten years of monitoring 1111
M. Ceccucci, G. Maranto & G. Mastroviti
Onset of rockslide by the peak-residual strength drop 1119
Q.G. Cheng & G.T. Hu
Analysis of mechanism of the K31 landslide of Changzhi-Jincheng express highway 1127
Y. Cheng

XII
A plane-torsion rockslide with a locked flank: A case study 1133
Q. Cheng
Monitoring of natural thermal strains using hollow cylinder strain cells: The case of a large
rock slope prone to rockfalls 1143
C. Clément, Y. Gunzburger, V. Merrien-Soukatchoff & C. Dünner
Landslide hazards mapping and permafrost slope InSAR monitoring, Mackenzie valley,
Northwest Territories, Canada 1151
R. Couture & S. Riopel
Advanced monitoring criteria for precocious alerting of rainfall-induced flowslides 1157
E. Damiano, L. Olivares, A. Minardo, R. Greco, L. Zeni & L. Picarelli
Investigation of slope failure mechanisms caused by discontinuous large scale geological
structures at the Cadia Hill Open Pit 1165
J. Franz & Y. Cai
Two approaches for public landslide awareness in the United States – U.S. geological survey
warning systems and a landslide film documentary 1173
L.M. Highland & P.L. Gori
Formation and mechanical analysis of Tiantai landslide of Xuanhan county, Sichuan province 1177
R.Q. Huang, S. Zhao & X. Song
Development of wireless sensor node for landslide detection 1183
H.W. Kim
Redox condition and landslide development 1189
Y.H. Lang, S.Y. Liang & G.D. Zheng
Prepa displacement mechanism and its treatment measures for Hancheng landslide 1195
T.F. Li & L.C. Dang
Investigation of the stability of colluvial landslide deposits 1205
X. Li & L.M. Zhang
Choice of surveying methods for landslides monitoring 1211
S.T. Liu & Z.W. Wang
No. 1 landslide on the eastern approach road to ErLang Mountain tunnel: Inference
factors and controlling measures 1217
H.M. Ma & Z.P. Zhang
Estimation of landslide load on multi-tier pile constructions with the help of a combined method 1225
S.I. Matsiy & Ph.N. Derevenets
The use of PSInSAR™ data in landslide detection and monitoring: The example of the
Piemonte region (Northern Italy) 1233
C. Meisina, F. Zucca, D. Notti, A. Colombo, A. Cucchi, G. Savio, C. Giannico & M. Bianchi
Fill slopes: Stability assessment based on monitoring during both heavy rainfall
and earthquake motion 1241
T. Mori, M. Kazama, R. Uzuoka & N. Sento
The mechanism of movement of mud flows in loess soils, successful and unsuccessful
cases of forecast 1247
R.A. Niyazov, Sh.B. Bazarov & A.M. Akhundjanov
Influence of fine soil particles on excess of basal pore-fluid pressure generation in granular
mass flows 1253
Y. Okada & H. Ochiai
An early warning system to predict flowslides in pyroclastic deposits 1259
L. Pagano, G. Rianna, M.C. Zingariello, G. Urciuoli & F. Vinale

XIII
Monitoring and modeling of slope movement on rock cliffs prior to failure 1265
N.J. Rosser & D.N. Petley
Active tectonic control of a large landslide: Results from Panagopoula landslide multi
parametric analyses 1273
S. El Bedoui, T. Lebourg & Y. Guglielmi
A warning system using chemical sensors and telecommunication technologies to
protect railroad operation from landslide disaster 1277
H. Sakai
Distributive monitoring of the slope engineering 1283
B. Shi, H. Sui, D. Zhang, B. Wang, G. Wei & C. Piao
Observational method in the design of high cutting slope around bridge 1289
S. Sun, B. Zhu, B. Zheng & J. Zhang
Ultrasonic monitoring of lab-scaled underwater landslides 1297
Q.H. Truong, C. Lee, H.K. Yoon, Y.H. Eom, J.H. Kim & J.S. Lee
Interaction between landslides and man-made works 1301
G. Urciuoli & L. Picarelli
Desiccation fissuring induced failure mechanisms for clay levees 1309
S. Utili, M. Dyer, M. Redaelli & M. Zielinski
Stability analysis by strength reduction finite element method and monitoring of
unstable slope during reinforcement 1315
Z.Q. Wang, H.F. Li & L.M. Zhang
Displacement monitoring on Shuping landslide in the Three Gorges Dam reservoir area,
China from August 2004 to July 2007 1321
F.W. Wang, G. Wang, Y.M. Zhang, Z.T. Huo, X.M. Peng, K. Araiba & A. Takeuchi
Deformation mechanism and prevention measure for strongly expansive soft-rock slope in
the Yanji basin 1329
X. Wu, N. Xu, H. Tian, Y. Sun & M. He
Twenty years of safety monitoring for the landslide of Hancheng PowerStation 1335
M.J. Wu, Z.C. Li, P.J. Yuan & Y.H. Jiang
A time-spatial deterministic approach to assessment of rainfall-induced shallow landslide 1343
M.W. Xie, C. Qiu & Z.F. Wang
Introduction of web-based remote-monitoring system and its application to landslide
disaster prevention 1349
M. Yamada & S. Tosa
Deformation mechanism for the front slope of the left bank deposits in Xiluodu
hydro-electrical power station, China 1355
M. Yan, Z. Wu, R. Huang, Y. Zhang & S. Wang
Monitoring of soil nailed slopes and dams using innovative technologies 1361
J.-H. Yin, H.-H. Zhu & W. Jin
Application of multi-antenna GPS technique in the stability monitoring of roadside slopes 1367
Q. Zhang, L. Wang, X.Y. Zhang, G.W. Huang, X.L. Ding, W.J. Dai & W.T. Yang

Effects of earthquakes on slopes

Influences of earthquake motion on slopes in a hilly area during the Mid-Niigata
Prefecture Earthquake, 2004 1375
S. Asano & H. Ochiai

XIV
The 1783 Scilla rock-avalanche (Calabria, southern Italy) 1381
F. Bozzano, S. Martino, A. Prestininzi, M. Gaeta, P. Mazzanti & A. Montagna
Self-excitation process due to local seismic amplification and earthquake-induced
reactivations of large landslides 1389
F. Bozzano, S. Martino, G. Scarascia Mugnozza, A. Paciello & L. Lenti
Geological constraints to the urban shape evolution of Ariano Irpino (Avellino province, Italy) 1397
D. Calcaterra, C. Dima & E. Grasso
Landslide zones and their relation with seismoactive fault systems in Azerbaijan, Iran 1405
E. Ghanbari
Ground movements caused by lateral spread during an earthquake 1409
S.C. Hsu, B.L. Chu & C.C. Lin
2-D analysis of slope stability of an infinite slope during earthquake 1415
J. Liu, J. Liu & J. Wang
High-cutting slopes at Qingshuichuan electric power plant in the North of Shaanxi:
Deformation and failure modes and treatment scheme 1421
H. Liu, Z. Liu & Z. Yan
GIS-based real time prediction of Arias intensity and earthquake-induced landslide hazards
in Alborz and Central Iran 1427
M. Mahdavifar, M.K. Jafari & M.R. Zolfaghari
Geomorphology of old earthquake-induced landslides in southeastern Sicily 1433
P.G. Nicoletti & E. Catalano
Coseismic movement of an active landslide resulting from the Mid-Niigata Prefecture
Earthquake, Japan 1439
T. Okamoto, S. Matsuura & S. Asano
Characteristics of large rock avalanches triggered by the November 3, 2002 Denali
Fault earthquake, Alaska, USA 1447
W.H. Schulz, E.L. Harp & R.W. Jibson
FE analysis of performance of the Lower and Upper San Fernando Dams under the 1971
San Fernando earthquake 1455
C. Takahashi, F. Cai & K. Ugai
Reduction of the stability of pre-existing landslides during earthquake 1463
B. Tiwari, I. Dhungana & C.F. Garcia
Probabilistic hazard mapping of earthquake-induced landslides 1469
H.B. Wang, S.R. Wu, G.H. Wang & F.W. Wang
Investigation on stability of landfill slopes in seismically active regions in Central Asia 1475
A.W. Wu, B.G. Tensay, S. Webb, B.T. Doanth, C.M. Ritzkowski, D.Z. Muhidinov & E.M. Anarbaev
Mechanism for loess seismic landslides in Northwest China 1481
L. Yuan, X. Cui, Y. Hu & L. Jiang

Climate, hydrology and landslides

Evaluation of the landslide potential in Chahr Chay dam reservoir slopes 1489
K. Badv & K. Emami
Effect of well pumping on groundwater level and slope stability in the Taiwan
Woo-Wan-Chai landslide area 1493
M. Chang, B.R. Li, Y.S. Zhang, H.S. Wang, Y.H. Chou & H.C. Liu

XV
Case study: Embankment failure of Cable-Ski Lake development in Cairns 1501
K. Chen
Analysis method for slope stability under rainfall action 1507
X.D. Chen, H.X. Guo & E.X. Song
Hydrological modelling of the Vallcebre landslide 1517
J. Corominas, R. Martín & E. Vázquez-Suñé
Landslides in stiff clay slopes along the Adriatic coast (Central Italy) 1525
F. Cotecchia, O. Bottiglieri, L. Monterisi & F. Santaloia
Research on the effect of atomized rain on underground water distribution in Dayantang landslide 1533
J. Ding
Landslide hydrogeological susceptibility in the Crati valley (Italy) 1539
P. Gattinoni
Sustainable landslide stabilisation using deep wells installed with siphon drains
and electro-pneumatic pumps 1547
A. Gillarduzzi
Biological and engineering impacts of climate on slopes – learning from full-scale 1553
S. Glendinning, P.N. Hughes, D.A.B. Hughes, D. Clarke, J. Smethurst, W. Powrie, N. Dixon,
T.A. Dijkstra, D.G. Toll & J. Mendes
Some attributes of road-slope failure caused by typhoons 1559
M.W. Gui, C.H. Chang & S.F. Chen
A small rock avalanche in toppled schist, Lake Wanaka, New Zealand 1565
G.S. Halliday
NRCS-based groundwater level analysis of sloping ground 1571
L.I. Ju, O.T. Suk, M.Y. Il & L.S. Gon
A numerical case study on load developments along soil nails installed in cut slope subjected
to high groundwater table 1575
A.K.L. Kwong & C.F. Lee
Landslides at active construction sites in Hong Kong 1581
T.M.F. Lau, H.W. Sun, H.M. Tsui & K.K.S. Ho
Landslide ‘‘Granice’’ in Zagreb (Croatia) 1587
Z. Mihalinec & Ž. Ortolan
Improvement of subsurface drainage provisions for recompacted soil fill slopes in Hong Kong 1595
K.K. Pang, J.M. Shen, K.K.S. Ho & T.M.F. Lau
Biotechnical slope stabilization and using Spyder Hoe to control steep slope failure 1603
P. Raymond
Rapid landslides threatening roads: Three case histories of risk mitigation in the Umbria
region of Central Italy 1609
D. Salciarini, P. Conversini, E. Martini, P. Tamburi & L. Tortoioli
Assessment of the slope stabilisation measures at the Cadas Pangeran road section,
Sumedang, West Java 1615
D. Sarah, A. Tohari & M.R. Daryono
Analysis of control factors on landslides in the Taiwan area 1621
K. Shou, B. Wu & H. Hsu
Inclined free face riverbank collapse by river scouring 1627
J.C. Sun & G.Q. Wang
Drainage control and slope stability at an open pit mine: A GIS-based hydrological modeling 1633
C. Sunwoo, Y.S. Choi, H.D. Park & Y.B. Jung

XVI
Assessment of regional rainfall-induced landslides using 3S-based hydro-geological model 1639
C.H. Tan, C.Y. Ku, S.Y. Chi, Y.H. Chen, L.Y. Fei, J.F. Lee & T.W. Su
Investigation of a landslide along a natural gas pipeline (Karacabey-Turkey) 1647
T. Topal & M. Akin
Influence of extreme rainfall on the stability of spoil heaps 1653
I. Vanicek & S. Chamra
Behavior of expansive soil slope reinforced with geo-grids 1659
M.Y. Wang, X.N. Gong, M.Y. Wang, J.T. Cai & H. Xu
Geotechnical properties for a rainstorm-triggered landslide in Kisawa village,
Tokushima Prefecture, Japan 1667
G. Wang & A. Suemine
Yigong rock avalanche-flow landslide event, Tibet, China 1675
Q. Xu, S.T. Wang, H.J. Chai, Z.Y. Zhang & S.M. Dong
Key issues of emergency measures and comprehensive remediation projects to control
the Danba landslide, Sichuan province, China 1681
Q. Xu, X.-M. Fan, L.-W. Jiang & P. Liu
Enhanced slope seepage resulting from localized torrential precipitation during a
flood discharge event at the Nuozhadu hydroelectric station 1689
M. Xu, Y. Ma, X.B. Kang & G.P. Lu
An issue in conventional approach for drainage design on slopes in mountainous regions 1697
Z.Q. Yue
Analysis of geo-hazards caused by climate changes 1703
L.M. Zhang

Slope stabilization and protection

Back experience of deep drainage for landslide stabilization through lines of siphon drains and
electro-pneumatics drains: A French railway slope stabilization example 1713
S. Bomont
Experimental geo-synthetic-reinforced segmental wall as bridge abutment 1721
R.M. Faure, D. Rossi, A. Nancey & G. Auray
Rock slope stability analysis for a slope in the vicinity of Take-off Yard of Karun-3 Dam 1725
M. Gharouni-Nik
Stabilization of a large paleo-landslide reactivated because of the works to install
a new ski lift in Formigal skiing resort 1731
J. González-Gallego, J. Moreno Robles, J.L. Garcı́a de la Oliva & F. Pardo de Santayana
A case study on rainfall infiltration effect on the stability of two slopes 1737
M.W. Gui & K.K. Han
Consolidation mechanism of fully grouted anchor bolts 1745
S. He, Y. Wu & X. Li
Stability analysis for cut slopes reinforced by an earth retention system by considering
the reinforcement stages 1751
W.P. Hong, Y.S. Song & T.H. Kim
Landslide stabilization for residential development 1757
I. Jworchan, A. O’Brien & E. Rizakalla
Influence of load transfer on anchored slope stability 1763
S.K. Kim, N.K. Kim, Y.S. Joo, J.S. Park, T.H. Kim & K.S. Cha

XVII
Review of slope surface drainage with reference to landslide studies and current practice
in Hong Kong 1769
T.M.F. Lau, H.W. Sun, T.H.H. Hui & K.K.S. Ho
Analysis of dynamic stability about prestressed anchor retaining structure 1775
H. Li, X. Yang, H. Liu & L. Du
Safety analysis of high engineering slopes along the west approach road of ZheGu
mountain tunnel 1781
T.B. Li, Y. Du & X.B. Wang
Landslide stabilizing piles: A design based on the results of slope failure back analysis 1787
M.E. Popescu & V.R. Schaefer
Landslides on the left abutment and engineering measures for Manwan hydropower project 1795
X. Tang & Q. Gao
Factors resulting in the instability of a 57.5 m high cut slope 1799
J.J. Wang, H.J. Chai, H.P. Li & J.G. Zhu
Orthogonal analysis and applications on anchorage parameters of rock slopes 1805
E.C. Yan, H.G. Li, M.J. Lv & D.L. Li
Waste rock dump slope stability for a gold mine in California 1811
H. Yang, G.C. Rollins & M. Kim
Properties of the high rock slope of Hongjiadu hydropower project and its engineering
treatment measures 1817
Z. Yang, W. Xiao & D. Cai
Typical harbor bank slopes in the Three Gorges reservoir: Landslide and collapse and their
stability control 1825
A. Yao, C. Heng, Z. Zhang & R. Xiang
Weighting predisposing factors for shallow slides susceptibility assessment
on the regional scale 1831
J.L. Zêzere, S.C. Oliveira, R.A.C. Garcia & E. Reis
Analyses of mechanism of landslides in Tongchuan-Huangling highway 1839
L. Zhang & H. He
Treatments of Loess-Bedrock landslides at Chuankou in Tongchuan-Huangling expressway 1847
J.B. Zhao
Types, characteristics and application conditions of anti-slide retaining structures 1855
J. Zheng & G. Wang
The stabilization of the huge alluvial deposit on the left bank and the high rock slope on
the right bank of the XiaoWan Hydropower Project 1863
L. Zou, X. Tang, H. Feng, G. Wang & H. Xu

Risk assessment and management

Malaysian National Slope Master Plan – Challenges to producing an effective plan 1873
C.H. Abdullah & A. Mohamed
Spatial landslide risk assessment in Guantánamo Province, Cuba 1879
E. Castellanos Abella & C.J. van Westen
Landslide risk management: Experiences in the metropolitan area
of Recife – Pernambuco, Brazil 1887
A.P. Nunes Bandeira, & R. Quental Coutinho

XVIII
Societal risk due to landslides in the Campania region (Southern Italy) 1893
L. Cascini, S. Ferlisi & E. Vitolo
Landslide risk in the San Francisco Bay region 1899
J.A. Coe & R.A. Crovelli
A first attempt to extend a subaerial landslide susceptibility analysis to submerged slopes:
The case of the Albano Lake (Rome, Italy) 1905
G.B. Fasani, C. Esposito, F. Bozzano, P. Mazzanti & M. Floris
Landslide susceptibility zonation of the Qazvin-Rasht-Anzali railway track, North Iran 1911
H. Hassani & M. Ghazanfari
Assessment of landslide hazard of a cut-slope using linear regression analysis 1919
S. Jamaludin, B.B.K. Huat & H. Omar
Global monitoring strategy applied to ground failure hazards 1925
E. Klein, C. Nadim, P. Bigarré & C. Dünner
Regional slope stability zonation based on the factor overlapping method 1933
J.F. Liu, G.Q. Ou, Y. You & J.F. Lui
Landslide hazard and risk assessment in the areas of dams and reservoirs of Serbia 1939
P. Lokin & B. Abolmasov
The evaluation of failure probability for rock slope based on fuzzy set theory
and Monte Carlo simulation 1943
H.J. Park, J.G. Um & I. Woo
Macro-zoning of areas susceptible to flowslide in pyroclastic soils in the Campania region 1951
L. Picarelli, A. Santo, G. Di Crescenzo & L. Olivares
Zoning methods for landslide hazard degree 1959
J. Qiao & L.L. Shi
A proposal for a reliability rating system for fluvial flood defence embankments in
the United Kingdom 1965
M. Redaelli, S. Utili & M. Dyer
Simplified risk analysis chart to prevent slope failure of highway embankment on soft
Bangkok clays 1971
A. Sawatparnich & J. Sunitsakul
Determining landslide susceptibility along natural gas pipelines in Northwest Oregon, USA 1979
J.I. Theule, S.F. Burns & H.J. Meyer
Landslide susceptibility assessment using fuzzy logic 1985
Z. Wang, D. Li & Q. Cheng
Prediction of the spatiotemporal distribution of landslides: Integrated landslide susceptibility
zoning techniques and real-time satellite rainfall 1991
H. Yang, R.F. Adler, G.J. Huffman & D. Bach
Entropy based typical landslide hazard degree assessment in Three Gorges 1995
Z. Yang & J. Qiao
The optimal hydraulic cross-section design of the ‘‘Trapezoid-V’’ shaped drainage canal
of debris flow 2001
Y. You, H.L. Pan, J.F. Liu, G.Q. Ou & H.L. Pan
Practice of establishing China’s Geo-Hazard Survey Information System 2005
K. Zhang, Y. Yin & H. Chen
A XML-supported database for landslides and engineered slopes related to China’s water
resources development 2011
Y. Zhao & Z. Chen

XIX
Landslide and engineered slopes in China
Failure and treatment technique of a canal in expansive soil in South to North
Water Diversion project 2019
Y.J. Cai, X.R. Xie, L. Luo, S.F. Chen & M. Zhao
High slope engineering for Three Gorges ship locks 2027
G.J. Cao & H.B. Zhu
Large-scale landslides in China: Case studies 2037
R.Q. Huang
Early warning for Geo-Hazards based on the weather condition in China 2055
C.Z. Liu, Y.H. Liu, M.S. Wen, C. Tang, T.F. Li & J.F. Lian
Slope engineering in railway and highway construction in China 2061
G. Wang, H. Ma, M. Feng & Y. Wang
Mining slope engineering in China 2075
S. Wang, Q. Gao & S. Sun
Structure and failure patterns of engineered slopes at the Three Gorges reservoir 2089
Y.P. Yin
Slope engineering in hydropower projects in China 2101
J.P. Zhou & G.F. Chen
A thunder at the beginning of the 21st century – The giant Yigong Landslide 2111
Z.H. Wang
Author index 2119

XX
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Preface

The city of Xi’an, China is privileged to have the honor of hosting the 10th International Symposium on
Landslides and Engineered Slopes, following its predecessors: Rio de Janeiro, Brazil, 2004; Cardiff, U.K.,
2000; Trondheim, Norway, 1996; Christchurch, New Zealand, 1992; Lausanne, Switzerland, 1988; Toronto,
Canada, 1984; New Delhi, India, 1980; Tokyo, Japan, 1977; and Kyoto, Japan, 1972.
China is one of the countries in the world that suffer severely from landslide hazards. Statistics have shown
that every year 700 to 900 people are killed by landslides. With the large scale infrastructure construction,
failures of engineered slopes are increasing, and have become a serious concern of the government, various
enterprises and technical societies. The Chinese geological and geotechnical communities look forward to this
unique opportunity of exchanging and sharing technical know-how and experience of combating landslides
disasters with our international peers.
Xi’an is a historical city of China. It has been the capital for China’s twelve dynasties, spanning over
1200 years, and was a starting point of the famous Silk Road. It is a nice place for participants from all
over the world to meet and review our past experience, and in the meanwhile, look forward to a productive future
against landslides and slope failures.
From this book, readers will find that there are 7 Symposium Themes, as allocated by the Steering Committee,
including 13 keynote and special papers. At the JTC1 meeting held at the 9th Symposium, it was decided that
a special session entitled ‘China Afternoon’ would be organized, whose papers are arranged under a separate
theme ’Landslides and Engineered Slopes in China’ in this Proceedings.
This symposium was jointly organized by the Chinese Institution of Soil Mechanics and Geotechnical
Engineering-CCES, Chinese National Commission on Engineering Geology, Chinese Society of Rock Mechan-
ics and Engineering, and the Geotechnical Division of the Hong Kong Institution of Engineers. The Organizing
Committee is grateful to the reviewers for reviewing more than 300 submitted papers. It is practically impossible
to list the large number of these volunteers here. However, their contributions must be fully acknowledged,
without which the quality of this book would not have been maintained.
Special thanks also go to the members of JTC1 for their constant attention and useful comments during the
preparation of this symposium.
Zuyu Chen

XXI
Keynote lectures
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Landslides: Seeing the ground

Norbert R. Morgenstern & C. Derek Martin

Dept. Civil & Environmental Engineering, University of Alberta, Edmonton, Canada

ABSTRACT: Landslide engineering requires the consideration of a number of complex processes ranging from
geological and hydrogeological characterization to geomechanical characterization, analyses and risk manage-
ment. This paper concentrates on recent advances that improve site characterization applied to landslide problems.
It presents the view that one of the most exciting developments is the growing potential for application of Geo-
graphical Information Systems (GIS) and that making GIS goetechnically smart is a transformative development.
Examples are given of integrating remote sensing data in GIS to improve visualization, mapping and movement
characterization. Application of analysis of rockfall within GIS and complex slope stability evaluation with the
aid of GIS are presented to illustrate recent developments and provide direction for future enhancements.

1 INTRODUCTION • geotechnical site characterization

• the geotechnical model, seepage, stability and
A landslide, whether it occurs in a natural or an engi- deformation analyses
neered slope, is a complex process. When Laurits • risk assessment and risk mitigation
Bjerrum, at the end of his Terzaghi Lecture (Bjerrrum,
1967), reminded us of the recognition in Japan of ‘‘a There has been very substantial progress in all of
landslide devil who seems to laugh at human incom- these areas in recent years. New tools are applied to site
petence’’, he was reminding us of the complexity of characterization. The range of geomechanical mod-
the landslide process. els that can be usefully applied in practice has grown
Managing complexity invariably requires simpli- substantially. The capacity to analyse often exceeds
fication into a Process Model. A Process Model the capacity to characterize. Risk assessment and
captures the essentials required to meet the objec- management of slopes is maturing quickly as a valu-
tives of using the model, without including details able tool for dealing with landslides both locally and
that are extraneous to these objectives. In geotechnical regionally. Yet much uncertainty persists in geotech-
engineering these objectives can range from ensuring nical practice. The intrinsic presence of uncertainty
that an engineered structure will perform as intended in geotechnical practice was emphasized by Morgen-
to managing the risk associated with natural hazards stern (2000) who provided numerous examples of
over a larger scale. Establishing the appropriate pro- unanticipated behaviour of geotechnically engineered
cess is both site and project dependent. It underpins facilities, often with unfavourable results.
the value associated with the practice of geotechnical In developing the theme for this paper, we have
engineering. drawn on our experience to conclude that the great-
Understanding the landslide process and being able est uncertainties in the process modeling of landslides
to simplify it effectively calls on interpreting a number arise from inadequacies in site characterization, in
of contributory processes and activities. The main ones the broadest sense, and therefore we concentrate on a
are as follows: discussion of recent advances that improve site chara-
cterization applied to landslide problems.
• geomorphology—the multiplicity of physical and
chemical processes that have affected the surface
2 VIEWING THE GROUND SURFACE
and near-surface of the site
• hydrology—the role of surface water in infiltration,
2.1 Geographical Information Systems (GIS)
erosion, etc.
• geology—the sequence and characteristics of the It is our view that one of the most exciting devel-
soils and rocks opments for landslide engineering is the growing
• hydrogeology—the factors affecting the groundwa- potential for application of GIS. The power of GIS
ter distribution is that it enables us to ask questions of a database,

3
perform spatial operations on databases and generate
graphic output that would be laborious or impossible
to do manually. Rhind (1992) observes that a GIS can
answer five generic questions:

Question Type of Task

1. What is at . . . ..? Inventory

2. Where is. . ...? Monitoring
3. What has changed since. . .? Inventory and
monitoring
4. What spatial pattern exists..? Spatial analysis
5. What if. . ...? Modelling

The first three questions are simple queries, while the

last two are more analytical.
GIS on its own adds enormously to our capacity
to see and interpret surface geospatial information
which is essential for landslide engineering. A first
fly-by experience in GIS soon convinces the landslide
engineer of its potential. However, GIS has limita-
tions in presenting three-dimensional (3D) geologic
and geotechnical data since it was originally devel-
oped to deal with two dimensional plane problems.
Some GIS systems, like ArcGIS, provide a functional
developer kit which can be used to develop the 3D
capability for geotechnical engineering problems. As
pointed out by Lan & Martin (2007), the many current
developments in 3D GIS are still not sufficient to meet
the needs of the geotechnical engineer. Mining soft- Figure 1. The architecture for an integrated system. It
is composed of three different stages which required the
ware such as Surpac Vision provides a comprehensive
implementation of specific tasks.
system for geological modeling, but not geotechni-
cal modeling. For example, while three-dimensional
solid modeling and two-dimensional sections can be
easily created in Surpac, querying inclinometer or the landslide engineer for many years, although often
piezometer data is not easily accomplished. However neglected in many geotechnical curricula. The latter
an integrated system, which is illustrated in Figure 1, has usually been the preserve of specialists.
can be developed. To recognize landslides, API relies on characteristic
Even within Stage 1, limited ground behaviour can morphology, vegetation and drainage. Parise (2003)
be modelled. The aim of Stage 2 is to establish a provides an example of how diagnostic surface fea-
comprehensive ground model of the site. Stage 3 tures can be related to certain types of movement, the
links geotechnical numerical analysis tools to conduct degree of activity and the depth of movement. The
geotechnical analyses and assist in decision-making. study of sequential photographs can provide informa-
Making GIS geotechnically smart is a transformative tion on the progressive evolution of landslides and
development for geotechnical engineering. Examples can lead to a better understanding of their causes
to illustrate this will follow in subsequent sections of (Chandler & Brunsden, 1995; Van Westen & Getahun,
the paper which will return to demonstrate the role of 2003). GIS facilitates the application of API, the
GIS in a number of slope related problems. archiving of the photos and the production of geomor-
phological maps that arise from the interpretation.
While the application of API is common, more
2.2 Aerial and terrestrial photographs
quantitative studies are rare, probably due to limited
Aerial photographs are a well established resource availability of good quality photographs, adequately
for landslide studies and this is well-understood fixed control points and cost. Modern photogrammet-
(Soeters & Van Westen 1996). Aerial photographs can ric software has been developed that should encourage
be used for interpretation (API) for qualitative anal- greater use of photogrammetry for the construction
ysis and photogrammetry for extracting quantitative of high quality digital elevation models (DEM). Dif-
information. The former has been an essential tool for ferential DEMs will quantify landslide movements.

4
 slopes are bare of vegetation and, under good condi-
tions, can produce DEM’s that are comparable to those
produced by terrestrial based LiDAR systems (Martin
et al. 2007).

2.3 Satellite sensing

Van Westen (2007) has recently summarized the use of
remote sensing imagery in creating landslide invento-
ries and notes that medium resolution satellite imagery
such as LANDSAT, SPOT, ASTER, IRIS-D etc. are
used routinely to create landuse maps and invento-
ries of landslides. At a broader scale, areas of global
landslide susceptibility have been determined by cor-
relating information on land surface features from the
NASA Shuttle Radar Topography Mission (SRTM)
which can resolve features up to 30 m in size. Cor-
relations with other satellite-based information, such
as that related to precipitation, results in a global
landslide susceptibility map that is broadly supported
empirically (Hong, Adler & Huffman 2007). While
technically fascinating, the restricted resolution of the
DEM and the lack of sub-surface information limits
the value of this very remote sensing information for
the landslide engineer.
While optical images with resolutions larger than
3 m (see medium resolution list above) have proven
Figure 2. Flowchart of the working procedures used by useful for interpretation of landslides in some indi-
Walstra et al (2007) for using historical aerial photographs vidual cases (Singhroy 2005), Van Westen (op. cit)
in landslide assessment.
advises that very high resolution imagery (QuickBird,
IKONOS, CARTOSAT-1, etc.) has become the best
option for landslide mapping from satellite imagery
and the number of earth observation satellites with
Walstra, Dixon & Chandler (2007) provide a flow stereo capabilities and resolution of 3 m or better is
chart for this process, reproduced here as Figure 2, and increasing rapidly.
then illustrate its application to the Mam Tor landslide, Van Westen (op cit) also draws attention to the
looking back over photographs of adequate quality to high resolution imagery that is available within Google
1953, and to the East Penwyn landslide over a com- Earth and notes that the 3-D capabilities and zooming
parable period. The need for accurate orthophotos are functions that are available in Google Earth, together
common to both the studies discussed in this paper with the possibility of drawing polygons on the image,
as well as any spatial quantification and referenc- greatly facilitates the interpretation and mapping of
ing within GIS. As the authors observe, the two case slopes and landslides. These images can be transported
histories have demonstrated the value of the aerial pho- into GIS for storage and mensuration.
tographic archive to extract spatial data necessary for
assessing landslide dynamics.
2.4 LiDAR technology
Terrestrial-based digital photography is also gain-
ing acceptance as an efficient method for creating LiDAR (Light Detection and Ranging) has become
DEMs as well as capturing geological structure of such an essential tool in the mapping and character-
rock slopes (Pötsch et al. 2007). When designing izing of landslides, that it is becoming difficult to
remedial measures, the DEM’s can be used to pro- imagine assessing all but the most local problems with-
duce sections as well as estimate volumes. Findley out it. Airborne LiDAR uses a powerful laser to map
(2007) describes the application of Sirovision tech- the ground surface in swathes. The literature on the
nology using terrestrial-based digital photography to application of LiDAR to landslide studies is growing
characterise a rock slope along Interstate 90 USA. He rapidly and the procedures are becoming almost rou-
notes that the while the technology saves time and tine in professional practice in some countries. In our
effort it is not a substitute for field mapping. It should view, the definitive studies published so far are those
also be noted that the technique works best when the carried out by the U.S. Geological Survey to assess

5
landslide susceptibility in Seattle, Washington (Schulz earth DEM’s. Examples of a bare earth DEM with
2004, Schulz 2007). a 1 m resolution applied to landslide studies and a
In this case, airborne laser pulses were uniformly bare-earth DEM with 25–50 cm resolution applied to
spaced within a 600 m wide swath with an average faulting studies are citied by Carter et al. (2007).
pulse density of 1/m2 . Up to four laser returns were Airborne LiDAR is increasingly being applied to
collected for each pulse resulting in a vertical pro- map landslides and contribute to infrastructure loca-
file of ground features for each pulse location. Each tions such as pipelines and to develop landslide maps
pulse generates multiple returns due to reflections which contribute to risk analysis. The ability to pene-
from features such as powerlines, buildings, trees, trate forest cover, even with reduced DEM accuracy, is
undergrowth and the ground surface. Simultaneous of enormous value. The increased accuracy of DEM’s
acquisitions of aircraft position and laser direction foresees the increasing use of differential LiDAR to
located laser returns with absolute vertical and hor- measure ground deformations with time.
izontal accuracy of 15 cm and less than 1 m, respec- As an illustration of the use of LiDAR in current
tively (Schulz 2004). Swathes are stitched together into practice, Figure 3 shows the bare earth projection
a seemless DEM during processing. side by side with conventional aerial photography
All ground features that produce returns are repre- of a potential pipeline crossing of a river in central
sented in the laser survey, including buildings, trees British Columbia. While most crossings are by hor-
and boulders. One of the most valuable develop- izontal directional drilling, design for conventional
ments is that the trees can be stripped away because crossings by excavation methods are needed as a
some pulses penetrate the tree canopy and others stand-by. Hence landslide identification is an impor-
are reflected off the forest floors. The latter can be tant consideration in route selection. In this case
separated from reflections from the trees to produce the river has downcut into a deep deposit (∼200 m)
bald-earth DEM’s. This processing for deforestation of glacio-lacustrine clays. The contrast between the
is a remarkable contribution but as pointed out by information revealed by the bare earth imagery and
Haugerud & Harding (2001), there are some limi- conventional airphotos is striking.
tations in the algorithms that need to be recognized While airborne LiDAR is commonly used for larger
in interpreting the bare-earth DEM’s. The technique scale studies, it is also of value to enhance detailed
has even been applied to faulting studies in high- geological studies at specific sites. Jaboyedoff et al.
relief Alpine landscapes, with spectacular results (2007) has suggested that the high resolution LiDAR
(Cunningham et al. 2006). DEM can be used to extract both regional and local
In the case of the Seattle bare-earth DEM, the scale geological structures. The advantages of such
vertical accuracy is typically about 30 cm, but is sig- techniques are obvious when dealing with steep moun-
nificantly less in areas of high vegetation. The data in tainous terrain. Figure 4 shows a portion of the famous
the DEM have a grid cell size of 1.8 m. This DEM Turtle Mountain-Frank Slide in Canada and the use of
was entered into a GIS to produce a landslide map shading relief of a high resolution LiDAR DEM to por-
using derivatives of the DEM such as shaded relief tray the extent of tension cracks that still exist beyond
maps (hill shades), a slope map, a topographic contour the scarp of the slide (Sturznegger et al. 2007). The
map and numerous ground surface profiles with a 2 m potential instability associated with these cracks is a
contour interval. This was supplemented by historical matter of concern.
information and ground mapping.
The strength and weaknesses of LiDAR mapping
are discussed in detail by Schulz (2007). It is of inter-
est to note his conclusions that aerial photographs
appeared to be more effective than LiDAR in the
Seattle area for discerning boundaries of recently
active landslides within landslide complexes. The res-
olution of the LiDAR data appeared inadequate to
resolve landslide boundaries within landslide com-
plexes. However, LiDAR was much more effective
for identifying presumably older landslides and the
boundaries of complexes in which recently active
landslides occurred. Another recent example illustrat-
ing the value of high resolution DEM’s provided by
LiDAR for mapping landslides has been provided by
Ardizzone et al. (2007).
Improvements in LiDAR technology, including pro- Figure 3. Example of a LiDAR bare-earth projection com-
cessing, are rapidly leading to even more accurate bare pared to the conventional aerial photograph.

6
 broader Engineering Geology or Geomorphology to
Geotechnical Engineering.
The development of a landslide map, an essential
for any hazard on risk assessment tool, relies primarily
on the visualization techniques summarized previ-
ously. Aerial photo interpretation remains widely used
and is increasingly enhanced by LiDAR imagery. GIS
and image processing software facilitate the process.
Soeters & Van Westen (1996) have summarized the
geomorphic features that are diagnostic of landslides
both recent and relict.
The assessment of landslide susceptibility goes
beyond the cataloguing of past and current landslides
by including areas that are susceptible to sliding.
Ideally a susceptibility assessment is based on field
Figure 4. Structural and tension crack mapping of the Frank reconnaissance to determine factors contributing to
Slide using LiDAR DEM. instability, utilizing the landslide inventory as a first
step. Landslide susceptibility maps have been pub-
lished for many decades (e.g., Radbruch & Crowther
In addition to airborne LiDAR, terrestrial-based 1973). However the ability to manipulate geomor-
LiDAR is also finding applications in slope stability phic data within GIS has proliferated the number of
studies. Examples of rock slope assessment, where landslide susceptibility studies and their associated
LiDAR has been used to evaluate rock structure are methodology. Even prior to the use of GIS based
given by Kemeny et al. (2006) & Martin et al. (2007). techniques, relative landslide susceptibility in terms
Using terrestrial based LiDAR portable scanners that of simple bivariate analyses or more complex multi-
can operate in the range of 50 m to 800 m greatly variate analyses had been developed. Early zonation
enhance our ability to map the slope discontinuities. methods based on these developments have been dis-
Sturzenegger et al. (2007) combined both airborne and cussed by Varnes (1984). More recent GIS-based
terrestrial-based LiDAR to map the structural features developments are listed in Chacón et al (op cit). Some
associated with Frank Slide (Fig. 4). Terrestrial-based highlights cited are:
LiDAR is also being used for direct monitoring of the • Franks et al. (1998) prepared detailed 1:1000 the-
process of hard rock coastal cliff erosion (Rosser et al. matic maps in GIS for landslide hazards on Hong
2005). Kong Island, based on a very rich database.
• Wachal & Huduk (2000) used GIS to assess lands-
liding in a 1,500–2,000 km2 area in the USA based
3 GIS AND LANDSLIDE SUSCEPTIBILITY on four factors—slope angle, geology, vegetation
and distance to faults.
As stated by Van Westen (op cit), GIS has determined, • Dai & Lee (2004) developed probabilistic measures
to a large degree, the current state of the art in landslide of landslide susceptibility for Lantau Island using
hazard and risk assessment, particularly for landslide multivariate logistic regression of presence-absence
studies that cover large areas. Chacón et al. (2006) have of dependent variables relating landslides and con-
recently conducted a comprehensive general review of tributing factors such as lithology, slope angle, slope
GIS landslide mapping techniques and basic concepts aspect, elevation, soil cover, and distance to stream
of landslide mapping. From this extensive investiga- channels.
tion they identify three main groups of maps that have
There is a tendency to incorporate increasingly
been propagated by means of GIS:
complex statistical methods in these landslide suscep-
1. Spatial incidence of landslides tibility analyses. Spatial validation is essential for any
2. Spatial-temporal incidence and forecasting of land- practical application.
slides (hazard susceptibility) Temporal considerations most commonly enter into
3. Consequence of landslides. landslide susceptibility forecasting by coupling rain-
fall probability assessment as an important triggering
Regional studies might characteristically have factor (Lan et al. 2005). This can be undertaken
scales of 1:50,000 and smaller, while site spe- empirically or on a more process-based consideration.
cific studies will have larger scales ranging from The work of Mejia-Navarvo et al (1994) provides an
1:1000–1:25,000 depending on the project. At these example of the former while that of Dietrich et al.
larger scales one is characteristically merging from (1995) is an early example of the latter.

7
 The inclusion of geomechanical and hydro- slowly been incorporated into geo-engineering prac-
logical process considerations within GIS based tice for mapping the rates and extents of ground defor-
modeling and landslide hazard analysis marks a mations associated with landslides. As the potential
convergence between the techniques for regional- applications and limitations of this tool are gradually
based studies developed by Engineering Geom- being understood, the range of terrains and situations
orphologists and Geologists and the inputs of the to which it may be applied are expanding. The strength
Geotechnical Engineer. Here the example offered by of this technique is that either available archives of data
Delmonaco et al. (2003) is of interest. In this case can be utilized to better understand historical move-
the infinite slope analysis was applied at a river basin ments or new data can be acquired for go-forward
scale with basin scale characterization of all of the monitoring for large areas (up to 2500 km2 ) using
inputs to this classical equation. In order to calcu- a remote platform that can acquire data at night or
late the likely pore pressure development, Green-Ampt through clouds.
infiltration analyses were also carried out over the Synthetic aperture radar (SAR) is an active sen-
basin scale, reflecting the variation of rainfall with sor that can be used to measure the distance between
different return periods. The relation between poten- the sensor and a point on the earth’s surface.
tial instability and return period was determined and A SAR satellite typically orbits the earth at an alti-
the predicted scenarios of instability were found to tude of approximately 800 km. The satellite constantly
correspond sensibly with observations made after an emits electromagnetic radiation to the earth’s sur-
extreme rainfall event in 1966. Examples like this face in the form of a sine wave, which reflects off
encourage the integration of process-based consid- the earth’s surface and returns back to the satel-
erations into GIS-based hazard and risk analyses. lite. The back-scattered microwave signal is used to
The coupling of landslide susceptibility forecasts with create a SAR satellite image (a black and white rep-
earthquake effects have already been investigated in a resentation of ground reflectivity) using SAR signal
GIS environment (Refice & Capolongo 2002). processing methods. SAR radar images are made
The centrality of GIS-based processing has greatly up of pixels, with the specific size influenced by
advanced regional landslide hazard and risk analy- the SAR sensor resolution; the higher the resolu-
sis as summarized by Chacón et al. (2006). There tion the smaller the pixel size. To measure dif-
has been some convergence between the tools used ferential ground movements over a specified time
in regional studies and those used by geotechnical period, InSAR requires two SAR images of the same
engineers in more site specific problems. As stated area taken from the same flight path, within typ-
in Section 2, current GIS technology has signifi- ically 500 m laterally. During InSAR processing
cant limitations in truly three-dimensional problems. the phase of the corresponding pixels of both images
Chacón et al. (2006) concluded that ‘‘the use of are subtracted. The phase difference between the two
three-dimensional GIS for large scale, detailed haz- SAR images can be used to determine the ground
ard or risk maps will be one of the significant movement in the line-of-sight of the satellite.
developments in the near future’’. This, applied to Froese et al. (2004) discussed some of the
landslide engineering, is the fundamental theme of limitations and applications of differential InSAR
this paper. Günther et al (2004) illustrate the kind (D-InSAR) in mapping ground deformations associ-
of progress that is being made in their extension ated with landslides. These included data availability,
to GIS, designated RSS-GIS, that incorporates the rate of motion, direction of movement, steep slope
deterministic evaluation of rock slope stability and is distortions and loss of coherence due to a variety of
particularly useful for regional stability assessment. factors such as vegetation, ground moisture and atmo-
It incorporates grid-based data on rock structures, spheric effects. Therefore the potential application of
kinematic analyses for hard rock failure modes, some InSAR to landslide mapping and monitoring requires
pore pressure effects and stability evaluations on a consideration on a case-by-case basis to determine the
pixel basis. Other examples of integration of geotech- suitability of this method to a particular set of site con-
nical considerations with GIS follow later in this paper, ditions. Over the past few years a number of advances
see Section 5, 6 and 7. have lead to an increased reliability of InSAR for mea-
suring ground motion in an ever increasing number of
ground conditions.
PS-InSAR: In the last fifteen years, the avail-
able number of spaceborne SAR sensors (ERS 1/2,
4 MAPPING GROUND MOVEMENT
Radarsat 1, JERS, ALOS), has increased signifi-
cantly. The capability of InSAR has been consider-
4.1 InSar
ably improved by using large stacks of SAR images
Since the late 1990’s the application of spaceborne acquired over the same area, instead of the classi-
Interferometric Synthetic Aperture Radar (InSAR) has cal two images used in the standard configurations.

8
This multi-image InSAR technique was introduced stabilizing the existing road versus a re-route away
as Permanent/Persistent Scatterer Interferometric Syn- from the area of most significant instability. As there
thetic Aperture Radar (PS-InSAR) (Ferretti et al., is limited point source geotechnical instrumentation
1999, 2000, 2001). With these advances the InSAR available in the areas that are easily accessible from
techniques are becoming more and more quantitative the highway, larger portions of the valley slope do not
geodetic tools for deformation monitoring, rather than have quantitative monitoring information.
simple qualitative tools. Numerous recent projects in In the fall of 2006, a series of 18 corner reflectors
Europe (Farina et al. 2006, Colesanti & Waskowkski, were installed on both the southwest and northeast
2006, Meisina et al. 2007) have shown good corre- valley walls in order to characterize the differential
lation between results obtained from PS-InSAR and movements of the various portions of each valley walls
traditional geotechnical instrumentation in urban areas (Figure 5). Between November 2006 and November
impacted by landslide movements. 2007, scenes of Radarsat-1 ascending F2N scenes
CR-InSAR: While the PS-InSAR technique is ide- were obtained and processed by the Canadian Cen-
ally suited to urban environments where buildings tre for Remote sensing using IPTA software (Froese
can be used as artificial reflectors or where suit- et al. 2008). The preliminary results available at the
able natural exposures exist, the application of this time of the preparation of this paper indicate that for
technique is more limited in northern boreal regions the reflectors that are situated on landslide blocks
with sparse development and more dense vegeta- moving with the line-of-sight of the satellite, the
tion cover. It is often in these remote northern areas movements observed from the CR-InSAR are greater
where large slowly moving landslides whose size than those found over the same time period as those
and rates of deformation are ideally suited for the observed on conventional slope inclinometers. As
InSAR technology are located. In order to overcome these slides are moving in colluvium, likely with a
the issues associated with loss of coherence in veg- rotational component, the CR-InSAR results may be
etated and moist ground conditions the introduction more representative of the actual deformations that
of artificial, phase stable reflectors is emerging. This are only represented in the horizontal plane by slope
technique has been called either Corner Reflector inclinometers. Evaluation of this data is currently on
InSAR (CR-InSAR) or Interferometric Point Target going.
Analysis (IPTA). One of the first documented case Future Development: While the available resolution
histories of the use of artificial reflectors for moni- of the SAR sensors and the number of satellites has in
toring of landslides was by Rizkalla & Randall (1999) the past been a limitation to the technique, the launch-
where five corner reflectors were installed on the Sim- ing of new, higher resolution satellites provides the
monette River pipeline crossing as a trial to monitor opportunities to overcome some of these limitations.
slope movements. More recently Petrobras has uti- With the launch of Radarsat-2 in December 2007, the
lized this technique along a pipeline crossing in Brazil ability of the satellite to look both right and left and
(McCardle et al. 2007). Both of these applications obtain 3 m pixel resolution data will likely increase the
have focused on the application of artificial reflectors directions of slope movements that can be measured
along linear corridors in vegetated terrain. Perhaps the and increase the amount of data that can be obtained.
most complex landslide monitoring attempted utiliz-
ing CR-InSAR is for the Little Smoky River crossing
of Highway 49 in northern Alberta, Canada. The appli-
cation of D-InSAR to this site was first attempted in
2003 (Froese et al. 2004) but the heavy vegetation and
ground moisture conditions limited the success of this
application.
Both valley walls at the Highway 49 crossing of
the Little Smoky River are subject to ongoing move-
ments of deep seated, retrogressive slides in glacial
materials and bedrock. The movements of each valley
wall are very complex as there are a variety of zones
of movement that differ in aspect and level of activ-
ity based on their proximity to the present day river.
Since the completion of the bridge across the Little
Smoky in 1957, there have been significant ongoing
maintenance issues due to slope instability impacting
on the highway. In order to provide a more stable long Figure 5. Layout of the corner reflector array in relation to
term solution to mitigate the impacts of slope move- recently installed instrumentation and profile locations (from
ments on the highway, options were considered for Froese et al. 2008).

9
The introduction of 1 m pixel resolution available from
the recently launched TerraSAR-X will also continue
to increase the density of data that is available for target
detection and monitoring.
As the quality and density of this data improves,
three dimensional deformation information for land-
slides may become a reality. Recent studies by Farina
et al. (2007) for the Ciro Marina village in Calabria,
Italy have shown the potential for using data from
different radar platforms to estimate the geometry of
movement patterns, an essential step for defining the
geometry of the three dimensional nature of the rupture
surface.
Figure 6. Example of rock fall hazard along a section of
railway in British Columbia.
4.2 Surface Radar (SSR)
The application of differential interferometry using rock fall analyses at regular intervals and often the
synethic aperature radar has been recently applied to hazards come from inaccessible natural rock slopes
the monitoring of rock slopes. The technique is called well upslope of the track with previously undeter-
Slope Stability Radar (SSR) and instead of using syn- minable flow paths (Figure 6). GIS has been used as
thetic aperature radar from a moving radar platform, an effective tool in hazard delineation, but seldom is
the SSR uses a real-aperture on a stationary plat- GIS used for rock fall process modeling (Dorren &
form positioned 50 to 1000 metres away from the Seijmonsbergen 2003). Stand alone computer soft-
foot of the slope (Harries & Roberts 2007). A major ware to assess rock falls have been developed to
advantage of the technique is that it provides full cov- analyze trajectories, run-out distance, kinetic energies,
erage of the rock slope without the need to install and the effect of remedial measures (Pfeiffer & Bowen
reflectors. According to Harries & Roberts (2007) the 1989, Guzzetti et al 2002, Jones et al. 2000). This soft-
technique offers sub-millimetre precision of slope wall ware typically does not interact directly with existing
movements without being affected by environmental GIS software. As a result to use these programs, one
conditions such as rain, dust, etc. The accuracy of this must first extract the digital elevation model (DEM)
technique diminishes in areas of vegetative cover and and then recompile it in a form that is suitable for the
hence the technique has been primarily used in open rock fall software.
pit mines. RockFall Analyst, a three dimensional rock fall pro-
gram that was developed as an extension to ArcGIS,
is used to illustrate the added value GIS technology
5 ROCK FALL PROCESS MODEL provides for hazard assessment for rock falls (Lan
et al. 2007).
Rock fall is the simplest of landslide processes and it Rock fall hazard assessment for engineering pur-
is a surface phenomenon. If GIS can be made geotech- poses must capture as many variables as possible
nically smart, the development of rock fall simulation in relation to the rock fall process, kinetic char-
within GIS, a Stage 3 development in Figure 1, should acteristics and their spatial distribution (Dorren &
form a starting point. Seijmonsbergen 2003). As a geomorphologic slope
The Canadian railway industry has been exposed process, rock falls are characterized by high energy
to various ground hazards since the first transconti- and mobility despite their limited volume. The dynam-
nental line was constructed in the 1800s. One of the ics of the rock fall process is dominated by spatially
frequently occurring ground hazards is rock fall. These distributed attributes such as: detachment conditions,
events in mountain regions occur as the result of ongo- geometry features and mechanical properties of both
ing natural geomorphologic processes (Figure 6). The rock blocks and slopes (Agliardi & Crosta 2003).
slope and rock properties controlling the initiation and Today accurate three dimensional morphology can
behaviour of these rock falls can vary widely and it be obtained from LiDAR data but the geotechnical
is not practical to eliminate these rock fall hazards parameters (the coefficient of restitution and friction)
due to the extent and area of potential rock fall source must be calibrated using historical rock fall events.
zones. Nonetheless, reducing this hazard to an accept- The historical rock fall database records provides the
able level of safety requires proactive risk management information of past rock fall events including loca-
strategies. tion of source and deposition, timing of events, size,
Due to the linear corridor occupied by railways influence on the railway operations and the effective-
there is often a need to conduct a large number of ness of existing barriers, should such barriers exist.

10
Lim (2008) showed that high resolution airphotos can
also be used to aid in the assigning the geotechnical
parameters to various regions of the slope and in the
source zone characterisation process.

5.1 Rock fall modelling

Spatial modeling of rock fall hazards along a section of
railway in the Rocky Mountains was carried out using
RockFall Analyst, a GIS extension, which combines
3 dimensional dynamic modeling of the rock fall phys-
ical process with distributed raster-image modeling of
rock fall spatial characteristics. The rock fall modeling
process involves the following steps:
1. The potential instabilities areas (source areas) are
evaluated using the DEM from LiDAR data and
the spatial database of historical rock fall events.
The available LiDAR DEM provided a spatial
resolution of 0.15 m.
2. The rock fall hazard was assessed using Rock-
Fall Analyst by taking into account the distributed
geometry and mechanical parameters, and spa-
tial pattern of rock fall characteristics. This step
included two parts: (1) rock fall process modelling
and (2) raster image modelling.
3. Raster image modelling of rock fall spatial
frequency, flying height (potential energy) and
kinetic energy was used to produce the rock fall
hazard map.
Once the DEM was created from the LiDAR data
and the spatial attributes assigned, potential rock falls
were simulated from all of the possible source areas.
Figure 7a shows the results from the simulation where
the source is simulated as a horizontal line source.
Using the ArcGIS tools, the number of rock fall tra-
jectories occurring was calculated for each grid cell
with an area of one square meter. Spatial geostatisti-
cal techniques within the GIS software were used to
analyze the trajectories and determine the rock fall spa-
tial frequency for the whole study area (Fig. 7b). The
rock fall frequency occurring in the region of railway
track was classified into 10 classes (0–9) and plotted in
ArcGIS in three-dimensional view (Fig. 7c). Figure 7
demonstrates the advantage of using GIS technology
to simulate the rock fall process model.
One of the concerns that frequently arises when
conducting rock fall analysis is the level of resolu-
tion required for the DEM to provide reliable results.
The resolution of the DEM controls the geometry of
the rock fall impacts and trajectories and may con-
trol the physics of the impact. In British Columbia,
where the case history is located, a DEM at a 10-m-grid
spacing is freely available. For comparison purposes
a 1-m-grid LiDAR survey was obtained to provide Figure 7. Spatial frequency analysis of rock falls using
more accurate geometry of the slope. The rock fall RockFall Analyst. All modeling and processing of the
simulations described above using the DEM from the information was carried out in ArcGIS.

11
Figure 8. Comparison of the historical rock fall frequency
impacting the railway tracks with the results from the Rock-
Fall Analyst simulations using the 1-m and 10-m digital
elevation models.

LiDAR survey were repeated using the 10-m DEM.

Figure 8 compares the historical rock fall data base
Figure 9. Rock fall hazard assessment based on rock fall
to the results from RockFall Analysts for both the
frequency and kinetic and potential energy.
1-m and 10-m-grid. It is evident from Figure 8, that
the rock fall simulations from the 1-m provides bet-
ter agreement with the historical rock fall events (Lim
as well as historical data for calibration purposes.
2008).
Once calibrated, the energy output from RockFall
Analyst may be used in the design of protective
5.2 Hazard zoning measures.
Once the spatial distribution of the rock fall has been
computed the energy from such events is required to
complete the hazard assessment. Two raster layers 6 INTEGRATING GIS INTO GEOTECHNICAL
were created in ArcGIS to assess the spatial distri- PRACTICE
bution of rock fall potential-energy (flying/bouncing
height relative to ground) and the kinetic energy Assessment of slope movement and associated hazards
(velocity). The energy raster layers combined with the demands an understanding of the site characteristics
rock fall frequency is used to produce the rock fall haz- and their spatial and temporal variability. Current
ard assessment shown in Figure 8. The rock fall hazard geotechnical modelling tools are focused on numeri-
map clearly identifies the section of railway with the cal analyses and are not generally designed to facilitate
greatest risk, consistent with the historical evidence. the requirements of site investigation and characteriza-
Once the hazard has been identified, the energy raster tion. Site characterisation must address key geospatial
layers can also be used to provide input to the design issues, e.g., complex geology, highly irregular pore-
of protective barriers. Rock fall protection requires water pressure, complex surface geometry and slip
an assessment of both the height (bouncing/flying) surface definition as appropriate. There is little doubt
and velocity of the rock falls. Without such infor- that capturing more complete geomorphological, geo-
mation the design of protective measures is usually logical and geotechnical information improves the
based on single 2 dimensional analysis or qualitative quality of geotechnical site investigation, particularly
methods. when the site is geologically and geotechnically com-
plex (Luna & Frost 1998, Tsai & Frost 1999, Parsons &
Frost 2002, Jaboyedoff et al. 2004, Kunapo et al.
5.3 Summary
2005). Culshaw (2005) in the fifth Glossop lecture
Rock falls are a significant hazard to Canadian rail- suggested that ‘‘the rapid development in technol-
ways. The assessment of such hazards over long ogy over the last twenty years and the digitization
sections of railway requires an efficient means for of increasing amounts of geological data has brought
storing historical data and conducting rock fall sim- engineering geology to a situation in which the produc-
ulations. The development of the three dimensional tion of meaningful three-dimensional spatial models
RockFall Analyst as an extension to ArcGIS, provides of the shallow subsurface is feasible’’. Despite these
the framework for rapid assessment of rock fall haz- advances there are very few spatial tools that help the
ards. The use of such tools requires a detailed DEM geotechnical engineer achieve this goal.

12
 While GIS is increasingly viewed as a key tool for The main aim of Stage 2 in Figure 1 is to establish
managing spatial distribution of data (Nathanail & a comprehensive ground model for the site. The con-
Rosenbaum 1998, Parsons & Frost 2000, Kunapo struction of the ground model is enhanced using three
et al. 2005) it has significant limitations in present- dimensional geological modelling tools commonly
ing three-dimensional (3D) geologic and geotechnical available in the mining industry.
data. Some GIS systems, like ArcGIS, developed by Finally, geotechnical analyses and engineering
the Environmental Systems Research Institute, Inc. decisions are performed in Stage 3 (see Figure 1).
(ESRI), provide a functional developer kit which can In this step the ground behaviour is analysed using
be used to create 3D capability. However the cur- commercially available geotechnical numerical tools.
rent developments in 3D GIS are still not sufficient It is essential that the tools used in Stage 2 commu-
to meet the needs of the geotechnical engineer. Min- nicate with the tools used in Stage 3 so that the data
ing software such as Surpac Vision developed by integration is maintained across all stages.
Gemcom Software International Incorporated pro- One of major issues in this integrated approach
vide a comprehensive system for geological mod- involves data input and output. In order to develop
elling but not geotechnical modelling. For example, an appropriate easy-to-use input and output function,
while three-dimensional solid modelling and two- some industrial standard file formats are employed
dimensional sections can be easily created in Sur- for data conversion and communication throughout
pac, querying inclinometer or piezometer data is not the three Stages. Shape file (.SHP) from ESRI and
readily accomplished. In the following section we Data Exchange File (.DXF) from AutoDesk are both
describe an integrated approach using ArcGIS, Sur- industrial standard formats supported by almost all
pac Vision and numerical modeling, to develop a PC-based CAD and GIS products. The communica-
three-dimensional spatial model of a shallow sub- tion between different Stages in the system developed
surface slide locally referred to as the Keillor Road by Lan & Martin (2007) was implemented using these
Slide. This integrated approach illustrates the added two file formats.
value obtained when data and analyses are tightly The integration of the tools described above offers
integrated. effective digital tools to model heterogeneous geol-
ogy, complex stratigraphy and slip surface geometry,
and variable pore pressure conditions which are critical
6.1 Development of an integrated approach
to complex slope stability problems or other analyses.
Nearly all slope site characterization efforts deal with The tools also provide for incorporating findings from
surface mapping, geological information from bore- monitoring data. In the following section the tools
hole data and monitoring data. The work flow from are demonstrated using a translational bedrock slide
data collection through to engineering analyses was in Edmonton Alberta.
outlined by Lan & Martin (2007) and can be summa-
rized in three stages (see Figure 1). Stage 1 involves
6.2 Case study: Keillor Road slide
the data collection, management and geosynthesis
of the data. Modern GIS software provides effec- A complete description of the Keillor Road bedrock
tive tools for the handling, integrating and visualizing slide was given by Soe Moe et al. (2005). The failure
diverse spatial data sets (Brimicombe 2003). There- of the slope occurred over a number of years with the
fore, in Stage 1, the functionality of GIS provides largest deformations occurring in 2002. The slide took
an essential role in collecting, storing, analyzing, place along the bank of the North Saskatchewan River
visualizing and disseminating geospatial information. valley in Edmonton, Alberta Canada (Fig. 10). The
This information could be basic site investigation site investigations for the slide were conducted over
data, such as geomorphology and geology condi- a period of 15 years using traditional boreholes and
tions, and diverse, continually evolving geotechnical monitoring systems.
parameters, such as displacement and pore pres- Figure 11 shows a plan view of the site created in
sure readings from geotechnical instruments. Most ArcGIS indicating the outline of the slide, the topog-
GIS tools have limitations in representing time series raphy of the area, location of the tension cracks and
data such as the displacement data from inclinome- location of the main scarp. Figure 11 also shows the
ter or pore pressures from piezometers. Therefore location of the boreholes that had been used in the site
additional functional tools were required for the stan- investigations over the 15 year period. The major bene-
dard ArcGIS software. These tools have been imple- fit of assembling the data in ArcGIS is that the borehole
mented using ArcObject, an ArcGIS developer kit, and symbols are dynamically linked to the data base and
Visual studio.net, a software developing package by instrumentation data installed in the boreholes.
Microsoft. These development tools provide capabil- To reconstruct the dynamics and kinematics of
ity for users to interact and communicate with various the processes acting on the slope and to determine
data sets. their spatial and temporal distribution, Lan & Martin

13
 deceleration of slope movement. In addition to the
displacement versus time-plots, plots of displacement
vector directions and displacement rate offer the abil-
ity to identify and evaluate the spatial and temporal
character of the deformations, all in a user-friendly
environment.
In addition to being able to process the data quickly
in both a visual manner as well as conduct specific
depth queries, the user can quickly assess the kinemat-
ics of the slide. Multiple movement zones at different
depth are often detected in slope inclinometer plots. In
this case, two movement zones were identified in the
inclinometer readings for borehole B02-2 (Fig. 12).
Figure 10. Photograph of the Keillor Road Slide, from Soe One zone was from depth 0.61 m to depth 2.44 m
Moe et al (2005). (zone 1) and the other zone was from depth 7.32 m
to depth 8.53 m (zone 2). Their deformation histories
can be rapidly shown on the plan map (Fig. 12). It
can be seen clearly that zones 1 and 2 show different
movement characteristics. The moving direction of the
shallow zone 1 changed direction frequently while the
direction of zone 2 was essentially unchanged.
Once the major rupture surface is identified, the
spatial deformation pattern from all the inclinometer

Figure 11. Plan view of Keillor Road slope showing loca-

tion of the site investigation boreholes, tension cracks and
outline of the slide. Contour elevations have been removed
for clarity.

(2007) developed add-on tools for the processing of

borehole information, plotting of time-series displace-
ment data from slope inclinometers, pore pressure
data from piezometers and relative geomorphologi-
cal features, such as tension cracks. The add-on tools
provide all of the standard types of plots for analysing
slope inclinometer data. From these standard plots dis-
crete movement zones can be defined by specifying Figure 12. Displacement history at different shear zones of
the from-to-depths. The resultant time-displacement site B02-2. Shearing zone 1 and zone 2 are characterized by
plots for these discrete zones show acceleration or obviously different displacement vectors.

14
data can be shown (Fig. 13). This provides a consis- As mentioned earlier, Surpac Vision provides
tency check within the data sets as well highlights advanced tools for viewing and interpretation of geol-
the more active portions of the slide as both total ogy data. Connecting to the same geological database
displacements as well as displacement rate can be as used by ArcGIS, the three-dimensional geology
shown. model for the Keillor Road Slide was created in
Geotechnical parameters are usually measured at Surpac Vision. Together with the other data imported
points during site investigation by in-situ tests or by from ArcGIS, such as the geomorphological surface,
laboratory tests. Geostatistical kriging and simula- displacement and pore pressure readings, and tension
tion techniques in GIS offer powerful spatial mod- crack planes, a comprehensive ground model for Keil-
eling tools for visualising the spatial variability of lor Road slope was created in Surpac Vision. From
these parameters (Nathanail & Rosenbaum 1998). the model, the spatial extent of the slope which is at
Parsons & Frost (2002) argued that such statistical risk from instability can be immediately defined by
approaches improve the quality of site investigation the displacement data and the surface mapping infor-
data. Pore pressure is an essential parameter in slope mation. Critical profile sections can be extracted in
stability studies. Lan & Martin (2007) used geosta- Surpac Vision along the section lines parallel to the
tistical techniques to interpret the point pore pressure displacement vectors. These sections now include all
data into a spatial pore pressure surface. When con- information managed and produced in GIS and Sur-
ducting such geostatistical analysis it is important to pac. These section profiles can be exported to DXF
ensure that the pore pressures are being measured on files which can then be optimized for slope stability
the same geological unit which can be readily verified analysis. Nearly all modern slope stability software
by comparing the borehole logs and piezometer instal- such as Slope/W or Slide can readily import DXF files.
lation locations. Lan & Martin (2007) also attached the However, it is important that the user examine these
three dimensional displacement curves obtained from DXF files to ensure that the relevant information is
the inclinometer data to the boreholes as lines to show captured.
a spatial relationship between displacement locations Slope/W is widely used in geotechnical engineer-
and pore pressure. ing practice for analyzing the stability of slopes. It uses
limit equilibrium theory to compute the critical factor
of safety (Krahn 2003). The essential geometry ele-
ments in Slope/W include the ground surface, complex
geological regions, and pore pressure line and tension
crack lines. In many slope stability problems it is very
important to establish an accurate representation of
the slope surface geometry because small changes in
the slope profile can have a significant impact on the
calculated factor of safety especially when the rupture
surface is relatively flat such as the translational slide at
Keillor Road. Creating the section profile from LiDAR
survey ensures that the most accurate surface geome-
try is captured. Figure 14 shows the final geometry and
geology modeled using Slope/W. The integration of the

Figure 13. Spatial distribution of slope deformation from

all the inclinometer data at the same discrete movement
zone. Two major slope portions with different displacement Figure 14. Slope stability analysis using Limit equilibrium
evolution are divided by Keillor Road. and/or Finite-element analysis.

15
GeoStudio Software means that this model can also be has cut through about 150 metres of glacial sed-
used for conducting deformation or stress analyses. iments (Porter et al. 2002). Quaternary sediments
occur within the major valleys where deep valley fills
have been dissected and terraced by postglacial down-
7 ANALYSING COMPLEX LANDSLIDES cutting of the trunk rivers. The landslides occurred on
the steep walls of an inner valley that formed during
In the previous section we showed the benefit of inte- the Holocene when Quaternary sediments filling the
grating technologies when analyzing a single slide. broader Thompson River valley were incised. The val-
In this section we demonstrate the added value when ley fill consists dominantly of permeable sediments,
considering multiple complex landslides. the exception being a unit of rhythmically-bedded silt
and clay in the Pleistocene sequence (Clague & Evans
2003). The surficial materials in the area are tills, flu-
7.1 Background vial, fluvioglacial, lacustrine and colluvial deposits
(Ryder 1976).
Large translational landslides with rupture surfaces
Individual investigations had been carried out for
through glacial lake sediments in preglacial valleys
the six most active of these earth slides in the Thomp-
are common hazards within river valleys of Western
son Valley since the early 1980s. A major effort
Canada (Evans et al. 2005). Eleven, retrogressive,
was initiated in 2003 to re-analyse the data that had
multiple, translational earth slides have occurred along
been collected over the past 20 years using the spa-
10 kilometres of the Thompson River valley between
tial capabilities inherent in GIS tools. Eshraghian
the communities of Ashcroft and Spences Bridge in
et al (2007) completed a comprehensive study of the
south-central British Columbia, Canada. The Cana-
slides and concluded that the rupture surfaces that
dian Pacific Railway (CPR) and Canadian National
had been detected in the individual slides followed the
Railway (CN) main rail lines were constructed through
highly plastic, overconsolidated clays within a Pleis-
the Thompson River valley in 1885 and 1905 respec-
tocene stratigraphic unit consisting of up to 45 metres
tively. Both have had recurring slope stability prob-
of rhythmically-bedded glaciolacustrine deposit of silt
lems along this valley (Fig. 15). Given that the two
and clay couplets, ranging from less than 1 cm to
national railroads traverse the same landslide prone
several tens of centimetres thick (Fig. 16). These sedi-
area, the evaluation of risk at this location is a matter
ments may be several hundred thousand years old and
of considerable significance.
The Ashcroft area is part of the Thompson Plateau,
a subdivision of the Interior Plateau of British
Columbia. The Thompson River flows south and

Figure 16. Geological units in the earth slides and highland

Figure 15. Major landslides south of Ashcroft, BC, (modi- terraces in Thompson Valley. The arrows indicate the rupture
fied from Eshraghian et al 2007). surfaces. Modified from Eshraghian et al 2007.

16
Figure 17. Example of an aerial photograph draped over
a digital elevation model produced from an airborne Lidar
survey (modified from Eshraghian et al 2007).

thus of Middle or Early Pleistocene age (Clague &

Evans 2003). Samples of this unit from boreholes in
South Slide show layers of brown, high-plastic clay
1 to 20 cm thick between thicker layers of olive silt
(Figure 16).
The deposits of the three glacial sequences are
separated by unconformities. Figure 16 shows the
geological succession synthesized from borehole logs
and outcrops in scarps and terraces in the Ashcroft
area based on units proposed by Clague & Evans
(2003). Eshraghian et al. (2007) used GIS technol-
ogy to estimate the slide volumes which varied from
1.8 to 21.4 Mm3 and spatially correlate the rupture sur-
face. They concluded that the stratigraphic boundaries
have tilts of 1.7 m/km similar to the glacial lake bot-
tom and that sliding was occurring along essentially
the same layers within the glaciolacustrine sediments Figure 18. Simplified multiblock model illustrating the
(Figure 16). They then examined the surface expo- sliding process since deglaciation at the Slide CN50.9 (mod-
sure of the larger slides using LiDAR technology. ified from Eshraghian et al 2007).
The airborne LiDAR provided a vertical resolution of
±150 mm and when combined with the aerial pho- horizontal movement by block A and horizontal and
tographs illustrated the multiple blocks associated with vertical movement by block B. This sliding and also the
retrogressive slides of this nature (Figure 17). Thompson River erosion caused progressive failure on
the deeper rupture surface and the slide was ready for
another retrogression (Stage 4, Figure 18). The most
7.2 Multiblock retrogressive model
recent retrogression of the Slide CN50.9 happened in
Eshraghian et al 2007 examined Slide CN50.9 using September 1897. During this stage, block D moved
traditional site investigation boreholes and the com- down on the main scarp and rest of slide material
bined Lidar DEM and aerial photograph model to moved horizontally toward the river (Stage 5, Fig. 4).
develop the geological and movement history since During this retrogression in the early morning of
deglaciation (Fig. 18). During the first stage a September 22, 1897, residents of Ashcroft were awak-
braided Thompson River started cutting through the ened by loud, thunder-like rumblings. The landslide
glacial sediments after deglaciation. Thompson River constricted the Thompson River without completely
continued its down-cutting until it reached the first blocking it (Clague & Evans 2003). From these expla-
weak layer and potential rupture surface (stage 2, nations, the movement rate during this retrogression
Fig. 18). Progressive failure within this weak layer is estimated to be rapid. Following the slide, Thomp-
caused sliding of blocks A and B on the shallower rup- son River removed part of the toe of the slide, mainly
ture surface. More down-cutting by Thompson River within block A (stage 6, Figure 18).
encountered a deeper weak layer (Stage 3, Fig. 18). Eshraghian et al 2007 used the conceptual model
This time, movement happened without retrogres- developed in Figure 18 to develop the geological
sion on the deeper rupture surface. It caused more multiblock model for the retrogressive slide shown

17
 the Thompson River levels and slide movements from
1970 to 2000. Data prior to 1970 are sparse. They
presented a correlation between the cumulative river
level difference from the average river level (CRLD)
and active years. They concluded that the main trigger
for reactivation of these slides was the discharge of
Thompson River that produced above average river
levels for prolonged periods (Figure 20).
Clague & Evans (2003) suggested ‘‘irrigation of
Figure 19. Slide CN50.9 cross-section showing the rupture the bench lands above the river, especially in the late
surfaces, stratification, and borehole locations (see Figure 15
for location, modified from Eshraghian et al 2007).
1800s, introduced large amounts of water into the
valley fill. High pore pressures probably developed
locally at the top of the rhythmically bedded silt-
clay unit, triggering large landslides’’. They added
in Figure 19. This slide is now moving on two rup- ‘‘although high pore pressures generated by irriga-
ture surfaces as a multiple translational earth slide. tion related groundwater discharge probably triggered
The positions of the rupture surfaces were determined most of the historical landslides in the Ashcroft area,
by inclinometers and the rate of movement differs the fundamental causes are geological’’. Morgenstern
not only on the main rupture surfaces but between (1986) also concluded that the primary trigger for
inclinometer measurements on the shallower rupture the movement in these complex slides was related
surface, suggesting possible more small blocks at the to their retrogressive nature characterised by toe-
toe within R-3 block. Developing a single factor of initiated movements. Eshraghian (2007) examined the
safety for such a complex slide does not convey the following possible trigger mechanisms: (1) rainfall,
geological complexity of the slide nor communicate (2) irrigation, (3) Thompson River Level and (4) toe
the risk to the railways. erosion. He concluded that the changes in the river
level had the largest impact on the stability of the slides
and that this occurred when the river remained high
7.3 Trigger mechanisms
for a long period and then retreated to cause a draw-
The location of the piezometers at the toe of Slide down effect on the slope toe blocks. Those slides which
CN50.9 indicate upward gradients near the toe of the when combined with toe-erosion showed the greatest
slope. All piezometers respond to changes in the river potential for rapid movements.
level but the shallower the piezometer and the closer
it is located to the river, the greater the response. The
piezometers also show a 7 to 10 days delay between the
river level changes and the piezometer on the deeper
rupture surface. The piezometers indicate that the slide
portion near to the scarp, i.e. the slide head, is gener-
ally a recharge zone and the toe is a discharge zone
when the Thompson River level is low. On the other
hand, when the Thompson River starts rising, the water
from the river seeps towards the slide mass and offsets
the upward gradient condition at the toe. However, the
river may not stay at these high levels for sufficient
time to allow the flow system to reach equilibrium.
Therefore, the top part of the rhythmically bedded
silt and clay layer (unit 2) is more affected by river
level changes than the lower part. In the years that
the Thompson River stays at high levels for longer
periods, the piezometers show the greatest increase in
pore water pressures because the water has more time
to seep through the soil mass.
The average rainfall in the area has been increasing
since the 1920s from 150 mm/year to 240 mm/year
(Porter et al. 2002). Despite this rainfall increase in the Figure 20. Illustration of the average Thompson River
area, Eshraghian et al. 2005a did not find a correlation Level compared to 1981 when typical slow movements were
between slide movements and short term or long term recorded and 1982 when the Goddard Slide was reactivated,
rainfall. Eshraghian et al. (2005a, 2005b) examined (modified from Eshraghian et al 2007).

18
7.4 Movement and risk well as uncertainties from the groundwater modeling
and toe erosion. The probabilistic rates of movement
It is obvious from the discussion above that attempt-
were calculated using the frequency of the trigger
ing to capture the risk from the movement associated
(the Thompson River flood, Figure 21), the historic
with such landslides with a single number is not prac-
movement rates, for each reactivation block.
tical. Eshraghian et al 2007 used a quantitative hazard
The result of calculating the probability of
analysis in a framework that considered the differ-
movement for reactivation blocks are a movement
ent post-failure movement rates. They demonstrated
probability distribution (Figure 22) which shows the
the approach using probabilistic stability analyses
probability distribution of different movement rates
that included material and trigger uncertainties as
which may happen during the design life time of the
project. They also reported the results for each reac-
tivation block in the form of probability of different
movement rates using the movement rate class sug-
gested by Cruden & Varnes (1996) calculated for the
designed life time of 100 years (Fig. 23).

7.5 Summary
The Introduction for this paper drew attention to
the large number of contributory processes that have
to be considered in developing an effective process
model of a landslide and its consequences. All of
these contributory processes have had to be consid-
ered in the example just presented; from geology and
geomorphology through hydrological and geotechni-
Figure 21. Thompson River level for different yearly dis-
charge return periods.
cal characterization and finally geotechnical and risk

Figure 22. Histogram frequency distribution of movement Figure 23. Frequency of different movement rate classes
rate for two translational blocks on shallower and deeper for reactivation blocks defined within Slide CN50.9 for a
rupture surfaces. 100 year return period.

19
analyses, utilizing both deterministic and probabilis- iv. The Gateway Pipeline Project for providing
tic considerations. The capacity for undertaking such Figure 3.
complex landslide analyses was enhanced by the avail-
ability of recently developed tools such as LiDAR The development of the ArcGIS tools and the
imagery. However, it is unlikely that the end product case histories described in this paper was supported
could have been achieved without utilizing GIS for by the Canadian Railway Ground Hazard Research
spatial data management, correlations and analysis. Program, a collaborative research program between
Canadian National Railway, Canadian Pacific Rail-
way, Transport Canada, Geological Survey of Canada,
8 CONCLUDING REMARKS University of Alberta, Queen’s University and the
Natural Sciences and Engineering Council of Canada.
A number of recent technical advances are leading
to dramatic improvements in the study of landslides
and the evaluation of appropriate risk mitigation REFERENCES
measures. This paper draws attention to some, such
as the application of LiDAR to delineate landslides Agliardi, F. & Crosta, G.B. 2003. High resolution three-
more clearly than aerial photographs, and the role of dimensional numerical modelling of rockfalls. Interna-
InSAR to monitor ground movements over large areas tional Journal of Rock Mechanics and Mining Science
with increasing accuracy. A number of other tools are 40 (4): 455–471.
entering practice that merit discussion but were beyond Ardizzone, F., Cardinali, M., Gzzetti, F. & Reichen-
the scope of this paper. bach, P. 2007. Identification and mapping of recent
rainfall-induced landslides using elevation data collected
It has been the central premise of this paper that
by airborne LiDAR. Natural Hazards and Earth System
the most important advances have been associated Sciences 7: 637–650.
with improved visualization of landslides and related Bjerrum, L. 1967. Progressive failure in slopes and
processes, both through surface and sub-surface fea- overconsolidated plastic clay and clay shales. Journal
tures. To this end, our experience leads us to the Soil Mechanics and Foundations Division, ASCE 93
view that GIS is capable of making transformative (SM5): 3–49.
contributions. Brimicombe, A. 2003. GIS, Environmental modelling and
Examples of use of GIS in geotechnical assessment, engineering, London; N.Y.: Taylor & Francis, 312 p.
beyond its routine application of archiving surface Carter, W.E., Shresthe, R.L. & Slatton, K.C. 2007. Geodetic
laser scanning. Physics Today 60: 41–47.
information, have been provided. A general rockfall
Chacón, J., Irigaray, C., Fernandez, T. & El Hamdouni, R.
simulation model has been developed with GIS. Other 2006. Engineering geology maps: landslides and geo-
examples illustrate the integration of GIS with subsur- graphical information systems. Bulletin of Engineering
face modelling capability with interfaces to any kind Geology and the Environment 65: 341–411.
of geotechnical analysis software. Chandler, J.H. & Brunsden, D. 1995. Steady state
Landslide data management and analysis of all behaviour of the Black Ven mudslide: the application of
kinds in GIS will be essential for future progress in archival analytical photogrammetry to studies of land-
landslide engineering as three-dimensional visualiza- form change. Earth Surface Process and Landforms
tion and modelling capabilities improve. 20: 255–275.
Clague, J.J. & Evans, S.G. 2003. Geological framework of
large historic landslides in Thompson River Valley, British
Columbia. Environmental and Engineering Geoscience
ACKNOWLEDGMENTS 9: 201–212.
Colesanti, C. & Wasowski, J. 2006. Investigating Land-
The authors wish to acknowledge the assistance of the slides with Space-borne Synthetic Aperture Radar (SAR)
following: Interferometry. Engineering Geology 88: 173–199.
Cruden, D.M. & Varnes, D.J. 1996. Landslide types and
i. Dr. Hengxing Lan, Research Engineer, for his processes. Transportation Research Board Special report
leadership in developing GIS based tools at the 247: 36–75.
University of Alberta. Culshaw, M.G. 2005. The Seventh Glossop Lecture—From
ii. Mr. Cory Froese, Team Leader—Geological Haz- concept towards reality: developing the attributed 3D
ards, Alberta Geological Survey/Energy and Util- geological model of the shallow subsurface Quarterly
ities Board, for assisting us with understanding Journal of Engineering Geology and Hydrogeology 38:
231–284.
the current status of InSAR applied to landslide Cunningham, D., Grebby, S., Tansey, K., Gosar, A. &
studies. Kastelic, V. 2006. Application of air borne LiDAR to
iii. The graduate students who have collaborated with mapping seismogenic faults in forested mountainous ter-
us in these and related studies over the past few rain, southeastern Alps, Slovenia. Geophysical Research
years. Letters 33: L20308.

20
Dai, F.C. & Lee, C.F. 2004. A spatiotemporal probabilis- Franks, C.A., Koor, N.P. & Campbell, S.D.G. 1998. An inte-
tic modelling of storm-induced shallow landsliding using grated approach to the assessment of slope stability in
aerial photographs and logistic regression. Earth Surface urban areas in Hong Kong using thematic maps. Proc 8th
Processes and Landforms 28: 527–545. IAEG Congress. Vancouver: Balkema, 1103–1111.
Delmonaco, G., Leoni, G., Margottini, C., Puglisi, C. & Froese, C.R., Kosar, K. & van der Kooij, M. 2004. Advances
Spizzichino, D. 2003. Large scale debris-flow hazard in the application of InSAR to complex, slowly moving
assessment: a geotechnical approach and GIS mod- landslides in dry and vegetated terrain; in Landslides:
elling. Natural Hazards and Earth System Sciences Evaluation and Stabilization, W. Lacerda, M. Erlich,
3: 433–455. S.A.B. Fontoura & A.S.F. Sayao (ed.), Proceedings of the
Dietrich, W.E. & Montgomery, D.R. 1998. A digital terrain 9th International Landslide Symposium, Rio de Janeiro,
model for mapping shallow landslide potential. Technical Brazil: 1255–1264.
Report NCASI. http://socrates.berkeley.edu/∼geomorph/ Froese, C.R., Poncos, V., Skirrow, R., Mansour, M. &
shalstab/ Martin, C.D. 2008. Characterizing Complex Deep Seated
Dorren, L.K.A. & Seijmonsbergen, A.C. 2003. Comparison Landslide Deformation using Corner Reflector InSAR
of three GIS-based models for prediction rock fall runout (CR-INSAR): Little Smoky Landslide, Alberta. Proceed-
zones at a regional scale. Geomorphology 56: 49–64. ings of the 4th Canadian Conference on Geohazards.
Eshraghian, A. 2007. Hazard analysis of reactivated earth Quebec: In Press.
slides along the Thompson River Valley, Ashcroft, British Günther, A., Carstensen, A. & Pohl, W. 2004. Automated
Columbia. PhD Thesis Dept. Civil & Environmental sliding susceptibility mapping of rock slopes. Natural
Engineering, University of Alberta, Edmonton, Alberta, Hazards and Earth Systems Sciences 4: 95–102.
Canada. Guzzetti, F., Crosta, G., Detti, R. & Agliardi, F. 2002.
Eshraghian, A., Martin, C.D. & Cruden, D.M. 2005a. Land- STONE: a computer program for the three-dimensional
slides in the Thompson River valley between Ashcroft and simulation of rock-falls. Computers & Sciences 28:
Spences Bridge, British Columbia. In Proceedings of the 1079–1093.
International Conference on Landslide Risk Management, Harries, N.J. & Roberts, H. 2007. The use of slope stabil-
Vancouver, Canada, May 31 to June 4, 2005: 437–446.s ity radar (SSR) in managing slope instability hazards. In
Eshraghian, A., Martin, C.D. & Cruden, D.M. 2005b. Earth Eberhardt, E., Stead, D. & Morrison, T. (eds), Proceedings
slide movements in the Thompson River valley, Ashcroft, 1st Canada-U.S. Rock Mechanics Symposium, Vancouver
British Columbia. In Proceedings of the 58th Cana- 1: 53–59. London: Taylor & Francis Group.
dian Geotechnical Conference, Saskatoon, Saskatchewan, Haugerud, R.A., & Harding, D.J. 2001. Some algorithms
Canada, September 18–21, 2005. for virtual deforestation (VDF) of lidar topographic sur-
Eshraghian, A., Martin, C.D. & Cruden, D.M. 2007. Com- vey data. International Archives of Photogrammetry and
plex Earth Slides in the Thompson River Valley, Ashcroft, Remote Sensing 34–3/W4: 211–217.
British Columbia. Environmental and Engineering Geo- Hong, Y., Adler, R.F. & Huffman, G.J. 2007. Satel-
science Journal XIII: 161–181. lite remote sensing for global landslide monitoring,
Evans, S.G., Cruden, D.M., Brobrowsky, P.T., Guthrie, EDS. Transactions of the American Geophysical Union
R.H., Keegan, T.R., Liverman, D.G.E. & Perret, D. 88: 357.
2005. Landslide risk assessment in Canada; a review of Jaboyedoff, M., Ornstein, P. & Rouiller, J.-D. 2004. Design
recent developments. In Proceedings of the International of a geodetic database and associated tools for monitoring
Conference on Landslide Risk Management, Vancouver, rock-slope movements: the example of the top of Randa
Canada, 31 May–3 June 2005. A.A. Balkema. 351–434. rockfall scar. Natural Hazards and Earth System Sciences
Farina, P., Casgli, N. & Ferretti, A. 2007. Radar- 4: 187–196 (pdf, 5755 Ko).
Interpretation of InSAR Measurements for Landslide Jaboyedoff, M., Metzger, R., Oppikofer, T., Coulture, R.,
Investigations in Civil Protection Practices. In Schaefer, Derron, M., Locat, J. & Turmel, D. 2007. New insight
V.R., Schuster, R.L., Turner, A.K. (eds), Landslides and techniques to analyse rock slope relief using DEM and
Society. AEG Special Publication 23: 272–283. 3D-imaging cloud points: COLTOP-3D software, In Eber-
Farina, P., Colombo, D., Fumagalli, A., Marks, F. & hardt, E., Stead, D. & Morrison, T. (eds), Proceedings 1st
Moretti, S. 2006. Permanent scatterers for landslide Canada-U.S. Rock Mechanics Symposium, Vancouver 1:
investigations: Outcomes from the ESA-SLAM Project. 61–68. London: Taylor & Francis Group.
Engineering Geology 88: 200–217. Jones, C.L., Higgins, J.D., & Andrew, R.D. 2000. Colorado
Ferretti, A., Prati, C. & Rocca, F. 1999. Permanent scat- Rock fall Simulation Program Version 4.0. Colorado
terers in SAR interferometry. International Geoscience Department of Transportation. Colorado Geological Sur-
and Remote Sensing Symposium, Hamburg, Germany, 28 vey, March 2000. 127 pp.
June-2 July, 1999. 1–3. Kemeny, J., Norton, B. & Turner, K. 2006. Rock slope sta-
Ferretti, A., Prati, C. & Rocca, F. 2000. Nonlinear subsidence bility analysis utilizing ground-based LiDAR and digital
rate estimation using permanent scatterers in differential image processing. Felsbau 24: 8–16.
SAR interferometry. IEEE Transactions on Geoscience Krahn, J. 2003. The 2001 R.M. Hardy Lecture: The lim-
and Remote Sensing 38(5): 2202–2212. its of limit equilibrium analyses. Canadian Geotechnical
Ferretti, A., Prati, C. & Rocca, F. 2001. Permanent scatterers Journal 40(3): 643–660.
in SAR interferometry. IEEE Transactions on Geoscience Kunapo, J., Dasari, G.R., Phoon, K.K. & Tan, T.S. 2005.
and Remote Sensing 39(1): 8–20. Development of a Web-GIS based geotechnical informa-
Findley, D.P. 2007. Rock Stars. Civil Engineering July: tion system. Journal of Computing in Civil Engineering
46–51. 19(3): 323–327.

21
Lan, H.X., Lee, C.F., Zhou, C.H. & Martin, C.D. 2005. Parsons, R.L. & Frost, J.D. 2002. Evaluating site investigation
Dynamic characteristics analysis of shallow landslides quality using GIS and geostatistics. Journal of Geotechni-
in response to rainfall event using GIS. Environmental cal and Geoenvironmental Engineering 128 (6): 451–461.
Geology 47(2): 254–267. Pfeiffer, T.J. & Bowen, T. 1989. Computer simulation of
Lan, H. & Martin, C.D. 2007. A digital approach for rock falls. Bulletin of the Association of Engineering
integrating geotechnical data and stability analyses in Geologists 26 (1): 135–146.
E.Eberhardt, D. Stead & Morrison, T. (eds), Rock Mech- Porter, M.J., Savigny, K.W., Keegan, T.R., Bunce, C.M. &
nanics: Meeting Society’s Challenges and Demands, MacKay, C. 2002. Controls on stability of the Thompson
45–2. London: Taylor and Francis Group. River landslides. In Proceedings 55th Canadian Geotech-
Lan, H., Martin, C.D. & Lim, C.H. 2007. Rockfall Ana- nical Conference: Ground and Water: Theory to Practice,
lyst: a GIS extension for three-dimensional and spa- Niagara Falls, Ontario, Vol. 2 , pp 1393–1400.
tially distributed rockfall hazard modelling. Computers & Pötsch, M., Schubert, W. & Gaich, A. 2007. The applica-
Geosciences 33: 262–279. tion of metric 3D images for the mechanical analysis of
Lim, C.H. 2008. A process model for rock fall, PhD Thesis keyblocks. In Eberhardt, E., Stead, D. & Morrison, T.
Dept. Civil & Environmental Engineering, University of (eds), Proceedings 1st Canada-U.S. Rock Mechanics Sym-
Alberta, Edmonton, Canada. posium, Vancouver 1: 77–84. London: Taylor & Francis
Luna, R. & Frost, J.D. 1998. Spatial liquefaction analysis sys- Group.
tem. Journal of Computing in Civil Engineering 12 (1): Radbruch, D.H. & Crowther, K.C. 1973. Map showing areas
48–56. of estimated relative susceptibility to landsliding in Cal-
Maffei, A., Martino, S. & Prestininzi, A. 2005. From the ifornia. U.S. Geological Survey, Miscellaneous Geologic
geological to the numerical model in the analysis of Investigation Map I-747.
gravity-induced slope deformations: An example from the Refice, A. & Capolongo, O. 2002. Probabilistic modelling
Central Apennines (Italy). Engineering Geology 78(3–4): of uncertainties in earthquake-induced landslide hazard
215–236. assessment. Computing and Geoscience 28: 735–749.
Martin, C.D., Tannant, D.D. & Lan, H. 2007. Comparison Rhina, D. 1992. Why GIS? ARC News 11: 1–4.
of terrestrial-based, high resolution, LiDAR and digital Rizkalla, M. & Randall, C. 1999. A Demonstration of
photogrammetry surveys of a rock slope. Eberhardt, E. Satellite-Based Remote Sensing Methods for Ground
Stead, D. & Morrison, T. (eds), Proceedings 1st Canada- Movement Monitoring in Pipeline Integrity Management.
U.S. Rock Mechanics Symposium, Vancouver. Taylor & Proceedings 18th International Conference on Offshore
Francis Group, London 1: 37–44. Mechanics and Arctic Engineering. St. John’s, OMAE
McCardle, A., Rabus, B., Ghuman, P., Rabaco, L, M.L., 99/PIPE-5004, ASME.
Amaral, C.S. & Rocha, R. 2007. Using Artificial Point Rosser, N.J., Petley, D.N., Lim, M., Dunning, S.A. &
Targets for Monitoring Landslides with Interferometric Allison, R.J. 2005. Terrestrial laser scanning for monitor-
Processing. Anais XIII Simposio Brasileiro de Sensoria- ing the process of hard rock coastal cliff erosion. Quar-
mento Remoto. INPE: 4933–4934. terly Journal of Engineering Geology and Hydrogeology
Meisina, C., Zucca, F., Conconi, F., Verri, F., Fossati, D., 38: 363–373.
Ceriani, M. & Allievi, J. 2007. Use of Permanent Scatter- Ryder, J.M. 1976. Terrain inventory and Quaternary geol-
ers Technique for Large-scale Mass Movement Investiga- ogy, Ashcroft, British Columbia, Geological Survey of
tion. Quaternary International 171–172: 90–107. Canada, Ottawa.
Mejia-Navarro, M., Wohl, E.W. & Oaks, S.D. 1994. Geologi- Schulz, W.H. 2004. Landslides mapped using LiDAR
cal hazards, vulnerability, and risk assessment using GIS: imagery, Seattle, Washington. U.S. Geological Survey,
model for Glenwood Springs, Colorado. Geomorphology Open-File Report 2004-1396.
43: 117–136. Schulz, W.H. 2007. Landslide susceptibility revealed by
Morgenstern, N.R. 1986. Goddard landslide of September, LiDAR imagery and historical records, Seattle, Washing-
1982, summary of the opinion of Norbert R. Morgenstern. ton. Engineering Geology 89: 67–87.
Report prepared for court between Canadian Pacific Singhroy, V. 2005. Remote sensing of landslides. In Glade, T.,
Limited and Highland Valley Cattle Company Limited. Anderson, M. & Crozier, M.J. (eds), Landslide Hazard
Supreme Court of British Columbia: Vancouver registry and Risk 469–492. Wiley.
No. C841694. Soe Moe, K.W., Cruden, D.M., Martin, C.D., Lewycky,
Morgenstern, N.R. 2000. Performance in geotechnical prac- D. & Lach, P.R. 2005. 15 years of movement at Keillor
tice: Inaugural Lumb Lecture. Transactions Hong Kong Road, Edmonton. Proceedings 58th Canadian Geotechni-
Institute of Engineers 7: 1–15. cal Conference & 6th Joint CGS/IAH-CNC Groundwater
Nathanail, C.P. & Rosenbaum, M.S. 1998. Spatial manage- Specialty Conference, Saskatoon, 1: 281–388.
ment of geotechnical data for site selection. Engineering Soeters, R. & Van Westen, C.J. 1996. Slope instability, recog-
Geology 50(3–4): 347–356. nition, analysis, and zonation in Turner, A.K. and Schus-
Parise, M. 2003. Observations of surface features on an active ter, R.L. (eds). Landslides: Investigation and Mitigation
landslide and implications for understanding its history of 129–177. Washington, D.C.: National Academy Press.
movement. Natural Hazards and Earth System Sciences Sturznegger, M., Stead, D., Froese, C., Moreno, F. & Jaboyed-
3: 569–580. off, M. 2007. Ground-based and airborne LiDAR for
Parsons, R.L. & Frost, J.D. 2000. Interactive analy- structural mapping of the Frank Slide. In Eberhardt, E.,
sis of spatial subsurface data using GIS-Based tool. Stead, D. & Morrison, T. (eds), Rock Mechanics: Meet-
Journal of Computing in Civil Engineering 14(4): ing Society’s Challenges and Demands 925–932. London:
215–222. Taylor and Francis.

22
Tsai, Y.C. & Frost, J.D. 1999. Using geographic information Varnes, D.J. 1984. International Association of Engineer-
system and knowledge base system technology for real- ing Geology Commission on Landslides and Other Mass
time planning of site characterization activities. Canadian Movements on Slopes. Landslide Hazard Zonation. Inter-
Geotechnical Journal 36 (2): 300–312. national Association of Engineering Geology. UNESCO
Van Westen, C.J. & Getahun, F.L. 2003. Analyzing the Natural Hazard Series 3(63) pp.
evolution of the Tessina landslide using aerial photo- Wachal, D.J. & Hudak. P.F. 2000. Mapping landslide sus-
graphs and digital elevation models. Geomorphology ceptibility in Travis County, Texas, USA. GeoJournal 51:
54: 77–89. 245–253.
Van Westen, C.J. 2007. Mapping landslides: Recent devel- Walstra, J., Dixon, N. & Chandler, J.H. 2007. Historical
opments in the use of the digital spatial information. aerial photographs for landslide assessment: two case his-
Proceedings 1st North American Landslide Conference, tories. Quarterly Journal of Engineering Geology and
Vail, CO 221–238. Hydrogeology 40: 315–332.

23
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Limit equilibrium and finite element analysis – A perspective

of recent advances

Zuyu Chen
China Institute of Water Resources and Hydropower Research, China

Keizo Ugai
Department of Civil Engineering, Guma University, Japan

ABSTRACT: This paper gives a general review of the recent advances in the applications of limit equilibrium
(LEM) and Finite Element Methods (FEM) for slope stability analysis. Accuracies of various LEM including
Sarma’s have been reviewed. Special attentions have been given to the strength reduction finite element method
regarding its applicability, criteria for failure indications, and the treatment for modulus, Poisson ratio and
dilation angles.

1 INTRODUCTION 2 THE LIMIT EQUILIBRIUM ANALYSIS

METHODS
Analyses play an important role in assessing the risks
involved in a potential landslide and the design of an 2.1 General
engineered slope.
When commenting the further efforts in updating the
The traditional approach based on the limit equilib-
general limit equilibrium approaches, Morgenstern
rium methods (LEM) has found wide applications in
(1992) said that various LEM, at least for two-
slope stability analysis. The bench-mark reports given
dimensional analyses, are well understood. The impact
by Morgenstern (1992, 1991) and Duncan (1996) have
of new studies can be slight.
covered almost all important aspects with regard to
Indeed, new papers dealing with the improvement
the use of LEM. Latest development in this area may
of the analytical aspects of LEM have not been many,
not be substantial, but the improvement in enhancing
compared to the early stages of the 1960s or 1970s.
its efficiency and widening its applicability deserve a
Some conclusive remarks on the accuracies of various
special review. On the other hand, 3D LEM is an area
LEM have been made (Duncan, 1996). More papers
that has received wide attentions whose review is also
have focused on the search techniques for locating
worthwhile.
the critical slip surfaces, especially on the stochastic
The finite element method (FEM) offers an alterna-
approaches of optimization methods.
tive that is more rigorous, being free of the assumptions
regarding static indeterminacy, and descriptive, being
2.2 The generalized formulations
able to provide both stress and deformation infor-
mation. FEM normally does not give an explicit A number of general formulations have been pro-
result for factor of safety, which limits its applicabil- posed in the literatures. The authors tried to fix various
ity in engineering judgments. The strength reduction LEMs in a unified framework. Among these research
finite element method (SRF), originally advocated outcomes (Zhu, et al 2003, Espinoza, et al 1994,
by Zienkiewicz, et al (1975), has received warm Li, 1992), we wish to briefly review the work by
response recently as it has similar theoretical back- Chen & Morgenstern (1983), whose analytical forms
ground and the associated factor of safety to those of will greatly facilitate the calculations by spread sheets
the conventional approach using LEM. (Chen. et al., 2008).

25
 The governing force and moment equilibrium
equations provided by Chen and Morgenstern are
 b
p(x)s(x)dx = 0 (1)
a
 b
p(x)s(x)t(x)dx − Me = 0 (2)
a

The conventional definition for factor of safety

reduces the shear strength parameters c and φ  by the
following equations.

ce = c/F (3)
tanφe = tan φ/F (4)

Eqs. (1) and (2) involve the following definitions:

dW
p(x) = sin(φe − α) + q sin(φe − α)
dx
dW
− ru · sec α sin φe + ce sec α cos φe
dx
dW
+η cos(φe − α) (5)
dx
s(x) = sec(φe − α + β)
  x 
dβ
× exp − tan(φe − α + β) dζ (6)
a dζ
 x
Figure 1. The generalized method of slices. (a) The failure
t(x) = (sin β − cos β tan α) mass, (b) Spencer method, (c) tanβ to be fixed at both ends.
a
 ξ 
dβ
× exp tan(φe − α + β) dζ dξ (7)
a dζ that is zero at x = a and x = b. Figure 1(c) is an
 example that adopts a sine function for f (x).
b
dW It is possible to find F and λ from Eqs. (1) and (2)
Me = η he dx (8) by iterations.
a dx

where α = inclination of the slice base; β = the incli-

2.3 The simplified methods
nation of the inter-slice force. dW/dx = weight of the
slice per unit width; q = vertical surface load; η = The various simplified methods in common use can
the coefficient of horizontal seismic force, he = the be derived from Eqs. (1) and (2) as follows.
distance between the horizontal seismic force and base
of the slice, ru = pore pressure coefficient. 1. Spencer, 1966
In order to avoid violating the Principle of comple- This method takes fo (x) = 0 and f (x) = 1 in
mentary shear stresses, Chen and Morgenstern (1983) Eq. (9), which means [refer to Figure 1(b)],
argued that β must be fixed at both ends of the sliding
mass. They suggested [Figure 1(b) and (c)]. dβ/dx = 0 (10)

tan β = fo (x) + λf (x) (9) Eqs. (1) and (2) thus respectively reduce to

in which f (x) is a linear function that allows the values s(x) = sec(φe − α + β) (11)
fo (a) and fo (b) to be equal to specified values of tan β at
x = a and x = b respectively. f (x) is another function t(x) = sin β(x − xa ) − cos β(y − ya ) (12)

26
  b 
The force and moment equilibrium equations (1) dW dW
and (2) are simplified as: cos α tan(φe − α) − ru sin φe
a dx dx
  b
 b
× sec(φe − α) dx + ce sec(φe − α)
p(x) sec(φe − α + β)dx = 0 (13) a
a 
 dW
b × cos φe − η Rd dx = 0 (20)
p(x) sec(φe − α + β) dx
a
which can be demonstrated to be identical to the
× (x sin β − y cos β)dx = Me (14)
original formulation given by Bishop.

2. U. S. Army, Corps of Engineers, 1967.

This method assumes that the inclination of the
inter-slice force of each slice is parallel to the aver- 2.4 Applicability of various simplified methods
age sloping of the slope surface, whose inclination 2.4.1 General remarks
is designated γa , The accuracies of various limit equilibrium methods
have been conclusively summarized by Duncan as
β = γa (15) follows:
• The Ordinary (Fellunius) Method is highly inaccu-
3. Low, J. III and Katafiath, 1960. rate for effective stress analyses of flat slopes with
The authors assumed that the inclination of the high pore-pressure;
inter-slice force of each slice is equal to the average • Bishop’s simplified method is accurate for all
of the slopings of the top and base of the slice. conditions;
• Factors of safety calculated by force equilibrium
(α + γ ) methods are sensitive with the assumed inclinations
β = β = (16) of the side forces between slices;
2 • Methods that satisfy all conditions of equilibrium
are accurate for any conditions.
4. Janbu, 1954
The simplified Janbu’s method assumes For years our profession has been puzzled by the fact
that Bishop’s simplified method always gives factors
of safety in good agreement with those that satisfy
β=0 (17) complete equilibrium conditions. Zhu (2008) found
that the omitted term on the summation of unbalanced
By specifying a particular value of β for each shear forces on the interfaces in Bishop’s simplified
slice, it is possible to solve for F in Eq. (1) for the approach can be set to zero if a particular shear force
methods 2, 3 and 4. distribution for inter-slices is assigned, which in the
5. Bishop 1952 meanwhile allows the force equilibrium condition to
This method is concerned with circular slip be satisfied.
surfaces whose center is taken to establish the
moment equilibrium equation. The Bishop’s sim-
plified method assumes β = 0 that makes 2.4.2 Illustrative examples
To illustrate the statements regarding the accuracies
of the LEM methods in Section 2.3, we present the
s = sec(φe − α) (18)
following two examples.
Example 1 Figure 2 shows a slope with simple
and geometry and material properties: c = 5 × 9.8 kN/m2 ,
 φ = 35◦ , γ = 1.7 × 9.8 kN/m3 . Table 1 summarizes
x
the factors of safety associated with different arc angle
t=− tan αdξ α of the circle and pore pressure coefficient ru . It can
a
 x be found that the Bishop’s simplified method in all
dy cases gives basically the same results of Spencer’s
=− dξ = −y = −R cos α (19)
a dξ method. The discrepancies of Sweden’s method, com-
pared to Spencer’s, increases rapidly as α and ru
where R is the radius of the circle. Substituting Eqs. enlarges. At α = 117.6 and ru = 0.6, the relative error
(18) and (19) into Eq. (2), we have is: (FS − FF )/FF = (1.381 − 0.769)/1.381 = 44.3%.

27
 Figure 3. Back analysis of the Huaihexin Dike.

Table 2. Geotechnical parameters for example 2.

Soil layer number γ (kN/m3 ) c (kPa) φ(◦ )

1 19.11 11.0 26.0

2 13.03 2.67 0
Figure 2. An illustrative example explaining the accuracies 3 17.12 27 2.84
of Bishop, Fellunius and Spencer methods.

Table 1. Factors of safety associated with various arc angle Table 3. Factors of safety associated with different values
α of the circle and pore pressure coefficient ru . of β.

ru α(◦ ) Fb FF Fs β(◦ ) 0.0 1.55* 5.0 10.0 15.0 20.0

0.0 117.6 3.020 3.009 2.544 F 1.013 1.027 1.070 1.147 1.232 1.341
95.2 2.614 2.608 2.322
81.1 2.451 2.446 2.245
70.9 2.371 2.368 2.216
63.2 2.332 2.329 2.209 compared with that obtained by Spencer’s method
0.2 117.6 2.444 2.444 1.953 that gives F = 1.027 and β = 1.55◦ . It can be found
95.2 2.121 2.122 1.820 that F varies with β considerably and the additional
81.1 1.994 1.995 1.782 moment equilibrium method is indeed necessary to
70.9 1.936 1.937 1.775 find a reasonable solution for F.
63.2 1.910 1.910 1.783
0.4 117.6 1.876 1.895 1.361
95.2 1.634 1.648 1.317 3 THE UPPER BOUND ANALYSIS
81.1 1.542 1.552 1.318
70.9 1.504 1.511 1.334
63.2 1.490 1.496 1.354 3.1 Sarma’s (the upper-bound) method
0.6 117.6 1.325 1.381 0.769 Sarma presented a method that divides the failure mass
95.2 1.157 1.195 0.814 into a number of slices with inclined interfaces. The
81.1 1.098 1.126 0.855 limit equilibrium condition has been applied to both
70.9 1.078 1.098 0.893 the base and inter-slice faces. This method is particu-
63.2 1.076 1.091 0.929 larly applicable to rock slopes as advocated by Hoek
NOTE: Fb , FF , Fs are factors of safety obtained by methods
(1983, 1987).
of Bishop, Fellunius and Spencer respectively. The original approach by Sarma (1979) is based on
the force equilibrium conditions (Figure 4), which has
complex recurrence formulations.
Example 2 Back analysis of the Huaihexin Dike Donald and Chen (1997) presented an identical
This example, shown in Figure 3, takes from a dike approach which is theoretically supported by the upper
case in which the authors tried to use different methods bound theorem and practically easy to handle.
to back analysis the failure (Chen, 1999). The strength This method starts with a kinematically admissible
parameters are shown in Table 2. velocity field, in which the slice moves in a direction
The factors of safety given by the approach of that inclined at a friction angle relative to its neigh-
satisfying the force equilibrium method only, associ- boring slice or the base (detailed discussion has been
ated with different input of β, are shown in Table 3, given in Chen (2008).

28
 Figure 5. Sketch for the analyses by the energy approach of
Sarma’s method in finite difference forms.

Figure 4. Schematic illustrations for Sarma’s method.

1. Formulations based on the finite differences

The velocity of a slice numbered i, designated
Vi can be determined by (refer to Figure 5)

Vi = κV1 (21)

where V1 is the velocity of the first slice. κ is Figure 6. Sketch for the analyses by the energy approach of
Sarma’s method in integral forms.
defined as
  x 

i j
sin(αil − φeil − θi ) dα
κ= (22) V = κ exp − cot(α − φe − θj ) dζ V1
j dζ
j=1 sin(αir − φeir − θi ) x0
(24)
θ is the angle of the velocity with reference to Eq. (23) then becomes
the positive x axis. The superscript j refers to the
variable on the interfaces, and l and r refer to the
 xn 
left and right sides of the interfaces. (ce cos φe − u sin φe ) sec α
x0
The factor of safety, based on Eqs. (3) and (4) is 
obtained by the work-energy balance equation, dW
− sin(α − φe ) E(x)dx
dx

n  xn
κ[(ce cos φe − u sin φe ) sec α x − (cej cos φej − uj sin φej )L csc(α − φe − θj )
i=1 x0

− W sin(α − φe )]i dα
× E(x)dx + Ki = 0 (25)
dx

n−1
− κ(cej cos φej − uj sin φej )i where Ki is a coefficient accounting for possible
i=1 discontinuities in α, φe and ce .
× csc(α r − φer − θj )i sin(α − φe )i Li = 0 
n
Ki = − (cej cos φej − uj sin φej )i
(23) i=1

× Li csc(α r − φer − θj ) sin(α − φe )i E l (xi )

The first term of the left-hand side of Eq. (23)
refers to the work done by the external loads and (26)
the energy dissipation on the slip surface, while
the second term is the energy dissipation on the
3.2 The optimization process
interfaces between two contiguous slices.
2. Formulations based on integrals The conventional limit equilibrium methods and the
Eq. (21) can be transformed to an integral if upper-bound method (Donald and Chen, 1997) include
the width of the slice approaches to infinitesimally an optimization process that finds the critical failure
small (Figure 6), mode associated with the minimum factor of safety.

29
Early research work, such as Chen and Shao’s (1988), problem and demonstrated that Eq. (25) is reducible
has been continued recently by a number of researchers to the closed-form solution.
(Goh, 1999, Pham and Fredlund, 2003, Cheng, et al., Figure 9 shows an example with the material prop-
2003, Sarma and Tan, 2006). erty parameters c = 98 kPa, φ = 30◦ . The inclination
The slip surface is discretized into a number of of the slope surface is γ  = 45◦ , the bearing capac-
nodal points that are connected by either smooth ity q calculated by the closed form solution (refer to
curves or straight lines designated A1 , A2 , . . . , A6 Chen, 2008) is 10921.1 kPa. In Figure 9(a) the initially
(Figure 7). The variables defining the failure mode guessed slip surface is represented by 5 nodal points
includes the co-ordinates of the nodal points and connected by straight lines. The inclinations of the
the inclinations of the interfaces if the upper-bound interfaces are set arbitrarily. The factor of safety given
method is adopted. The optimization method will find by Eq. (23) is 1.047. Figure 9(b) shows the critical
these variables that give the minimum factor of safety mode associated with F = 1.013. Figure 9(c) shows a
designated B1 , B2 , . . . , B6 . more accurate result that employs 16 nodal points with
F = 1.006. It can be seen that the upper-bound method
gives an accurate result both in terms of the factor of
3.3 Test examples safety and the critical failure mode, compared to the
slip-line field method.
3.3.1 Theoretical verifications
A series of test problems based on the closed-
form solutions provided by the slip-field method 3.3.2 Comparisons with the conventional methods
(Sokolovski, 1960) has been performed using the Test examples have also shown that the upper-bound
numerical approaches described in this Section method is also able to give comparative results of factor
(Donald and Chen, 1997; Chen, 1999). The results of safety to those obtained by the conventional meth-
showed good agreements, demonstrating that the ods. The following two examples are taken from the
upper-bound method approach is more rigorous than ACADS slope stability programs review by Donald
the conventional method of vertical slices. As an exten- and Giam (1992) from which one may find the details
sion, this method has been successfully applied to including the material and geometry parameters.
the calculation for bearing capacity analysis, in which Example 4 The ACADS test example EX1(a)
the conventional method is generally not applicable For a simple test example shown in Figure 10
(Wang et al, 2001). This means that various empirical the ‘referee answer’ based on the simplified Bishop’s
coefficients involved in the conventional approaches
accounting for the effect of soil weight, embedment of
footing, complicated ground heterogeneities and water
conditions are no longer necessary.
Example 3 Comparisons with the closed-form
solution
Figure 8 shows a uniform slope subjected to a ver-
tical surface load q. The weight of the soil mass is
neglected. The closed-form solution for the ultimate
vertical surface load has been provided by Sokolovski
(1960). Chen (2008) gives a detailed description of the

Figure 8. Verifications of Eq. (24) compared to the the-

oretical solution provided by the slip-line field method,
example 3.

Weak seam

Figure 7. The optimization process for locating the critical Figure 9. Example 3, an example describing the upper
failure mode. bound approach.

30
method for the critical slip surface is 1.00. The upper 3.4 Practical considerations with Sarma’s method
bound method defined 4 nodal points designated A, B,
The experience of using Sarma’s method shows that
C, D with arbitrary interface inclinations as shown in
the following two issues are frequently encountered,
the Figure 10(a). Factor of safety for this initial failure
which require proper treatments.
mode is 1.304. Figure 10(b) shows the critical failure
mode associated with a minimum F of 0.997, which 1. The alternative directions of shear force or relative
is very close to the ‘referee answer’ velocity on the inter-slice surface
Example 5 The ACADS test example EX1(c) It has been shown that there are two possi-
Using the similar algorithms the factor of safety ble directions for a relative velocity between two
for the initial failure mode was 1.630 as shown in contiguous slices. The conventional method only
Figure 11(a), and the minimum F for the critical fail- considers the condition that the left slice moves
ure mode shown in Figure 11(b) was 1.401. The upper upward relative to the right one, as shown in Fig 7.
bound results can be compared with those given by the However, it is sometimes likely that the left slice
conventional methods shown in Figure 12. The slip moves down ward relative to the right one. Failure
surfaces 1, 2, 3 are related to the methods of Spencer, to identify this alternative may occasionally yield
Bishop, and Sarma respectively. wrong results.
Chen (2008) illustrated that the limit equilibrium 2. Treatment when tension develops on the interfaces
methods with the vertical and inclined slices can be and/or the base of a slice
approximately fixed in the theoretical framework of Sarma’s method assumes that shear failure
the lower and upper theorems of Plasticity. develops along the slip surface and the interfaces.
However the calculated results may show some
tensile internal forces, which is contradictive to
the original assumptions. In his program SARMA,
Hoek gives a warning but no solution is offered. An
approximate treatment is proposed by Chen (2008).
Details working on the two issues deserve a spe-
cial paper which is contained in this Proceedings
Figure 10. The ACADS test example EX1(a). (a) The initial (Chen, 2008).
failure mode; (b) the critical one.

4 THE 3D LIMIT EQUILIBRIUM

AND UPPER BOUND ANALYSIS

4.1 3D analysis based on the ‘method of columns’

In their keynote and state-of-the-art reports,
Morgenstern (1992) and Ducan (1996) advocated
the importance of the development of 3D limit
equilibrium methods for slope stability analyses.
A great number of papers on 3D slope stability anal-
ysis methods have emerged during the subsequent
Figure 11. The upper bound solutions for ACADS test 10 years (e.g., Stark and Eid 1998, Chen et al.
example EX1(c). (a) The initial failure mode; (b) the critical 2001, Huang and Tsai 2000, Jiang and Yamagami
one. 2004). All these papers have dealt with the ‘method
of column’ that can be considered to an extension
of the ‘‘method of slices’’ in the two-dimensional
area (Figure 13). On the other hand, the method
employing columns or blocks with inclined inter-
faces have been developed (Michalowski, 1989, Chen
et al. 2001a,b, Farzaneh and Askari 2003), which
can be regarded as an extension of the 2D Sarma’s
method.
Figure 12. Comparisons of the critical slip surfaces Review of the development of 3D limit equilib-
obtained various methods. (1) Spencer, F = 1.366; (2) rium and upper bound analysis deserves a full paper,
Bishop, F = 1.378; (3) Sarma, F = 1.401. and indeed has been tried by Chen et al. (2006) in

31
Figure 13. Sliding mass consisting of prisms with vertical Figure 14. The rock wedge failure ρl , ρr = the dilatant
interfaces. angles. The subscripts ‘l’ and ‘r’ stand for the left and
right planes respectively. For other parameters, refer to Chen
(2004).
their keynote paper of GeoShangai. Therefore no more
elaboration will be made in this Paper.

4.2 The generalized solution to rock wedge analysis

Among a variety of three-dimensional stability prob-
lems of slopes, wedge is a special and also the simplest
case that requires a special study.
It has been found that the limit equilibrium
approach commonly used for tetrahedral rock wedge
stability analysis actually involves an assumption that
the shear forces applied on the failure planes are par-
allel to the line of intersection. It is because of this
assumption that makes the solutions for normal forces
applied on the left and right failure planes possible, as Figure 15. Factors of safety associated with various shear
depicted in Figure 14. force directions on the failure planes.
To illustrate the impact of this finding, Chen took an
example that has a symmetric geometry and material
properties with respect to the line of intersection. The 5 FACTORS OF SAFETY BY THE FINITE
cohesion of the two failure surfaces is set to zero. The ELEMENT METHODS
angle between the line of intersection and the shear
force applied on the failure surface is denoted by γ . 5.1 Definition of the factor of safety
For this symmetric wedge with simple geometry, it
The finite element method normally gives information
is possible to establish a formulation to calculate F
of stress and strain fields. Various approaches have
associated with different values of γ . The case γ = 0◦
been proposed to transfer them to the factor of safety
corresponds to the conventional method and gives a
that is a common concern in engineering practice.
value F = 0.727. However F increases as γ becomes
larger and eventually reaches a maximum of 1.002 at 1. Based on the stress levels
γ = 42.5◦ , as shown in Figure 15. Keep σ3 of an element unchanged, draw a circle
A new method that allows an input of various that is tangent to the Mohr-Coulomb’s failure enve-
shear force directions has been presented by Chen lope with a corresponding diameter (σ1 − σ3 )f and
(2004). The controlling equation is reducible to the the stress level (σ1 − σ3 )/(σ1 − σ3 )f . The factor of
conventional solution and permits a formal demon- safety is defined as
stration to confirm that when ρl = φel and ρr = φer

factor of safety will obtain its maximum. Chen also dl
discussed the theoretical implications of these find- FFE1 =  (27)
σ1 − σ3
ings regarding some fundamental understanding in dl
Plasticity. (σ1 − σ3 )f

32
 where the integral represents the scalar summations
along a potential slip surface
2. Based on the shear stress on an element
For a stress state σx , σy , τxy , the normal and
shear stresses σn and τ on the slip surface can be
determinated by:

1
τ= (σy − σx ) sin 2α + τxy cos 2α (28)
2
σn = σx sin2 α + σy cos2 α − τxy sin 2α (29)
Figure 16. Factors of safety associated with various shear
force directions on the failure planes.
where α the inclination of the slip surface to the x
axis.
Table 4. Comparisons of factors of safety obtained by
The shear strength that can be developed along different definitions.
the slip surface is
Factor of safety

τf = c + σn tan φ 
(30) No. Types FFE1 FFE2 FFE3

The factor of safety along the entire slip surface 1 Embankment by layers 1.001 1.001 1.001
2 One-layer embankment 1.000 1.003 1.000
can be defined as 3 Excavation 1.082 1.044 1.044

(c + σn tan φ  )dl
FFE2 =  (31)
τ dl 6 THE STRENGTH REDUCTION
FINITE ELEMENT METHOD
3. Based on the weighted stress levels
This approach defines the factor of safety by the 6.1 The advances SRF has made
following equation: SRF was used for slope stability analysis as early as
1975 by Zienkiewicz et al. In this method, the shear
 strength parameters are reduced by Eqs. (3) and (4),
(c + σn tan φ  )dl
FFE3 =  (32) allowing the evolution of large area of plastic yielding.
σ1 − σ3
(c + σn tan φ  )dl Firstly, a gravity turn-on is implemented under elas-
(σ1 − σ3 )f tic state to determine the initial stress distribution
inside the slope. Then, stresses and strains are calcu-
lated by the elasto-plastic finite element method. The
5.2 Search for the critical slip surface shear strength reduction factor, F, is then increased
incrementally until the global failure of the slope
Having given the definition of the factor of safety, a
reaches, which means that the finite element calcula-
search technique, similar to that commonly used in
tion diverges under a physically realistic convergence
the limit equilibrium analysis area, can be employed
criterion (Refer to 6.3.1).
to find the critical slip surface associated with the min-
In a benchmark paper, Griiffiths and Lane (1999)
imum factor of safety (Zhou et al., 1995). Donald et al.
provided a series of test examples that show good
(1985) used an algorithm called CRISS to calculate the
agreements both in terms of factor of safety and plastic
factor of safety of a slope taken from ex1(a) of the test
zones with the conventional LE method.
problems issued by ACADS.
SRF can be a powerful alternative to the tradi-
Example 5 Reevaluations of Example 2 by FEM
tional limit equilibrium methods. This technique has
The referee answer based on the conventional limit
also been adopted in some well-known commercial
equilibrium method by the ACADS review program
software, such as FLAC, for practical applications.
Ex1(a) is 1.00 associated with a critical slip surface
The main advantages of the SRM can be summa-
passing through the toe of the slope as shown in
rized as follows:
Figure 16.
Table 4 summarizes the associated minimum fac- • It requires no assumptions which have been com-
tors of safety. It can be found that the results are close monly involved in LEM.
to one other. • The critical failure surface is found automatically.

33
• It offers much more detailed information such as and Lane, 1999; Cheng et al, 2007) has confirmed
the plastic zone, stress and deformation field, etc., good agreements between LEM and SRF in terms of
compared to LEM. factor of safety based on the definitions given by Eqs.
• It is possible for SRF to include piles and anchors (1) and (2). In this Proceedings, Duan et al. (2008)
that produce the coupled stress fields for soil and reported the calculated results by SRF for the gravity
structure simultaneously (Cai et al. 1998; Cai & dam stability problems with weak seams, which are
Ugai 2000). briefly summarized here.
Example 6 An example that compares the results of
Perhaps, the most important contribution of SRF is FEM and Sarma’s method (Duan, et al., 2008).
that it makes geotechnical calculations by FEM self- Figure 17 shows the geometry of the example with
checkable and reproducible. Since a large-scale non- material properties listed in Table 5. For the analysis
linear finite element analysis involves complicated along the weak seam ABC, Sarma’s method gave a
constitutive equations and iterations, it always happens factor of safety F = 2.12.
that different computer programs cannot give same Figure 18 shows the plastic zones from which one
analytical results. Lack of unique and widely accepted may find that the plastic zones along the weak seam
solutions has discouraged the extensive use of FEM extend as FOS increases. At F = 2.20, yielding domi-
in geotechnical practice. Now SRF can be a tool to nates throughout the seam with a plastic zone near the
test the applicability of an EEM program which, as dam toe shown in Figure 17, compared to the result
reliable software, should provide comparative results F = 2.12 by Sarma’s method.
with LEM if SRF is performed using this program. Duan et al (2008) further investigated a case
where section BC no longer exists as shown in
6.2 An illustrative example Figure 17(b). The critical location is determined by
an automatic search process in Sarma’s method. The
A number of research work (Naylor, 1981; Donald results obtained by Sarma and SRF were also in good
and Giam, 1992; Matsui and San, 1992; Ugai and agreement.
Leshchinsky, 1995; Dawson et al, 1999; Griffiths
6.3 General issues with SRF
6.3.1 The failure criteria
Dam
There have been a number of criteria that define
failure at which the calculation by SRF terminates:
(1) nonconvergence of the numerical process; (2) rapid
A C increase of displacement at some critical points; or
B (3) development of basically continuous plastic zones.
Bedrock Seam The value of F at this moment is believed to be the
D
solution for factor of safety of this problem.
(a) Experience has shown that these criteria do not lead
to substantially different values of F.
Dam
6.3.2 The constitutive laws
In SRF, both associated and non-associated elasto-
A plastic constitutive models can be adopted. The Mohr-
Bedrock Coulomb yield criterion is used to define the yield
Seam function if non-associated flow law is used.
D
1
f = −c cos φ  − I1 sin φ 
3
Figure 17. An example that compares the results of FEM  1

and Sarma’s method, the calculation by Sarma. + J2 cos  − sin  sin φ  (33)
3

Table 5. Geotechnical properties for example 6.

Material Density (kN/m3 ) Modulus (GPa) Poisson ratio Friction angle Cohesion (MPa) Tensile strength (MPa)

Dam 24.0 20 0.17 35 2.0 1.85

Bedrock 25.6 10 0.26 35 1.3 0
Seam 18.0 2.5 0.35 19.8 0.115 0

34
where Eq. (37) was derived on the assumption that c is
 √  negligible and the overburden h is very big. Duan
1 −1 3 3 J3 π π et al. (2008) gave a more generalized criterion,
= sin − , − ≤≤
3 2 J23/2 6 6
sin(φ + α)/ cos α ≥ (1 − 2μ) (38)
(34)
where α is defined by
and the Drucker-Prager criterion is normally adopted
to define the plastic potential function 2c/γ h
sin α = 
2 (39)
 2c
g = −αI1 + J2 − κ (35) (1 + K)2 +
γh
where
K is approximately taken to be coefficient of
tan  3c earth pressure at rest, which can be taken as:
α= √ , κ= √ (36)
9 + 12 tan2  9 + 12 tan2  μ
K= (40)
1−μ
In the above equations c, φ, and  are the effective
cohesion, friction angle, and dilatant angle, respec- The detailed work is documented in a paper of
tively. I1 , J2 , and J3 are the first invariant of the this Proceedings (Duan et al., 2008).
effective stress, and the second and third invariants 2. Young’s modulus
of the deviatoric stress, respectively. From physical point of view, it is obviously
advantageous to reduce both E and μ.
Duan et al. (2008) also suggested a hyperbolic
6.3.3 Treatment of other parameters stress strain relationship, similar to that proposed
The normal practice of SRF reduces the shear strength by Duncan and Chang (1970) to derive the crite-
parameters during calculations while keeping other rion for the reduced Young’s modulus. Figure 19
parameters constant. The necessity of treating these illustrates how the modulus can be reduced based
unchanged parameters has been discussed. on the reduction of the strength envelop.
1. The Poisson ratio As a matter of fact, the results shown in
Zheng et al. (2002) found that if c and tan φ Example 4 and Figure 20 are based on the reduced
are reduced considerably while Poisson ratio is values of μ. If μ keeps unreduced, a large area of
still kept unchanged, it is likely that an element
will inevitably yield, which is unrealistic. An
approximate condition was suggested:

sin φ ≥ 1 − 2μ (37)

where μ is the Poisson ratio. Therefore a bet-

ter approach can be reducing the Poisson ratio
simultaneously with the reduction of c and tan φ.

Figure 19. The concept of reducing Young’s modulus E

based on the hyperbolic stress strain relationship, the sub-
Figure 18. An example that compares the results of FEM script ‘o’ and ‘r’ stand for the original and reduced variables
and Sarma’s method, the calculation by SRF. respectively.

35
 This paper also reviews the use of finite element
method for slope stability analysis with particular
attentions to the strength reduction approach (SRF).
Main findings are:

• In the 2D areas, good agreements between LEM and

Figure 20. The large area of plasticity at the bottom by using SRF can be found. SRF can thus provide a useful
unreduced values of μ by SRF.
approach for stability analysis;
• It would be advantageous to make appropriate
plastic zone would develop at the bottom of the reductions for other material parameters, such as
foundation, as shown in Figure 20. Although the Poisson ratio, Young’s modulus, etc. Some associ-
factor of safety obtained was still F = 2.12, this ated criteria have been proposed.
unrealistic stress distribution would limit the credit
and applicability of SRF.
3. The dilatant angles
A certain elasto-plastic constitutive law will be REFERENCES
employed in the nonlinear finite element calcula-
tions, which can be associative or non-associative. Bishop, A.W. 1955. The use of the slip circle in the stability
The latter means a dilatant angle other than the fric- analysis of slopes. Geotechnique 5(1):7–17.
tion angle can be assigned to an element. A number Cai, F., Ugai, K., Wakai, A. & Li, Q. 1998. Effects
of horizontal drains under rainfall by three-dimensional
of papers (e.g. Cheng et al., 2007) investigated the finite element analysis. Computers and Geotechnics
influence of different values of dilatant angles to 23:255–275.
the final calculated results. The general conclusion Cai, F. & Ugai, K. 2000. Numerical analysis of the stability
is that adoptions of different values do not affect of a slope reinforced with piles. Soils and Foundations
the final solution to factor of safety substantially in 40(1):73–84.
the 2D analysis. Chen, L.H., Chen, Z.Y. & Sun, P. 2008. Slope stabil-
ity analysis using graphic acquisitions and spreadsheets.
Proceedings of the 10th International Symposium on
landslide and engineered slopes. Xi’an.
7 CONCLUSIONS Chen, Z.Y. & Morgenstern, N.R. 1983. Extensions to the gen-
eralized method of slices for stability analysis. Canadian
Geotechnical Journal 20(1):104–119.
This paper summarizes the recent advances in the tra- Chen, Z.Y. & Shao, C.M. 1988. Evaluation of minimum
ditional limit equilibrium and upper-bound methods, factor of safety in slope stability analysis. Canadian
referred to as the method of slices with vertical and Geotechnical Journal 25(4):735–748.
inclined interfaces respectively. Main findings in this Chen, Z.Y. 1999. Discussions: Prior and back stability
area are: analysis of the Huaihongxin Dike. Chinese Journal of
Geotechnical Engineering 21(4):518–519 (in Chinese).
• The analytical forms, such as Eqs. (1), (2) and Chen Z.Y., Wang xX.G., Haberfield C., Yin, J.H. &
Eqs. (13) and (14), can be derived for the gener- Wang, Y.J. 2001. A three-dimensional slope stability
alized method of slices. They are particular useful analysis method using the upper bound theorem-part I:
for calculation by a spread sheet. theory and methods. International Journal of Rock
• Theoretical studies and test examples have shown Mechanics & Mining Sciences 38(3):369–378.
Chen Z.Y., Wang J., Wang Y.J., Yin, J.H. & Haberfield C.
that various LEM basically yield factors of safety 2001. A three-dimensional slope stability analysis method
close to one another, all lie on the lower bound side. using upper bound theorem-part II: numerical approaches,
• Sarma’s method with the slices of inclined interfaces applications and extensions. International Journal of Rock
can be formulated by an upper-bound approach rep- Mechanics & Mining Sciences 38(3):379–397.
resented by Eq. (23) or (25). This method enjoys a Chen, Z.Y. 2004. A generalized solution for tetrahedral rock
sound mechanical background. Some treatments for wedge stability analysis. International Journal of Rock
the alternative inter-slice shear force directions and Mechanics & Mining Sciences 41:613–628.
internal tensions have been proposed. Chen, Z.Y., Yin, J.H. & Wang, Y.J. 2006. Keynote lecture:
• The three-dimensional limit equilibrium and upper- The Three-Dimensional Slope Stability Analysis: Recent
Advances and a Forward Look. Advances in Earth Struc-
bound methods with vertical and inclined columns tures, Research to Practice, Proceedings of Sessions of
have been made possible. As a special case, the Geoshanghai, ASCE Special Publication No. 151:1–42.
traditional rock wedge stability analysis method has Chen, Z.Y. 2007. The limit analysis in soil and rock: a
been generalized allowing different input of shear mature discipline of geomechanics. Journal of Zhejiang
force directions in the failure surfaces. University SCIENCE 8(11):1712–1724.

36
Chen, Z.Y. 2008. Some notes on the upper-bound and Hoek, E. 1987. General two-dimensional slope stability
Sarma’s methods with inclined slices for stability anal- analysis-Analytical and Computational Methods in Engi-
ysis. Proceedings of the 10th International Symposium neering Rock Mechanics. AlIen Unwin, London.
on landslide and engineered slopes. Xi’an. Huang, C.C. & Tsai, C.C. 2000. New method for 3D
Cheng, Y.M. 2003. Locations of critical failure surface and asymmetric slope stability analysis. ASCE. Jour-
and some further studies on slope stability analysis. nal of Geotechnical and Environmental Engineering
Computers and Geotechnics 30:255–267. 126(9):917–927.
Cheng, Y.M., Lansivaara, T.B. & Wei, W.B. 2007. Janbu, N. 1954. Application of composite slip surfaces for
Two-dimensional slope stability analysis by limit stability analysis. Proceedings of European Conference
equilibrium and strength reduction methods. Computers on Stability of Earth Slopes. 3:43–49. Sweden.
and Geotechnics 34:137–150. Jiang, J.C. & Yamagami, T. 2004. Three-dimensional slope
Dawson, E.M., Roth, W.H. & Drescher A. 1999. Slope stability analysis using an extended Spencer method.
stability analysis by strength reduction. Geotechnique Soils and Foundations 44(4):127–135.
49(6):835–840. Li, K.S. 1992. A unified solution scheme for slope stabil-
Donald, I.B., Tan, C.P. & Goh, T.C.A. 1985. Stability ity analysis. Proceeding, 6th International symposium on
of geomechanical structures assessed by finite element landslides. 481–487. Christchurch.
method. Proc. 2nd Int. Conf. In Civil Engr. Hangzhou, Lowe, J. III. & Karaflath, L. 1960. Stability of earth dams
845–856. Beijing: Science Press. upon drawdown. Proc. 1st Panamer. Conf. Soil Mech,
Donald I.B. & Giam, S.K. 1988. Application of the nodal 2:537–552. Mexico City.
displacement method to slope stability analysis. In: Pro- Matsui, T. & San, K.C. 1992. Finite element slope stability
ceedings of the fifth Australia—New Zealand conference analysis by shear strength reduction technique. Soils and
on geomechanics. 456–460. Sydney, Australia. Foundations 32(1):59–70.
Donald, I.B. & Chen, Z.Y. 1997. Slope stability analysis by Michalowski, R.L. 1989. Three-dimensional analysis of
the upper bound approach: fundamentals and methods. locally loaded slopes. Geotechnique 39:27–38.
Canadian Geotechnical Journal 34:853–862. Morgenstern, N.R. & Price, V. 1965. The analysis of the
Donald, I.B. & Giam, P. 1992. The ACADS slope stability stability of general slip surface. Geotechnique 15(l):
programs review. Proc. 6th International Symposium on 79–93.
Landslides. 3:1665–1670. Morgenstern, 1991. The evaluation of slope stability—a
Duan, Q.W., Chen, Z.Y., Wang, Y., Yang, J. & Shao, Y. 2008. 25 year perspective. Proc. ASCE Conf. on Stability
Applications of the strength reduction finite element and Performance of Slopes and Embankments, 1:1–26.
method to a gravity dam stability analysis. Proceedings Berkeley.
of the 10th International Symposium on landslide and Morgenstern, 1992. Keynote paper: The role of analysis
engineered slopes. Xi’an. in the evaluation of slope stability. Proceedings of 6th
Duan, Q. & Zhang, P.W. 2008. On the treatments for the Internatioanl Symposium of Landslides: 1615–1629.
deformation parameters in the strength reduction finite Naylor, D.J. 1981. Finite elements and slope stability. Numer.
element method. Proceedings of the 10th International Meth. In:Geomech., Proceedings of the NATO Advanced
Symposium on landslide and engineered slopes. Xi’an. Study Institute, Lisbon, Portugal:229–244.
Duncan, J.M. & Chang, C.Y. 1970. Nonlinear anal- Pham, H.T.V. & Fredlund D.G. 2003. The application of
ysis of stress and strain in soils. Journal of Soil dynamic programming to slope stability analysis. Cana-
Mechanics and Foundation Engineering Division, ASCE dian Geotechnical Journal 40:830–847.
96(5):1629–1653. Sarma, S.K. 1979. Stability analysis of embankments and
Duncan, J.M. 1996. State of the art: Limit equilibrium and slopes. Journal of the Geotechnical Engineering Division,
finite element analysis of slopes. Journal of Geotechnical ASCE 105(GT12):1511–1524.
Engineering 122(7):577–596. Sarma, S.K. & Tan, D. 2006. Determination of critical
Espinoza, R.D., Burdeau, P.L.P.C. & Mohunthan, B. slip surface in slope analysis.Geotechnique 56(8):
1994. Unified formulation for analysis of slopes 539–550.
with general slip surface. J. Geotech. Engng, ASCE Sokolovski, V.V. 1960. Statics of soil media. (Translated by
120(7):1185–1104. Jones DH and Scholfield AN). London: Butterworth.
Farzaneh, O. & Askari, F. 2003. Three-Dimensional Analysis Spencer, E. 1967. A method of analysis of embank-
of Nonhomogeneous Slopes. Journal of Geotechnical and ments assuming parallel inter-slice forces. Geotechnique
Geoenvironmental Engineering 129(2). 17:11–26.
Fellenius, W. 1927. Erdstatisch Berechnungen, Berlin Stark, T.D. & Eid, H.T. 1998. Performance of three-
W.Ernst und Sohn revised edition, 1939. dimensional slope stability analysis method in practice.
Goh, A.T.C. 1999. Genetic algorithm search for critical Journal of Geotechnical Engineering, ASCE 124:
slip surface in multi-wedgestability analysis. Canadian 1049–1060.
Geotechnical Journal 36(2):383–391. Tan, C.P. & Donald, I.B. 1980. Finite element calculation of
Griffiths, D.V. & Lane, P.A. 1999. Slope stability analysis by dam stability. Proc. 11th Int. Conf. Soil Mech. and Fnd.
finite elements. Geotechnique 49(3):387–403. Engr. San Francisco.
Hoek, E. & Bray, J. 1977. Rock slope engineering. The U.S. Army, Corps of Engineers. 1967. Stability of slopes and
Institute of Mining and Metallurgy. foundations, Engineering Manual, Visckburg, Miss.
Hoek, E. 1983. Strength of jointed rock masses. Geotech- Ugai, K. 1985. Three-dimensional stability analysis of verti-
nique 33(3):187–223. cal cohesive slopes. Soils and Foundations 25(3):41–48.

37
Ugai, K. & Leshchinsky, D. 1995. Three-dimensional limit Zhu, D.Y., Lee, C.F. & Jiang, H.D. 2003. Generalized frame-
equilibrium and finite element analysis: a comparison of work of limit equilibrium methods for slope stability
results. Soils and Foundations 35(4):1–7. analysis. Geotechnique 53(4):377–395.
Wang, Y.J., Yin, J.H. & Chen, Z.Y. 2001. Calculation Zhu, D.Y., Dun, J.H. & Tai, J.J. 2007. Theoretical verifi-
of bearing capacity of a strip footing using an upper cation of rigorous nature of simplified Bishop’s method.
bound method. International Journal for Numerical and Chinese Journal of Rock Mechanics and Engineering
Analytical Methods in Geomechanics 25:841–851. 26(3):455–458 (in Chinese).
Whitman, R.V. & Bailey, W. 1967. Use of computers for Zhu, D.Y. 2008. Investigation on the accuracy of the simpli-
slope stability analysis. Journal of Soil Mechanics and fied Bishop method. Proceedings of the 10th International
Foundation Engineering Division, ASCE 93(SM4). Symposium on landslide and engineered slopes. Xi’an.
Wright, S.G., Kulhawy, F.H. & Duncan, J.M. 1973. Zienkiewicz, O.C. Humpheson, C. & Lewis R.W. 1975. Asso-
Accuracy of equilibrium slope stability analysis. Journal ciated and nonassociated visco-plasticity and plasticity in
of Soil Mechanics and Foundation Engineering Division, soil mechanics. Geotechnique 25(4):671–89.
ASCE 99(SM10):783–791. Zou, J.Z., Williams, D.J. & Xiong, W.L. 1995. Search for
Wright, S.G. 1978. Slope stability analysis. Proceedings critical slip surfaces based on finite element method.
on Analysis and Design in Geotechnical Engineering, Canadian Geotechnical Journal 32:233–246.
Vol. 2. 153. Zheng, H. & Li, C. 2002. Solution to the factor of safety by
Yamagami, T. & Ueta, Y. 1988. Search for critical slip finite element method. Chinese Journal of Geotechnical
lines in finite element stress fields by dynamic program- Engineering 24(5):626–628 (in Chinese).
ming. Proceedings of the 6th International Conference
on Numerical Methods in Geomechanics. Innsbruck,
Australia: 1335–1339.

38
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Improving the interpretation of slope monitoring and early warning data

through better understanding of complex deep-seated landslide
failure mechanisms

E. Eberhardt
Geological Engineering, University of British Columbia, Vancouver, Canada

A.D. Watson
BC Hydro, Burnaby, British Columbia, Canada

S. Loew
Engineering Geology, ETH Zurich, Switzerland

ABSTRACT: The past several years have seen significant advances in landslide monitoring technologies.
Remote sensing techniques based on satellite and terrestrial radar can now provide high-resolution full area
spatial coverage of a slope as opposed to relying on geodetic point measurements. Automation in the form of
wireless data acquisition has enabled the collection of data with increased temporal resolution. These tools provide
increased capacity to detect pre-failure indicators and changes in landslide behavior. Yet the interpretation of
slope monitoring data, especially that for early warning, still remains largely subjective as geological complexity
and uncertainty continue to pose major obstacles. This paper reviews several recent developments in landslide
monitoring techniques but questions the phenomenological approach generally taken. Examples are then provided
from several recent experimental studies involving ‘‘in situ laboratories’’ in which detailed instrumentation
systems and numerical modeling have been used to better understand the mechanisms controlling pre-failure
deformations over time and their evolution leading to catastrophic failure. Preliminary results from these studies
demonstrate that by better integrating the different data sets collected, geological uncertainty can be minimized
and better controlled with respect to the improved interpretation of slope monitoring and early warning data.

1 INTRODUCTION Comprehensive reviews of these methods are pro-

vided by Bhandari (1988), Glastonbury & Fell (2002),
Monitoring forms a key component of most landslide Crosta & Agliardi (2003) and Rose & Hungr (2007).
hazard assessments, providing data that may be used to Inherently, these approaches are ‘holistic’, disregard-
quantify the nature of the hazard, its extent, kinematics ing details pertaining to the underlying slope failure
and stability state, sensitivity to triggering mecha- mechanism. Whether the displacement measurements
nisms, response to mitigation works, etc., or to provide are made using an extensometer positioned across a
early warning of an impending failure especially those tension crack or a system of geodetic reflectors across
where lives or infrastructure may be at risk. In both a slope, the analysis is often carried out in the same
cases, issues of uncertainty relating to the geological way—surface displacements are recorded over time,
conditions, slope kinematics and failure mode provide which are then extrapolated or analyzed for accel-
major obstacles that contribute to a lack of definition erations in order to predict catastrophic/impending
of the problem. failure. Generally, the kinematics and causes of failure
Techniques used for forecasting impending failure are not well defined, and instead, the surface manifes-
(i.e. temporal prediction) are largely phenomenolog- tation of the instability (i.e. surface displacements) is
ical, relying on surface-based point measurements of relied upon for predictive analysis. Not surprisingly,
displacement monitored over time, which are then only a few cases have been reported where these tech-
extrapolated or analyzed for accelerations that exceed niques have been successfully applied as part of a
set thresholds based on earlier patterns. Fukuzono’s reliable forward prediction (e.g. Rose & Hungr 2007);
(1985) inverse velocity method is one such example. most involve back analyses.

39
 Numerical modeling offers a means to account for movements can be measured to several millimeters
complex subsurface processes by breaking problems accuracy (Froese et al. 2005). Temporal decorrela-
down into their constituent parts and analyzing the tion due to vegetation coverage, however, dramat-
cause/effect relationships (and their evolution), which ically affects interferometric coherence and limits
govern the behavior of the system as a function of the detection resolution, although multi-image based
changing environmental factors. However, these anal- approaches like the Permanent Scatterers (Ferretti
yses require tight controls on the representation of et al. 2001) and Small Baseline (Berardino et al. 2002)
geological heterogeneity and structure, soil and/or methods work to limit spatial decorrelation effects and
rock mass behavior, and special boundary conditions. topography errors.
This information may be derived in part from sur- Several studies have now been published where
face and borehole data, but more often it is limited to coherence for a study site could be maintained and
subsurface projections based on surface observations. DInSAR successfully applied. Most of these are from
Slope monitoring data provides an important means slides showing coherent movements over larger areas,
to calibrate and constrain detailed numerical models. often involving rock masses or colluvium with strong
At the same time it must be recognized that most plastic deformations (e.g. Rott et al. 1999, Froese
in situ measurements are affected by the same issues et al. 2005, Singhroy et al. 2005). Figure 1a shows
of rock mass complexity and variability as the numer- deformations detected along a major fault and over
ical analyses they are meant to constrain. In many old mine workings and colluvium towards the bottom
situations the interpretation of monitoring data is far
from straight forward. In turn, it has been demon-
strated that numerical modeling can be used to help
constrain interpretations of complex field measure-
ments (Eberhardt & Willenberg 2005, Watson et al.
2006). Thus it must be emphasized that a counter-
balance and close association should exist between
field measurements and analysis to develop a more
complete understanding of the slope hazard problem
(Sakurai 1991).

2 CONTINUING DEVELOPMENTS
IN LANDSLIDE MONITORING

2.1 Remote sensing and InSAR

Although technological advancements continue with
respect to traditional total station and prism monitoring
systems (e.g. combined robotic total station and global
navigation satellite systems; Brown et al. 2007), many
of the key recent advances in slope monitoring are
those related to remote sensing technologies, primar-
ily satellite and ground-based radar but also non-radar
variants like airborne and terrestrial laser scanning
(e.g. Rosser et al. 2005).
At the regional scale, Differential Interferometric
Synthetic Aperture Radar (DInSAR) is proving to be
a useful means for identifying landslides within large
coverage areas (e.g. 100 × 100 km using ERS data;
Meisina et al. 2005) to help in the development of
landslide inventories. InSAR uses satellite emitted
electromagnetic signals to measure the phase differ-
ence resulting from the path length change between
satellite passes of the same area taken from the same
flight path. The difference in phase can be used to Figure 1. a) Vertical surface deformation map based on
determine ground movement in the line of site of satellite DInSAR for Turtle Mountain for a 2-yr period (after
the SAR satellite, and with an emitted electromag- Singhroy et al. 2005). b) Line of sight displacements derived
netic wavelength of a few centimeters, these ground from terrestrial radar for Randa for a 59-day period.

40
of Turtle Mountain in western Canada, location of the long-term in situ monitoring efforts. However, new
1903 Frank Slide (Singhroy et al. 2005). These drew technologies like fiber optics are also being tested to
attention to a possible link between slope movements capitalize on the stability and insensitivity of fiber
in the upper slope and de-stressing of the slope’s toe optic sensors to external perturbations. A review of
due to the slow collapse of the old workings (Froese new developments in fiber optic sensing technologies
& Moreno 2007). In comparison, brittle rock masses for geotechnical monitoring is provided in Inaudi &
showing smaller-scale complex block movements are Glisic (2007).
more difficult to study with satellite-based DInSAR Data reliability is equally a key issue. Recent studies
due to the strong spatial variability in block veloci- involving cases where the displacements being mea-
ties. Instead, ground-based DInSAR (e.g. Tarchi et al. sured are particularly small or where deep inclinome-
2003) with a typical pixel size of a few meters may ters are involved, have led to improved algorithms and
prove more suitable. Figure 1b shows preliminary procedures for identifying and correcting systematic
results from Randa in southern Switzerland, loca- errors (e.g. Mikkelsen 2003, Willenberg et al. 2003).
tion of the 1993 Randa rockslide, which are being Studies involving complex, deep-seated, rock slope
used to provide important information about the area instabilities have seen attempts to better integrate mul-
affected by slope movements and to support kinematic tiple geological and geotechnical data sets to improve
analyses. data interpretation (e.g. Willenberg 2004, Watson
et al. 2006, Bonzanigo et al. 2007, Hutchinson et al.
2007). These attempts at data ‘‘fusion’’ are moving
2.2 Slope Stability Radar (SSR)
towards the adoption of Virtual Reality (VR) technol-
Other new developments in the use of radar have ogy, where the identification of hidden relationships,
involved moving away from synthetic aperture radar discovery and explanation of complex data interdepen-
and instead using real-aperture from a stationary plat- dencies, and means to compare and resolve differing
form, as in the case of Slope Stability Radar (SSR). interpretations can be facilitated (Kaiser et al. 2002).
With SSR, the system is typically set up 50 to 1000 m Concomitant with data integration is data man-
from the foot of the slope and the region of interest agement. Important new elements include Web GIS
is continuously scanned, comparing the phase mea- services integrated into the operational resources of
surement in each image pixel with previous scans to decision makers. These services are linked to early
determine the amount of movement. (Fig. 2; Harries warning systems through wireless data acquisition
& Roberts 2007). The combination of near real-time and transmission technologies, which enable real-
measurement, sub-millimeter precision and broad area time data from multiple remote monitoring sites to be
coverage to quickly identify the size and extent of a accessed and viewed off-site by means of the internet.
developing failure is helping to establish SSR as a key This is proving to be a highly valuable resource where
tool for managing unstable rock slopes, especially in an unstable slope threatens a community, critical facil-
open pit mining (e.g. Harries et al. 2006, Little 2006, ity or, in the case of large open pit mine slopes, worker
Day & Seery 2007). Further advantages of the sys- safety. Furthermore, Hutchinson et al. (2007) propose
tem is that it is not adversely affected by rain, fog, that spatially and temporally distributed measurements
dust or haze. Vegetation on the slope, however, may should be combined with a knowledge engine and an
reduce the precision in pixels where there is low phase evolving rule base to form the hub of a decision support
correlation between scans (Harries & Roberts 2007). system.
Again, the advantage of these remote sensing-type One of the more comprehensive systems in place is
systems is that they provide fast and updatable data that installed by BC Hydro for its chain of hydroelec-
acquisition over broad areas as opposed to point mea- tric dams on the Columbia River in British Columbia,
surements, coincident with prism placement, when Canada (Fig. 3). The system connects dataloggers at
using traditional geodetic monitoring systems. six large landslide sites along different dam reser-
voirs to a central monitoring computer using radio
and microwave communication (Fig. 3c). The land-
2.3 Wireless data acquisition and data management
slides sites are separated by up to 150 km and range
Conventional slope monitoring (e.g. inclinometers, in size from less than one million to over a billion
total stations, tiltmeters, extensometers, crackmeters, cubic meters. These include: 731 Block, Checker-
etc.), continues to represent the favored means by board Creek, Downie Slide, Dutchmans Ridge, Little
which to measure and monitor slope deformation Chief Ridge and Little Chief Slide.
directions, magnitudes and rates, both on surface The extensive system in place is used for both inves-
and at depth. Instrument reliability is of paramount tigative and predictive monitoring, and involves the
importance, for which continuous improvement to the continuous monitoring of a large number of piezo-
performance of vibrating-wire technology has led to meters, in-place inclinometers, extensometers, water
it being widely recognized as the preferred choice for weirs, load cells and tiltmeters. The basis of the data

41
Figure 2. GroundProbe’s Slope Stability Radar (SSR) system showing the continuous monitoring of millimeter-scale
movements across the entire face of an unstable open pit mine slope (after Harries et al. 2006).

Figure 3. a) Example of the detailed instrumentation used by BC Hydro to monitor slope displacements, temperature and
pore pressures at depth for a rock slope, Checkerboard Creek, above one of their dam reservoirs (after Watson et al. 2007).
b) Monitoring and alarm interface used by BC Hydro to remotely monitor several reservoir slopes along their system of
hydroelectric dams on the Upper Columbia River. c) Schematics of the wireless data acquisition/transmission system used for
the remote monitoring system.

42
communication is a mix of UHF radio and spread- construction of a deep drainage adit that success-
spectrum radio which allow the dataloggers to com- fully led to the stabilization of the landslide (Fig. 4,
municate with the central monitoring computer. The Eberhardt et al. 2007). A second example is that pro-
fundamentals of the communication system include a vided by Watson et al. (2006, 2007) for BC Hydro’s
transmitter, receiver and surge arrestor which allows Checkerboard Creek (Fig. 3a). Again, detailed moni-
reliable communication over distances of a few kilo- toring of the slope revealed a similar persistent annual
meters, up to 100 kilometers with proper antennae. displacement cycle dominated by an active sliding
Advantages at these sites over satellite or cell phone phase in autumn to late winter and inactivity dur-
communication include cost and ease of use. Disad- ing the spring and summer. Although at first these
vantages include the requirement to use repeaters if periods of displacement activity appeared to corre-
line of sight on long distances is not possible. spond to periods of increased precipitation as in the
case of Campo Vallemaggia, it wasn’t until several
years worth of measurements were collected that it
was observed that the annual displacement cycle was
3 MONITORING OF LANDSLIDE BEHAVIOUR repeated each year regardless of the amount of pre-
cipitation. Instead, the annual displacement cycle
3.1 Current state of practice was more strongly correlated to seasonal tempera-
Landslide monitoring serves two important functions ture variations in the near surface bedrock and the
(Moore et al. 1991): deformation mechanism explained in terms of ther-
mally induced slip along sub-vertical joints (Watson
i. Investigative Monitoring: To provide an under- et al. 2006). In both cases, the assessments and sub-
standing of the slope and thus enable an appropriate sequent approaches taken were completed based on
action to be implemented. the results of investigative monitoring used to develop
ii. Predictive Monitoring: To provide a warning of a reliable geological and hydrogeological models later
change in behavior and thus enable the possibil- aided by numerical modeling to more fully understand
ity of limiting damage or intervening to prevent the deformation mechanisms involved.
hazardous sliding. Predictive monitoring systems usually evolve after
investigative monitoring and assessment. Without this
Instrumentation typically includes: piezometers, in- investigation phase it is unlikely that a predictive sys-
place inclinometers, extensometers/crack monitoring, tem could be designed, or early warning thresholds
tiltmeters and surface geodetic monuments. Often for set, with confidence. Monitoring information must be
large landslides the most reliable instruments are those assessed in the context of the physical setting and the
installed in deep subsurface exploratory boreholes conclusions of the investigation phase. Landslides can
(e.g. Fig. 3a). change their behavior within a few weeks or a few
Investigative monitoring can be used to obtain a days. Thus, for a predictive monitoring system to be
greater understanding of the slope behavior, thus effective the frequency at which the instruments are
enabling the correct approach to be taken or to con- monitored must be a fraction of this response time and
firm that the approach taken was correct. Monitoring the system must be in place to react to the instrumenta-
for investigative purposes can be as little as an annual tion results. Furthermore, in order for decision makers
visual inspection or as much as continuous measure- to be in a position to react correctly to predictive mon-
ment of a comprehensive instrumentation network itoring data, the chance of faulty alarms or misleading
(e.g. Fig. 3c). For the case of slow moving slides, instrument readings should be minimized by the use
many years of monitoring and annual cycles may be of the most reliable instrumentation possible.
required to identify relationships between water lev-
els, movement and other seasonal effects such as
3.2 In situ rockslide laboratories
temperature. This was demonstrated by Bonzanigo
et al. (2007) for the case of Campo Vallemaggia, Efforts to improve predictive monitoring have seen
an 800 million m3 deep-seated landslide in strongly several recent multi-disciplinary studies focused on
fractured and weathered crystalline rocks in southern improving our understanding of complex landslide
Switzerland that threatened two villages founded on deformation mechanisms. These include the Randa
the landslide. Detailed inclinometer and piezometer In Situ Rockslide Laboratory in Switzerland, the Tur-
measurements collected over a five year period were tle Mountain Field Laboratory in western Canada and
used to cross-correlate the stick-slip behavior of the the Åknes/Tafjord Project in Norway.
landslide with pore pressures exceeding a threshold The first of these, the Randa In Situ Rockslide Lab-
value tied to longer-term precipitation events (Fig. oratory (Fig. 5), was a comprehensive experimental
4). These measurements were subsequently used in investigation into the spatial and temporal evolution of
the decision making process to go forward with the large rock slope failures in fractured crystalline rock.

43
Figure 4. Correlation between the downslope velocities of the Campo Vallemaggia landslide and borehole pore pressures
measured before and after the opening of a drainage adit to stabilize the slope. Slide velocities were measured using an
automated geodetic station; pore pressures are expressed as the hydraulic head in the piezometer (after Eberhardt et al. 2007).

One of the prime motivating factors was to develop define the rock mass structure (e.g. Fig. 6), aided
a better understanding of rock mass strength degrada- in the positioning of borehole instruments (e.g. in-
tion (e.g. through the destruction of intact rock bridges place inclinometers) and was essential for reliably
between non-persistent discontinuities) and the pro- interpreting the monitoring data.
gressive development of internal shear zones, and their A similar multi-disciplinary field campaign was
accommodation of larger slope displacements, leading carried out for the Åknes/Tafjord Project, where up to
to increased extensional strains and eventually sudden 30–40 million m3 of unstable rock moving with a mean
collapse (e.g. Eberhardt et al. 2004b). Focus was also rate of 2–4 cm/year has been identified as a potential
placed on improving early warning capabilities in the threat and tsunami generating hazard for people and
presence of persistent and non-persistent discontinu- infrastructure living along the inner Storfjord (Blikra
ities, and multiple moving blocks and internal shear et al. 2005). Geological, geodetic and geophysical
surfaces. studies (including GPS, resistivity, georadar, reflec-
For this, a high-alpine facility was constructed tion and refraction seismics, airborne laser scanning,
above the scarp of the 1991 Randa rockslide (Fig. 5a), and high-resolution air photography) were carried out
where ongoing movements of 1–2 cm/year are being to define the geometry and volume of unstable areas.
recorded in gneissic rock for a volume of up to Through these detailed studies it was found that a pre-
10 million m3 . This facility included the installation of viously unmonitored section of the slide was moving
a variety of instrumentation systems designed to mea- at 15 cm/year, leading to revisions in the unstable vol-
sure temporal and 3-D spatial relationships between ume and thus the magnitude of the potential hazard
fracture systems, displacements, pore pressures and (Roth et al. 2006).
microseismicity (Fig. 5). The monitoring was com- The fracture network, both existing and newly gen-
plimented by a detailed geophysical field campaign, erated, was a central focus in the designs of the
which included 3-D surface seismic refraction and instrumentation networks at Randa, Åknes and Turtle
georadar surveys to resolve subsurface 3-D fracture Mountain. In different ways, displacement monitor-
distributions (Heincke 2005), and crosshole georadar ing targeted resolving the complex displacement field
and seismic tomography to identify key geological fea- generated by multiple moving blocks. At Turtle Moun-
tures (Spillmann 2007). These were then compared tain, a combination of crackmeters, wire-line exten-
to those mapped on surface and in the boreholes to someters and tiltmeters, were used to monitor surface
develop a 3-D geological model of the unstable rock tension cracks, enabling seasonal displacement pat-
mass (Willenberg 2004). This information helped to terns to be resolved (Moreno & Froese 2007). Similar

44
Figure 5. The Randa In Situ Rockslide Laboratory in southern Switzerland. a) Area of investigation (solid white line)
and outline of present-day instability (black dashed line) above the scarp of the 1991 rockslide. Photo by H. Willenberg.
b) Installation of surface and subsurface monitoring instruments and central data acquisition station housing batteries, power
generation sources (solar and wind) and data acquisition and transmission hardware. c) Plan view map showing location of
boreholes, geodetic reflectors and geophones relative to the active slide area and open tension cracks (after Willenberg 2004).

at Åknes and Randa. At Randa, the deepest incli-

nometer was also fitted with an Increx electromagnetic
induction sliding extensometer system to enable the
profile of 3-D displacement vectors at depth to be
determined. Biaxial vibrating wire in-place inclinome-
ters were installed and positioned at depth intervals
coinciding with key fractures identified through the
borehole televiewer surveys to provide continuous
monitoring of subsurface deformations along these
structures. Integration of these different displacement
data sets with the 3-D geological model showed that
the displacements recorded on surface and at depth
were localized across active discontinuities (Fig. 7)
and that the kinematic behavior of the slope was dom-
inated by complex internal block movements rather
than those of a coherently-sliding mass (Willenberg
Figure 6. Cross-section through the Randa study area 2004). Willenberg et al. (2003) demonstrated that to
showing the network of discontinuities (F-1 to F-3) and resolve these complex displacement patterns, rigorous
faults (highlighted in black) mapped using geological and correction algorithms must be carried out to attain the
geophysical methods (after Loew et al. 2007). requisite resolution.
The fracture network also plays a controlling role
monitoring was carried out at the Åknes and Randa with respect to the distribution of pore pressures at
project sites. This traditional focus on surface dis- depth and their coupled relationship with unstable
placement monitoring addresses certain logistic and rock slope movements. The design of the experimen-
economic realities in terms of what may be feasible tal monitoring networks at Randa, Åknes and Turtle
for on-site monitoring, yet it must also be asserted that Mountain each included borehole monitoring of pore
only so much can be inferred on surface for a prob- pressures to correlate with measured displacements.
lem that develops at depth. For this, deep inclinometer At Randa, piezometers were positioned and packed
measurements were added to the monitoring networks off along zones indicating potentially higher fracture

45
Figure 7. Integrated borehole data set for the 120-m deep borehole at Randa, showing from left to right: fracture frequency
log, optical televiewer log (highlighting traces of major fractures), cumulative inclination changes for a two-year period
(and corresponding preliminary kinematic interpretation), cumulative axial displacements for the same two-year period, and
corresponding 3-D displacement vector magnitudes and orientations (after Willenberg 2004).

permeability as determined from borehole televiewer ditions at ground surface is observed in landslides
data. These data showed several water tables dis- showing artesian pressures at depth.
tributed within the rock mass and different types of The final key monitoring strategy common to the
pore pressure interactions with infiltrating surface Randa, Åknes and Turtle Mountain projects was the
water and atmospheric pressure variations (Willenberg use of microseismic monitoring to detect and study
2004). This is a common feature for most deep-seated subsurface brittle fracture processes. Spatially clus-
slides in crystalline rock, where preferential fracture tered microseismic events in numerous fields (e.g.
permeability and hydraulic barriers (e.g. from fault mining, geothermal energy, nuclear waste disposal,
gouge) result in isolated compartments of groundwater etc.) have proven effective in providing critical infor-
flow and reaction delays between surface precipita- mation with respect to stress-induced tensile fracturing
tion and pressure responses at depth, making corre- mechanisms and/or shear slip along internal fracture
lations between slope movements and precipitation planes. The microseismic network at Randa was the
events extremely difficult (e.g. Moore & Imrie 1992, most detailed of the three and included three geo-
Bonzanigo et al. 2007). A strong sensitivity of slope phones (28 Hz) mounted in deep boreholes, nine
movements to changing groundwater recharge con- geophones (8 Hz) mounted in shallow boreholes, and

46
 Figure 9. Cross-section showing the relationship between
microseismic activity (high PDF values indicate microseis-
Figure 8. Vertical components of a locatable microseismic mogenic zones), faults (dashed lines) and the approximate
event: a) raw; b) 100–500 Hz bandpass filtered signals. Sig- limit of slope movements. The cumulative PDF represents
nals are sorted according to the source-receiver distance, with the sum of the hypocenter probability density functions for
sen-sor A1 being the farthest and B5 the closest. Absolute all microseismic events. After Spillmann et al. (2007).
time scale is arbitrary. After Eberhardt et al. (2004a).

two 24-channel seismographs (Fig. 5c). The spatial

distribution of the twelve triaxial geophones was cho-
sen to ensure that the hypocenter parameters generated
from the seismic sources could be reliably constrained
within the area of interest (Spillmann et al. 2007). One
of the initial findings from this system was that the
higher frequency content of the recorded microseismic
events was strongly attenuated (Fig. 8), pointing to the
presence of large open fractures at depth. This was
fully compatible with the geological model and bore-
hole televiewer and pore pressure data (Willenberg
et al. 2004). Larger low frequency events, such as those
generated from natural seismic activity in the region,
did not suffer as much from signal quality degrada- Figure 10. UDEC modeled thermal-induced horizontal
tion. Interpretation of the recorded local microseismic displacements compared to inclinometer measured slope
activity involved comparing the cumulative hypocen- displacements (after Watson et al. 2006).
ter probability density functions, which incorporates
uncertainties in the arrival times and travel times for
each event (see Spillmann et al. 2007), with the results local earth tremors and rock falls, together with seis-
from the geological, geophysical and geotechnical micity believed to be related to deformation and stress
investigations. From this, it was observed that the relief within Turtle Mountain and the ongoing collapse
microseismic activity was concentrated in two main of mine workings at the base of the mountain (Read
zones (Fig. 9): that near the scarp of the 1991 rock slide et al. 2005). Similar attempts will be made at Åknes
events and that coinciding with the highest density of to classify the microseismic events recorded by their
faults (Spillmann 2007). These two zones are bound system, based on rock falls, small-scale slides, and
by the geodetically determined limits of the moving those directly related to the main slide body (Roth et al.
mass (e.g. Fig. 5c). A full interpretation of the micro- 2006). As with Randa, signal quality has been noted as
seismic activity recorded during the Randa experiment being a key limiting factor with respect to constraining
is reported in Spillmann et al. (2007). source locations and mechanisms.
The system at Turtle Mountain also combines It is expected that the application of emerging mon-
surface- and borehole-mounted geophones, and builds itoring technologies like those used in the Randa,
on earlier experiences with seismic monitoring at Tur- Åknes and Turtle Mountain project studies, will pro-
tle Mountain carried out between 1983 and 1992. vide an improved means to gain a better understanding
These identified several different sources including of rock slope deformation kinematics and failure.

47
Figure 11. Distinct-element modeling of complex rock slope displacements at Randa and comparison between measured
and modeled cumulative displacement profiles, assuming slip along discontinuities and elasto-plastic block deformation. Note
that model boundaries extend beyond those shown (as indicated by the dashed boundary line).

At Turtle Mountain and Åknes, these are being inte- displacement patterns induced during thermal cycling
grated into early warning monitoring and response and those measured in situ by the instrumentation (Fig.
plans that comprise detailed monitoring procedures, 10). Results for the latter revealed that the cyclic nature
threshold and alert level development, notification of the displacements seen in the monitoring data was
protocols and emergency response (e.g. Froese et al. controlled by the thermal response of the rock mass
2005). Froese et al. (2005) conclude that the use of to seasonal temperature changes in the upper 10 m
multiple systems that provide different spatial and tem- of ground. This is similar to preliminary findings at
poral coverage, together with sufficient redundancy, Turtle Mountain, which suggest that thermal cycling
provide a higher level of confidence in interpreting the (i.e. thermal induced stress changes) contribute more
kinematics of movement and impending failure than towards measured slope deformations than do heavy
would be available from single sensor readings. precipitation events (Moreno & Froese 2007).
Thus, not only does monitoring data provide an
important means to constrain numerical models, but
3.3 Use of numerical models to interpret slope numerical modeling provides a means to better under-
monitoring responses stand monitoring data (Eberhardt & Willenberg 2005).
Numerical analysis can be a useful tool to provide At both Randa and Åknes, simplified numerical mod-
confirmation of the geological model and/or conclu- eling has been carried out together with the mapping
sions drawn from investigative monitoring, as well and monitoring programs to help identify and con-
as to explore possible future behavior. An exam- strain possible sliding surface/instability scenarios that
ple of this is the numerical analysis completed for would produce displacement patterns similar to those
the Checkerboard Creek rock slope to explore and measured in situ. Figure 11 provides an example from
confirm the indications from investigative monitor- a series of distinct element models generated for the
ing that a thermal control existed for the deep-seated Randa study, incorporating the key active geological
rock slope movements being measured (Watson et al. structures identified through mapping and geotechni-
2006). Modeling was initially carried out using the cal monitoring. The blocks are modeled using a Mohr-
Itasca finite-difference code FLAC, but results sug- Coulomb elasto-plastic constitutive model where the
gested that a continuum approach could not capture properties are scaled to those for an equivalent con-
the pattern of displacements and pore pressures mea- tinuum to account for smaller scale discontinuities
sured in situ. Itasca’s distinct element code UDEC not explicitly included in the distinct-element model.
was next used because of its capability to explicitly Results show a correspondence between the measured
include joints and shear zones, with groundwater flow and modeled block movements with toppling and
restricted to those joints, together with its ability to translational movements in the upper part of the slide
model thermal and dynamic loads. Models investi- and outward rotation of the blocks at depth. The latter
gated the stability of the slope during a 1 in 10, 000 year suggests that deep-seated yield together with shearing
earthquake, as well as correlations between modeled along persistent discontinuities may be an important

48
contributing factor in the complex block deformation field mapping and instrumentation data to constrain
patterns measured. It should be noted that uncertainty numerical models but also in the use of numerical
does exist for the inclinometer readings over the lower modeling to provide a means to better interpret and
portions of the borehole, and the modeling results understand complex monitoring and early warning
incorporate the limitations inherent in a 2-D repre- data. Thus, by better integrating the different data
sentation of a strongly 3-D problem. Still, any insights sets collected through all phases of an investigation,
gained into the instability mechanism, either support- from mapping to monitoring to analysis, geological
ing or refuting a current interpretation, provides a uncertainty can be minimized and controlled with
means to better plan and design future in situ inves- respect to the comprehension of complex rock slope
tigation, instrumentation and monitoring schemes for failure mechanisms, thereby improving our ability
the site. to effectively assess, monitor, mitigate and predict
the potential for catastrophic rock slope failure and
provide early warning to those endangered.
4 CONCLUSIONS

Issues related to geological complexity and uncer- ACKNOWLEDGEMENTS

tainty represent a significant obstacle to better predict-
ing the spatial and temporal evolution of catastrophic The authors wish to acknowledge and thank the dif-
rock slope failures. Advances in satellite and terres- ferent researchers, whose results were drawn on for
trial radar technologies (e.g. Slope Stability Radar) the material presented in this paper, including those at
and automation through wireless data acquisition are BC Hydro, GroundProbe and those connected to the
helping to address these issues. However, the findings Randa, Åknes and Turtle Mountain rockslide research
summarized in this paper emphasize the additional studies. Special thanks are extended to the Randa
need to not only improve landslide monitoring tech- In Situ Rockslide Laboratory team, including Dr.
nologies but to also better integrate the various data Heike Willenberg, Dr. Keith Evans, Dr. Hansruedi
sets collected during field-based investigations, mon- Mauer, Dr. Tom Spillmann, Dr. Björn Heincke, Prof.
itoring and stability analyses in order to overcome Alan Green and Prof. Doug Stead.
these challenges. Through the integration of these
data, a more reliable model of the controlling landslide
mechanism can be incorporated into the engineering
decision-making process. REFERENCES
Several examples are provided from recent studies
that aim to improve our mechanistic understanding of Berardino, P., Fornaro, G., Lanari, R. & Sansosti, E. 2002.
deep-seated rock slope behavior. These include experi- A new algorithm for surface deformation monitoring
based on Small Baseline Differential SAR Interferograms.
mental efforts involving the construction of high alpine Transactions of Geoscience and Remote Sensing 40 (11):
research facilities or ‘‘in situ rockslide laboratories’’, 2375–2383.
where the integration of state-of-the-art site charac- Bhandari, R.K. 1988. Some practical lessons in the inves-
terization and instrumentation systems and numerical tigation and field monitoring of landslides. In Bonnard
modeling are being used to better understand the mech- (ed.), Proc., 5th Int. Symp. on Landslides, Lausanne.
anisms controlling pre-failure deformations over time. Rotterdam: Balkema, v3, 1435–1457.
At depth, passive monitoring of microseismic activity Blikra, L.H., Longva, O., Harbitz, C. & Løvholt, F. 2005.
offers a means to detect subsurface tensile fracturing Quantification of rock-avalanche and tsunami hazard in
and/or shear slip along internal fracture planes that Storfjorden, western Norway. In Senneset et al. (eds.),
Landslides and Avalanches, Proceedings of the 11th Inter-
may provide insights into the evolution of a progres- national Conference and Field Trip on Landslides (ICFL),
sively developing rock slope failure. Data from both Norway. London: Taylor & Francis Group, pp. 57–64.
surface geodetic and subsurface instrumentation sys- Bonzanigo, L., Eberhardt, E. & Loew, S. 2007. Long-
tems can be integrated to obtain a description of the term investigation of a deep-seated creeping landslide in
3-D displacement field. When further integrated with crystalline rock—Part 1: Geological and hydromechani-
subsurface geological information, derived through cal factors controlling the Campo Vallemaggia landslide.
geophysical and/or borehole investigations, complex Canadian Geotechnical Journal 44 (10): 1157–1180.
block movements can be resolved relative to major Brown, N., Kaloustian, S. & Roeckle, M. 2007. Monitoring
persistent fractures and shears. of open pit mines using combined GNSS satellite receivesr
and robotic total stations. In Potvin (ed.), Proc. 2007 Int.
The results presented also emphasize the impor- Symp. on Rock Slope Stability in Open Pit Mining and
tance of numerical modeling to provide support Civil Engineering, Perth. Perth: ACG, pp. 417–429.
for and/or refute interpretations drawn from inves- Crosta, G.B. & Agliardi, F. 2003. Failure forecast for large
tigative monitoring as well as to explore possible rock slides by surface displacement measurements. Can.
future behavior. Focus is placed not only on using Geotech. J. 40 (1): 176–191.

49
Day, A.P. & Seery, J.M. 2007. Monitoring of a large wall fail- Hutchinson, D.J., Diederichs, M.S., Carranza-Torres, C.,
ure at Tom Price Iron Ore Mine. In Potvin (ed.), Proc. 2007 Harrap, R., Rozic, S. & Graniero, P. 2007. Four dimen-
Int. Symp. on Rock Slope Stability in Open Pit Mining and sional considerations in forensic and predictive simula-
Civil Engineering, Perth. Perth: ACG, pp. 333–340. tion of hazardous slope movement. In Eberhardt et al.
Eberhardt, E., Bonzanigo, L. & Loew, S. 2007. Long- (eds.), Proc. 1st Canada-U.S. Rock Mechanics Sympo-
term investigation of a deep-seated creeping landslide in sium, Vancouver. London: Taylor & Francis, v1, pp.
crystalline rock—Part 2: Mitigation measures and numer- 11–19.
ical modelling of deep drainage at Campo Vallemaggia. Inaudi, D. & Glisic, B. 2007. Overview of fiber optic sens-
Canadian Geotechnical Journal 44 (10): 1181–1199. ing technologies for geotechnical instrumentation and
Eberhardt, E., Spillmann, T., Maurer, H., Willenberg, H., monitoring. Geotechnical News 25 (3): 27–31.
Loew, S., and Stead, D. 2004a. The Randa Rockslide Kaiser, P.K., Henning, J.G., Cotesta, L. & Dasys, A. 2002.
Laboratory: Establishing brittle and ductile instability Innovations in mine planning and design utilizing collab-
mechanisms using numerical modelling and microseis- orative virtual reality (CIRV). 104th CIM Annual General
micity. In Lacerda et al. (eds.), Proc. 9th Int. Symp. on Meeting, Vancouver.
Landslides, Rio de Janeiro. Leiden: A.A. Balkema, pp. Little, M.J. 2006. Slope monitoring strategy at PPRust
481–487. open pit operation. In Stability of Rock Slopes in
Eberhardt, E., Stead, D. & Coggan, J.S. 2004b. Numerical Open Pit Mining and Civil Engineering Situations.
analysis of initiation and progressive failure in natural Johannesburg: SAIMM, Symposium Series S44, pp.
rock slopes—the 1991 Randa rockslide. Int. J. Rock Mech. 211–230.
Min. Sci. 41 (1): 69–87. Loew, S., Willenberg, H., Spillmann, T., Heincke, B.,
Eberhardt, E. & Willenberg, H. 2005. Using rock slope defor- Maurer, H., Eberhardt, E. & Evans, K. 2007 Structure
mation measurements to constrain numerical analyses, and kinematics of a large complex rockslide as determined
and numerical analyses to constrain rock slope defor- from integrated geological and geophysical investigations
mation measurements. In Barla & Barla (eds.), Proc. (Randa, Switzerland). In Schaefer et al. (eds.), Proc., 1st
11th Int. Conf. on Computer Methods and Advances in North American Landslide Conference, Vail. CD-ROM.
Geomechanics, Torino. Bologna: Pàtron Editore, v4, pp. Meisina, C., Zucca, F., Fossati, D., Ceriani, M. & Allievi, J.
683–692. 2005. PS InSAR integrated with geotechnical GIS: Some
Ferretti, A., Prati, C. & Rocca, F. 2001. Permanent Scatterers examples from southern Lombardia. In Sansò & Gil (eds.),
in SAR interferometry. Transactions of Geoscience and Geodetic Deformation Monitoring: From Geophysical
Remote Sensing 39 (1): 8–20. to Engineering Roles. Berlin: Springer, IAG Symposia,
Froese, C.R., Keegn, T.R., Cavers, D.S. & van der Kooij, M. v131, pp. 65–72.
2005. Detection and monitoring of complex landslides Mikkelsen, P.E. 2003. Advances in inclinometer data analy-
along the Ashcroft Rail corridor using spaceborne InSAR. sis. In Myrvoll (ed.), Proc. 6th Int. Symp. on Field Mea-
In Hungr et al. (eds.), Landslide Risk Management, Proc. surements in Geomechanics, Oslo. Lisse: A.A. Balkema,
of the Int. Conf., Vancouver. Leiden: Balkema, pp. pp. 555–567.
565–570. Moore, D.P. Imrie, A.S. & Baker D.G. 1991. Rockslide risk
Froese, C.R. & Moreno, F. 2007. Turtle Mountain Field reduction using monitoring. In Proc., Canadian Dam
Laboratory (TMFL): Part 1—Overview and activities. In Association Meeting, Whistler, BC. Canadian Dam Safety
Schaefer et al. (eds.), Proc., 1st North American Landslide Association.
Conference, Vail. CD-ROM. Moore, D.P. & Imrie, A.S. 1992. Stabilization of Dutchman’s
Fukuzono, T. 1985. A new method for predicting the fail- Ridge. In Bel (ed.), Proc. 6th Int. Symp. on Landslides,
ure time of a slope. In Proc. IVth Int. Conf. and Field Christchurch. Rotterdam: A.A. Balkema, pp. 1783–1788.
Workshop on Landslides, Tokyo. Tokyo: NRCDP, pp. Moreno, F. & Froese, C.R. 2007. Turtle Mountain Field Labo-
145–150. ratory (TMFL): Part 2—Review of trends 2005 to 2006. In
Glastonbury, J. & Fell, R. 2002. Report on the analysis of Schaefer et al. (eds.), Proc., 1st North American Landslide
slow, very slow and extremely slow natural landslides. Conference, Vail. CD-ROM.
The University of New South Wales, UNICIV Report No. Read, R.S., Langenberg, W., Cruden, D., Field, M.,
R-402. Stewart, R., Bland, H., Chen, Z., Froese, C.R., Cavers,
Harries, N., Noon, D. & Rowley, K. 2006. Case studies of D.S., Bidwell, A.K., Murray, C., Anderson, W.S.,
slope stability radar used in open cut mines. In Stability of Jones, A., Chen, J., McIntyre, D., Kenway, D., Bingham,
Rock Slopes in Open Pit Mining and Civil Engineering D.K., Weir-Jones, I., Seraphim, J., Freeman, J., Spratt, D.,
Situations. Johannesburg: SAIMM, Symposium Series Lamb, M., Herd, E., Martin, D., McLellan, P. & Pana, D.
S44, pp. 335–342. 2005. Frank Slide a century later: The Turtle Mountain
Harries, N.J. & Roberts, H. 2007. The use of Slope Stabil- monitoring project. In Hungr et al. (eds.), Landslide Risk
ity Radar (SSR) in managing slope instability hazards. Management: Proceedings of the International Confer-
In Eberhardt et al. (eds.), Proc. 1st Canada-U.S. Rock ence on Landslide Risk Management, Vancouver. Leiden:
Mechanics Symposium, Vancouver. London: Taylor & A.A. Balkema, pp. 713–723.
Francis, v1, pp. 53–59. Rose, N.D. & Hungr, O. 2007. Forecasting potential rock
Heincke, B. 2005. Determination of 3-D fracture distribu- slope failure in open pit mines using the inverse-velocity
tion on an unstable mountain slope using georadar and method. Int. J. Rock Mech. Min. Sci. 44 (2): 308–320.
tomographic seismic refraction techniques. D.Sc. thesis, Rosser, N.J., Petley, D.N., Lim, M., Dunning, S.A. &
Applied and Environmental Geophysics, Swiss Federal Allison, R.J. 2005. Terrestrial laser scanning for monitor-
Institute of Technology (ETH Zurich). 157 pp. ing the process of hard rock coastal cliff erosion. Quarterly

50
 Journal of Engineering Geology & Hydrogeology 38: by using ground-based synthetic aperture radar inter-
363–375. ferometry: Application to the Ruinon landslide in
Roth, M., Dietrich, M., Blikra, L.H. & Lecomte, I. 2006. the Italian Alps. J. Geophys. Res. 108(B8): 2387,
Seismic monitoring of the unstable rock slope site at doi:10.1029/2002JB002204.
Åknes, Norway. In Gamey (ed.), Proc. 19th Annual Symp. Watson, A.D., Martin, C.D., Moore, D.P., Stewart, T.W.G. &
on the Application of Geophysics to Engineering and Lorig, L.L. 2006. Integration of geology, monitoring and
Environmental Problems (SAGEEP), Seattle. CD-ROM. modeling to assess rockslide risk. Felsbau 24 (3): 50–58.
Rott, H., Scheuchel, B. & Siegel, A. 1999. Monitoring very Watson, A.D., Moore, D.P., Stewart, T.W. & Psutka, J.F.
slow slope movements by means of SAR interferome- 2007. Investigations and monitoring of rock slopes at
try: A case study from mass waste above a reservoir in Checkerboard Creek and Little Chief Slide. In Eberhardt
the Otztal Alps, Austria. Geophys. Res. Lett. 26 (11): et al. (eds.), Proc. 1st Canada-U.S. Rock Mechanics Sym-
1629–1632. posium, Vancouver. London: Taylor & Francis, v2, pp.
Sakurai, S. 1991. Field measurements versus analysis in 1015–1022.
geotechnical engineering problems. In Sorum (ed.), Field Willenberg, H. 2004. Geologic and kinematic model of a
Measurements in Geomechanics, Proc. of the 3rd Int. complex landslide in crystalline rock (Randa, Switzer-
Symp., Oslo. Rotterdam: Balkema, v3, pp. 405–414. land). D.Sc. thesis, Engineering Geology, Swiss Federal
Singhroy, V., Couture, R. & Molch, K. 2005. InSAR moni- Institute of Technology (ETH Zurich). 187 pp.
toring of the Frank Slide. In Hungr et al. (eds.), Landslide Willenberg, H., Evans, K.F., Eberhardt, E. & Loew, S. 2003.
Risk Management, Proc. of the Int. Conf., Vancouver. Monitoring of complex rock slope instabilities—correc-
Leiden: Balkema, pp. 611–614. tion and analysis of inclinometer/extensometer surveys
Spillmann, T. 2007. Borehole radar experiments and micro- and integration with surface displacement data. In Myr-
seismic monitoring on the unstable Randa rockslide voll (ed.), Proc. 6th Int. Symp. on Field Measurements in
(Switzerland). D.Sc. thesis, Applied and Environmental Geomechanics, Oslo. Lisse: A.A. Balkema, pp. 393–400.
Geophysics, Swiss Federal Institute of Technology (ETH Willenberg, H., Evans, K.F., Eberhardt, E., Loew, S.,
Zurich). 205 pp. Spillmann, T. & Maurer, H.R. 2004. Geological, geo-
Spillmann, T., Maurer, H., Green, A.G., Heincke, physical and geotechnical investigations into the internal
B., Willenberg, H. & Husen, S. 2007. Microseis- structure and kinematics of an unstable, complex sliding
mic investigation of an unstable mountain slope in mass in crystalline rock. In Lacerda et al. (eds.), Proc. 9th
the Swiss Alps. J. Geophys. Res. 112(B07301): Int. Symp. on Landslides, Rio de Janeiro. Leiden: A.A.
doi:10.1029/2006JB004723. Balkema, pp. 489–494.
Tarchi, D., Casagli, N., Moretti, S., Leva, D. &
Sieber, A.J. 2003. Monitoring landslide displacements

51
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Effects of earthquakes on slopes

Ikuo Towhata
University of Tokyo, Tokyo, Japan

Tetsuo Shimomura
Ohbayashi Corporation, Tokyo, Japan

Masanori Mizuhashi
Public Works Research Institute, Tsukuba, Japan

ABSTRACT: This paper First addresses examples of large and small slope failures that were triggered by
earthquakes. Their significance comes from the negative effects to the human community, which consist not
only of the number of casualties but also of the difficulties in post-earthquake rescue and restoration. To mitigate
the problems, identification of seismic instability and assessment of debris fun-out distance are important.
Because the existing practical methods for these have several problems, the direction of their improvement is
presented with examples and case studies.

1 INTRODUCTION the failed slope, and 300 m in depth. The total volume
of the failed mass is evaluated to be 24 billion m3 .
Slope failure is one of the most significant problems This event occurred in a prehistoric time, probably
during strong earthquakes. It is important in that it 10,000 years ago, according to C14 dating (Watson &
might claim thousands of lives and destroy both pub- Wright 1969). Oberlander (1965) attributed this event
lic and private properties. When local transportation to the foot erosion by river, but it is possible that the
is blocked by slope failures, moreover, rescue and unstable slope was finally destroyed by seismic shak-
restoration become significantly difficult. It is true, on ing. The important feature of this landslide lies in the
the other hand, that slope failure is a part of geomor- enormous run-out length of debris, which traveled as
phological processes and may be beyond the human long as 20 km, overtopping an anticline hill (Fig. 2).
control. Therefore, all what human beings can do is Gigantic slope failures in more recent times often
not to worsen the risk of slope failure, but to mitigate claimed many victims. Figure 3 demonstrates the ruin
negative effects to the human community. The present of Yungay City of Peru that was destroyed by seismic
paper addresses what happened in slopes during past failure of Huascaran Mountain slope upon an earth-
and recent earthquakes, their effects to the public, and quake of magnitude = 7.9 in 1970. The number of
various kinds of efforts to mitigate the negative effects. victims exceeded 17, 000. When a similar slope fail-
Note that attention is focused mostly on natural slopes, ure occurred in 1960, the debris flowed along a nearby
but some events in manmade slopes will be introduced Rio Santa channel, and the city was protected from it
in relation to recent urban developments. by a hill behind the city (Fig. 3). This experience gave
people a wrong idea that debris would never hit the
city in future. After the 1970 tragedy, the entire city
2 EXAMPLES OF ON-SHORE LARGE SLOPE was relocated to a safer place.
FAILURES DURING EARTHQUAKES One of the consequences of a gigantic landslide is
creation of a natural dam. The slope of Tsao-ling in
This chapter addresses different kinds of seismically- Taiwan failed several times in a large scale in the past:
induced slope failures that occurred both on shore and in 1862 and 1941 due to earthquakes and in 1942 due
in the sea, in either big or small scale, and their con- to rainfall (Kawata 1943). Figure 4 shows a lake after
sequences to the human communities. The discussion the 1999 ChiChi earthquake.
is initiated with the largest on-shore landslide in the Although a lake thus created may become a good
world. Seimareh landslide is located in south-west Iran tourist spot, problem is the possibility of the collapse
and measures 16 km in width (Fig. 1), 5 km in length of of the dam and flooding in the downstream region.

53
Figure 1. View of Seimareh landslide from Pol-e-Doghtar.

Figure 2. Debris deposit in Seimareh. Figure 4. Natural dam in Tsao-ling after 1999 ChiChi
earthquake.

Figure 3. Ruin of Yungay City in Peru. Figure 5. Statistics on cause of failure of natural dams (after
Tabata et al., 2002).

Tabata et al. (2002) reported that the dam collapse due has increased. To mitigate this situation, continuous
to overtopping is substantially more frequent than pip- efforts have been made to improve dikes and stabilize
ing failure (Fig. 5). Another negative consequence is slopes.
the long-term slope instability after the seismic failure.
Fig. 6 shows a part of Ohya slide in the upstream area of
Abe River, Japan. This slope collapsed upon the 1707
Hoei earthquake (M = 8.4). Since then, the destabi- 3 PROBLEMS OF ON-SHORE MINOR
lized slope has been producing debris flow at heavy SLOPE FAILURES
rainfalls frequently (Imaizumi et al. 2005). Conse-
quently, the river floor has raised significantly (Fig. 7) When the surface of a slope is subjected to weathering
and the chance of flooding and overtopping river dikes and disintegration, strong earthquake shaking triggers

54
Figure 6. Ohya slide in Shizuoka Prefecture.

Figure 9. Seismic failure of road embankment during 2004

Niigata-Chuetsu earthquake, Japan.

Instability problems occur in artificial fills as well.

Figure 9 illustrates an example in which a road
embankment, which was constructed on a slope, failed
upon an earthquake. Many local roads are subjected
to a similar problem. The cause of the problem is that
insufficient construction budget leads to the fill con-
struction without removing the soft surface deposits.
Those debris and organic materials at the bottom of
the road embankment cannot resist against the seis-
Figure 7. Valley of Abe River filled with debris. mic force, and the entire embankment collapses. What
is important is the choice between earthquake resis-
tance and money saving. In case the seismic failure
would not be fatal, it may be reasonable to allow such
a failure from the viewpoint of performance-based
design and the policy of minimizing the life cycle
cost. A similar but more serious problem lies in res-
idential development. Around a sprawling mega city,
inexpensive residential development is needed for low-
income citizens, and such a development is conducted
in hilly areas. Since cost has to be saved, soft valley
deposits may not be removed, and, upon earthquakes,
private lands and houses collapse. Such a loss of pri-
vate properties is never allowed by limited budget of
residents.

4 PROBLEMS OF SUBAQUEOUS
SLOPE FAILURES
Figure 8. Surface failure of weathered slope (Kashmir of
Pakistan in 2005). It is possible that earthquakes trigger failure of sub-
marine slopes. In addition to the famous examples of
Grand Banks in 1929 (Heezen & Ewing 1952) and
off Orléansville in 1954 (Heezen & Ewing 1955), the
falling of the surface soils. Figure 8 illustrates a slope 1964 Alaska earthquake of magnitude = 9.2 destabi-
of weathered limestone near Muzaffarabad, Pakistan, lized the sea bed deposit in front of Valdez (Fig. 10).
in 2005. The surface slide can easily destroy roads and A huge volume of soil mass (75 million m3 according
prevent local transportation. Accordingly, emergency to Coulter & Migliaccio (1966)) collapsed upon shak-
rescue and restoration become very difficult. ing, and this mass movement caused tsunami in the

55
Figure 10. Aerial view of Valdez (photograph taken by
Dr. K. Horikoshi).

Figure 11. Reinforcement of cliff behind house (Kamakura,

Valdez Bay. This tsunami destroyed the Valdez munic- Japan).
ipality and the town was later relocated in a safer place.
The problem was that the submarine deposit of the
Valdez Bay was most probably composed of fine non- 2. Relocation of human settlement from potentially
plastic silt that was produced by the glacial erosion and hazardous areas
was transported into the sea by the river in Figure 10. 3. Emergency evacuation
Being fine, silt grains deposit in water very slowly and The first choice is costly as stated above. The third
the resultant density becomes very low. Since there is choice may not be rapid enough during an earthquake.
no cohesion, moreover, its undrained shear strength is Both (1) and (2) require identification of potentially
very low. The possible excess pore water pressure dur- unstable slopes with practically reasonable cost. This
ing undrained shear does not dissipate during shaking, assessment of failure risk may be made by determin-
because the grain size is small. These situations made ing shear strength of soil experimentally or by field
the bay deposit highly vulnerable to undrained failure investigation, and then running a stability analysis with
and liquefaction. Moreover, it should be recalled that reasonable seismic load taken into account. This pro-
the highest tsunami so far recorded was triggered by cedure is unfortunately costly and can be conducted
rock slide impact induced by an earthquake (in Lituya only in special situations. Hence, less expensive and
Bay of Alaska: Miller 1960). time-efficient methodology is required, although the
Stability of submarine slopes is reduced due to accuracy may be sacrificed to some extent.
rapid rate of sedimentation (or human reclamation). The identification of potentially hazardous area for
If strong earthquake loading is superimposed on such (2) further requires assessing the run-out distance of
a situation, slope failure is likely to occur. Note that the failed mass. Fig. 12 compares two slope failures
a submarine landslide is one of the causative mecha- that were caused by the 2004 Niigata-Chuetsu earth-
nisms of tsunami as was the case in North Sea (Long quake. The one in front was of 110,000 m3 in soil
et al. 1989). volume and traveled over some distance at the bottom
of the valley. Consequently, this soil mass stopped the
river flow and formed a natural dam. In contrast, the
5 MITIGATION OF DAMAGE CAUSED surface failure in the same figure had approximately
BY EARTHQUAKE-INDUCED SLOPE 50 cm in thickness, suggesting that only the surface
weathered soil fell down. This smaller soil mass did
The previous sections introduced examples in which not travel laterally in the bottom of the valley and did
human communities were seriously damaged by seis- not therein affect either the river flow or road traffic.
mic slope failures. Although it is advisable to stabilize Thus, the flow characteristics of soil mass depends at
all the potentially hazardous slopes (Fig. 11), finan- least partially on its volume as pointed out earlier by
cial restrictions do not allow it. Accordingly, there are Hsü (1975). Furthermore, the effect of water content
three alternative options to be taken: is illustrated in Fig. 13 in which a dry cliff in Muzaf-
farabad of Pakistan fell down upon the 2005 Kashmir
1. Reinforcement of slope by means of retaining earthquake. Noteworthy is that the soil did not translate
walls, ground anchors, and other reinforcements laterally and houses in the bottom of the valley were

56
Figure 12. Failure of slope at Naranoki in Yamakoshi
Village upon 2004 Niigata-Chuetsu earthquake (Mizuhashi
et al., 2006).

Figure 14. Maximum possible epicentral distance of failed

slopes during past earthquakes (Yasuda, 1993).

would occur in a 500 m × 500 m grid during a future

earthquake (Table 2).
The weight points in Table 1 were determined by
regression analyses of past experiences. Therefore,
some consistency is expected between evaluation and
Figure 13. Collapse of dry slope in Muzzafarabad, Pakistan. reality. There are problems, however, in Table 1. First,
Table 1 is not intended to evaluate the safety of a par-
ticular slope. Second, the weight point for hardness
not affected. This nature is different from the behavior of rock (W4 ) increases as the rock becomes harder,
of bigger soil mass in Fig. 12 that was of high water implying more risk. This strange nature comes from
content. the regression data in which failures in hard rock
slopes were more numerous than those in soil, prob-
ably because slopes of weak soil did not exist or had
6 ASSESSMENT OF SEISMIC RISK failed during heavy rainfalls or in other circumstances
DUE TO SLOPE FAILURE prior to earthquakes. Consequently, Table 1 does not
have a reasonable consideration of soil or rock material
Any seismic risk depends primarily on the regional properties.
earthquake activities. Risk in seismic countries is To solve those two problems mentioned above, an
higher than that in nonseismic countries. This fea- attempt was made to improve the method by taking
ture is expressed in an empirical diagram in Figure 14 into account the characteristics of soil. The weight
where the maximum distance of slope failure from an in Table 1 was counted for individual slopes that
earthquake epicenter is plotted against the earthquake failed during the 2004 Niigata-Chuetsu earthquake
magnitude. In case a concerned slope lies within the and their calculated risk was compared with reality
critical epicentral distance, the failure risk has to be (Mizuhashi et al. 2006). The results in Figure 16 are
evaluated. understandable to a certain extent. The good point is
Financial reasons, however, eliminate in most cases that the assessed rank of risk changes with the increas-
the use of in-situ tests or laboratory shear tests on ing volume of landslide mass. Since larger events are
shear strength of undisturbed samples collected from associated with greater risk assessment, the results are
a concerned slope. Accordingly, risk is judged on the good. However, one of the biggest events at Naranoki
basis of easily-available information. Table 1 shows is of lower assessed risk, and there is a need for further
an example in which various site factors are evaluated improvement.
and weight points from them are added to make the The study puts emphasis on the effect of water.
final evaluation on the number of slope failures that In particular, the deterioration of properties due to

57
 Table 1. Weighting for factors related to slope instability
(See Fig. 15 for W : Kanagawa Prefectural Government, 1986).

Table 2. Assessed number of slope instabilities per 500 m ×

500 m grid (Kanagawa Prefectural Government 1986).

2004. The significant difference in strength was prob-

ably caused by deterioration or hydration of soil
minerals upon submergence. Furthermore, Figure 18
compares both peak and residual strengths of different
soils to confirm the submergence-induced deteriora-
tion. With these in mind, it was attempted to add a
water-concerned index to risk assessment, which is
obtained without running expensive shear tests.
Figure 15. Vertical cross section of slopes. The present study employed swelling characteris-
tics and plasticity index. For swelling tests, in-situ soil
grains were ground to the size less than 75 microns
so that water effects would occur rapidly. Then,
specimens for standard oedometer tests were recon-
stituted in the laboratory. Dry soil was compacted
in two layers, with 20 impacts per layer, so that
the dry mass density would be 1.061 g/cm3 (60 g
for a cylindrical sample). After compression under
20 kPa for 1,000 minutes in a dry state, specimens
were submerged in water and the volume change was
recorded (Fig. 19 for example). After some contraction
occurred during the first 30 minutes, volume expan-
sion (swelling) was observed in water-sensitive soils.
Figure 16. Seismic hazard assessment of slopes by using
Kanagawa Prefectural method. Note that the extent of swelling was remarkably vari-
able; some soils exhibited no volume expansion upon
submergence.
The other index is plasticity index (liquid limit–
water submergence is focused on. Figure 17 illus- plastic limit in %). Although this index is not directly
trates drained shear behaviors of air-dry and water- related with mechanical properties of soils, it stands
submerged specimens of soil collected from the Urase for the extent of mineral-water interaction and also is
site where rainfall-induced landslide occurred in July, easy to measure in the laboratory. Figure 20 suggests

58
Figure 17. Drained triaxial compression tests on air-dry and
Figure 19. Water submergence of soil powder collected
submerged specimens.
from Haguro Tunnel site in Niigata-Chuetsu.

Figure 20. Correlation between swelling strain and plastic-

ity index.

Figure 18. Reduction of shear strength of soils collected

from rainfall-induced landslide sites.

that there is a positive correlation between swelling

strain and plasticity index. Figure 21 demonstrates that
the swelling strain is greater for soils collected from
sites of greater slope failure. However, the correlation
is not clear in this figure because the volume of the
soil mass is also affected by the topography and other
geological conditions. Thus, swelling strain may be Figure 21. Correlation between landslide mass volume and
used as one of the indices to better assess the risk of swelling strain.
earthquake-induced slope failure. Similarly, Fig. 22
illustrates that there is a positive correlation between
the landslide volume and plasticity index, if such
exceptional cases as sandy slopes and a reactivated (swelling strain, %)/10 or W + (plasticity index)/100.
landslide are eliminated from this figure. The results are shown in Figures 23, 24. The consis-
On the basis of Figures 21, 22, the point, W , in tency with the damage extent and the determined rank
Table 2 was adjusted by employing either W + are better than that in Figure 16.

59
 Figure 25. Apparent friction angle of landslides obtained
from geometry.

Figure 22. Correlation between landslide mass volume and

plasticity index.

Figure 26. Tsao-ling slide after 1999 ChiChi earthquake.

Figure 23. Correction of Kanagawa Prefectural method by This diagram was drawn by using data collected by the
swelling strain.
present study. As is well known, the apparent friction
angle, φ, or H /L = tan(φ), decreases as the volume
of landslide mass increases.
A second approach to assess the travel distance
is the use of Newmark sliding block analogy (New-
mark 1965). It is more advanced than the foregoing
method because it employs the shear strength of soil
and the time history of (design) earthquake acceler-
ation. However, there is not yet a clear agreement on
what kind of shear strength of soil should be used out of
cyclic strength, undrained monotonic strength or else.
In the present study, the Newmark method was used
to reproduce the gigantic failure of Tsao-ling slope in
Taiwan which was triggered by the 1999 ChiChi earth-
Figure 24. Correction of Kanagawa Prefectural method by quake. The appearance of a part of the slide is shown in
plasticity index (Ip ). Figure 26, and the geological cross section is presented
in Figure 27. It may be found that the slope is made
of interbedding sandstone and mudstone (shale). To
7 ASSESSMENT OF RUN-OUT DISTANCE determine the shear strength of those materials, block
OF DEBRIS samples were collected from the remaining part of the
failed slope after the quake. Direct shear tests revealed
The extent of risk due to earthquake-induced landslide that the mudstone was of the least shear strength and
depends not only on the instability of a slope but also was likely to be the cause of the failure (Towhata et al.
on the travel distance of a failed debris mass. In this 2002).
regard, attention has to be paid to the dynamic char- Figure 28 indicates the stress-displacement rela-
acteristics of debris. Figure 25 illustrates the classical tionship of mudstone specimens. It is therein found
concept of the apparent friction angle by Hsü (1975). first that both peak and residual strengths increase

60
 Figure 29. Acceleration time history for Tsao-ling analysis.

Figure 30. Calculated sliding displacement with c = 417

kPa.

Figure 27. Geological cross section of Tsao-ling slope prior

to 1999 event (Ishihara, 1985). by substituting Amax into an empirical formula by Noda
(1975),


Kh = 3
Amax /g/3 (1)

in which g stands for the gravity acceleration. By

assuming the slope angle of 14 degrees and the
thickness of the sliding rock slab to be 80 m, the
pseudo-static factor of safety was obtained to be 0.73.
In order to conduct the Newmark-type analyses,
the CHY028 acceleration records were converted to
a component parallel to the slope (Fig. 29). Then the
time history of sliding displacement was calculated.
Figure 30, however, shows that the calculated displace-
Figure 28. Direct shear tests on Tsao-ling mudstone speci- ment was far less than the reality. Furthermore, the
men (Towhata et al. 2002). displacement terminated at the end of strong shaking,
which is contradictory to the complete failure of the
slope in reality.
with the confining pressure. A second finding is that Sassa et al. (2004) used a ring shear device to study
a cycle of unloading and reloading after the post-peak in more details the behavior of soil undergoing rapid
softening does not accomplish the peak strength again. flow.
Since the Tsao-ling slope experienced instability sev- The problem in underestimation of displacement is
eral times in the past, the present study decided to that the real failed rock mass was broken into pieces,
use the residual strength for stability analyses. The as will be shown in the next chapter (Fig. 32), and
employed strength parameters were c = 417 kPa and the assumed strength parameters were not relevant. In
φ = 19.4 degrees. particular, broken debris hardly had cohesion. There-
The first analysis was pseudo-static, in which the fore, a second analysis was carried out by using c =
maximum acceleration (Amax ) recorded at a nearby 0 kPa, while using the same friction angle. The result in
observatory (CHY028 at 8 km distance) of 748 cm/s2 Figure 31 shows that displacement increases towards
was converted to a seismic coefficient of Kh = 0.305 infinity even after the end of shaking.

61
Figure 31. Calculated sliding displacement without
cohesion.
Figure 32. Debris deposit at the bottom of Tsao-ling failed
slope.

8 NUMERICAL ANALYSIS ON LAYERED

MODEL OF SOIL MASS

The unsatisfactory performance of the Newmark-type

analysis in the previous section implies that the cal-
culation of debris over a long distance is beyond
its capacity. This is because the Newmark method
assumes a rigid block movement, while real debris
flow consists of decomposed grains (Fig. 32). As an
alternative, therefore, a more advanced analysis such
as distinct element analysis is promising. The present
paper, however, addresses a simpler choice.
To get an idea about the type of analysis, model
tests were conducted on the nature of debris flow.
Figure 33 shows the change of appearance of a model Figure 33. Distortion of debris flow with time.
of flow. At 0.8 second, there are three targets near
the front of the flow (in a white frame). Those tar-
gets are still visible at 1.0 second. However, they are
overlain by following soil at 1.2 second, and have
completely disappeared at 1.4 second. This suggests
that there is a significant distortion and movement of
grains inside the flow mass, which makes Newmark-
type analysis totally irrelevant for the present type of
Figure 34. Layered model of debris flow.
problem.
Figure 34 illustrates the layered idealization of a
debris mass in which the mass is composed of N -
layers of finite thickness. The internal friction angle
between layers is designated by φ, while the base fric-
tion by φb . When the bottom friction is greater than
the internal friction, the bottom layer stops its move-
ment after some translation, while overlying layers can
continue their movement further. This idea seems to
be consistent with the observation in Fig. 33.
The overall view of the analysis is illustrated in
Figure 36 where the initial kinetic energy is supplied
by free fall of Hi , and the velocity decreases gradu-
ally with the progress of down-slope movement. The
apparent friction angle is assessed by the overall falling Figure 35. Internal and base friction angles.
and lateral displacement (H /L).
An example analysis was made by assuming 100
layers (N = 100), the slope angle (θ ) = 10 degrees, increases with time. The lower the internal friction is,
the base friction angle (φb ) = 40 degrees, and varying the longer is the run-out distance. Second, Figure 38
international friction angles (φ). First, Figure 37 indi- shows the number of layers (n) that arrive at differ-
cates the travel distance of the front of the debris that ent travel distances. As illustrated in Figure 34, lower

62
Figure 36. Initiation and termination of movement in layer
analysis. Figure 38. Number of layers that reach different travel
distances.

Figure 37. Travel distance of debris front increasing with

Figure 39. Height of debris deposit changing with horizon-
time.
tal distance.

layers stop their movement at shorter travel distance

because of the greater bottom friction, while upper lay-
ers can travel longer. This feature makes a significant
difference in the ultimate travel distance; the smaller
internal friction allows longer travels.
Third, the thickness of debris deposit is assessed.
When many layers stop within a short distance (case
of φ = 30 degrees in Figure 38), the accumulation
of debris produces a high debris hill. This feature is
somehow assessed by dn/d (distance). As shown in
Fig. 39, the greater internal friction creates a high hill
of accumulated debris at the front, and, in contrast,
the lower internal friction allows more spread deposit Figure 40. Reproduced apparent friction angle changing
over a greater area. The latter corresponds to a flow of with debris volume.
liquefied (fluidized) debris with high water content in
which the reduced effective stress makes the internal
frictional resistance very small.
Finally an attempt was made to reproduce the empir- in Figure 40 that H /L (tangent of the apparent fric-
ical relationship in Figure 25. Because the thickness tion angle) decreases with increasing volume when
is represented by the number of layers, N , the vol- the internal friction angle is remarkably smaller than
ume is approximately represented by N 3 . It is seen the bottom friction.

63
9 CONSIDERATION ON LAYER THICKNESS
IN LAYERED FLOW SIMULATION

The numerical simulation in the previous section did

not clearly indicate how to determine the layer thick-
ness and, accordingly, the number of layers, N . At
present, it is supposed that the thickness depends
upon grain size of the debris. Figure 41 demon-
strates a debris deposit in Niigata that was triggered by
heavy rainfall. In this valley, the deposits developed a
step in the surface and its height was approximately
3 m. Figure 42 shows a deposit of slope failure near
Joganji in Nagaoka City that was caused by the 2004
Niigata-Chuetsu earthquake. The height of the step Figure 43. Earthquake-induced slope failure at Tsukidate
was approximately 50 cm. Finally, Figs. 43 and 44 during the 2003 Miyagi-ken earthquake.
reveal a fluidized slope failure in Tsukidate during the
2003 Miyagi-ken earthquake. The height of the step
near the front was 58 cm.

Figure 41. Rainfall-induced debris flow deposit Chu-ei

Tunnel in Niigata-Chuetsu in 2005.
Figure 44. Step of deposit in earthquake-induced slope
failure at Tsukidate.

Information was collected from case studies as well

as literatures on the above-mentioned layer height
and grain size distribution. The results are plotted in
Figure 45. There is no good correlation between the
layer thickness and the mean grain size, D50 . In con-
trast, D90 or probably the maximum grain size has
some correlation. Tentatively, therefore, an empirical
formula of

Layer thickness (m) = 30 × D90 (m) (2)

is proposed. Consequently, the number of layers is

determined by
Figure 42. Step of deposit in earthquake-induced slope
failure near Joganji, 2005 Niigata-Chuetsu earthquake. N = Total thickness of debris / layer thickness (3)

64
 REFERENCES

Coulter, H.W. & Migliaccio, R.R. 1966. Effects of the earth-

quake of March 27, 1964 at Valdez, Alaska, US Geological
Survey Professional Report 542-C.
Heezen, B.C. & Ewing, M. 1952. Turbidity currents and sub-
marine slumps, and the 1929 Grand Banks Earthquake,
Am. J. Science. 250: 849–873.
Heezen, B.C. & Ewing, M. 1955. Orléansville earthquake
and turbidity currents, Bull. Am. Assoc. Petroleum
Geologist 39(12): 2505–2514.
Hsü, K.J. 1975. Catastrophic debris streams (Sturzstroms)
Figure 45. Empirical relationship between thickness of
generated by rockfalls. Geol. Soc. Am. Bull. 86: 129–140.
debris layer and grain size.
Imaizumi, F., Tsuchiya, S. & Ohsaka, O. 2005. Behaviour of
debris flows located in a mountainous torrent on the Ohya
landslide, Japan. Can. Geotech. J. 42: 919–931.
10 CONCLUSIONS Ishihara, K. 1985. Stability of natural deposits during earth-
quakes, Theme Lecture. 11th ICSMFE, San Francisco,
The present paper discussed knowledge collected from Vol. 1: 321–376.
recent earthquakes. In mitigation of negative effects Kanagawa Prefectural Government. 1986. Report on Seis-
of slope failure to the human community, it is very mically induced Damage during future earthquake. pp.
important to identify potentially hazardous slopes and 13–63 (quoted by Manual for Zonation on Seismic
Geotechnical Hazards, 1998, TC4 of ICSMGE).
to assess the run-out distance. In particular, this has to
Kawata, S. 1943. Study of new lake created by the earthquake
be done with a reasonable cost. With regard to these, in 1941 in Taiwan. Bull. Earthq. Res. Inst., Univ. Tokyo
the following conclusions are drawn. 21: 317–325 (in Japanese).
1. Problems of large slope failures are well known. Long, D.E. & Dawson, A.G. 1989. A Holocene tsunami
deposit in eastern Scotland. J. Quaternary Science 4(1):
Those of minor failures, however, have to be under-
61–66.
stood more, because they may stop local trans- Miller, D.J. 1960. Giant Waves in Lituya Bay Alaska, Shorter
portations and make rescue or restoration more Contributions to General Geology, US Geological Survey
difficult. Professional Paper 354-C.
2. Seismically hazardous slopes should be identified Mizuhashi, M., Towhata, I., Sato, J. & Tsujimura, T. 2006.
with some consideration on characteristics of soils. Examination of slope hazard assessment by using case
3. For the hazard assessment, a use of either swelling studies of earthquake- and rainfall-induced landslides.
properties or plasticity index, both of which can be Soils and Foundations 46(6): 843–853.
measured with minor efforts, is proposed. Newmark, N.M. 1965. Effects of earthquakes on dams and
embankments, Geotech. 5(2): 137–160.
4. Hazard assessment further needs to assess the travel
Noda, S., Uwabe, T. & Chiba, T. 1975. Relation between
distance of debris. seismic coefficient and ground acceleration for grav-
5. Newmark rigid block analogy is not relevant for ity quaywall. Report of the Port and Harbor Research
this purpose. This study proposes a similarly sim- Institute 67–111 (in Japanese).
ple layer idealization for the assessment of run-out Oberlander, T. 1965. The Zagros streams, Syracuse Geo-
distance. graphical Series, No. 1. Syracuse University Press.
6. To facilitate the determination of layer thickness, a Sassa, K., Fukuoka, H., Wang, G.-H. & Ishikawa, N. 2004.
correlation between the thickness and particle size Undrained dynamic-loading ring-shear apparatus and its
was presented. application to landslide dynamics. Landslides 1(1): 7–19.
Tabata, S., Mizuyama, T. & Inoue, K. 2002. Natural dam and
disaster Kokin Shoin Publ. pp. 50–53 (in Japanese).
Towhata, I., Yamazaki, H., Kanatani, M., Ling, C.-E., &
ACKNOWLEDGMENT Oyama, T. 2002. Laboratory shear tests of rock specimens
collected from site of Tsao-ling earthquake-induced land-
Direct shear tests on Tsao-ling rock block samples slide. Tamkang Journal of Science and Engineering 4(3):
were conducted at the Central Research Institute of 209–219.
Electric Power Industries. Model tests on debris flow Yasuda, S. 1993. Zoning for slope instability. Manual for
were carried out by Mr. Y. Nishimura for his under- zonation of seismic geotechnical hazards, TC 4, Int. Soc.
graduate thesis. Field studies were made by T. Ito of Soil Mech. Found. Eng. 49–49.
Watson, R.A. & Wright Jr., H.E. 1969. The Saidmarreh land-
Fudo-Tetra Corporation. Deep thanks are expressed to
slide, Iran. Geol. Soc. Am. Special Paper 123: 115–139.
those supports and contributions.

65
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Monitoring and modeling of slope response to climate changes

H. Rahardjo
School of Civil and Environmental Engineering, Nanyang Technological University, Singapore

R.B. Rezaur
Department of Civil Engineering, Universiti Teknologi Petronas, Perak, Malaysia

E.C. Leong
School of Civil and Environmental Engineering, Nanyang Technological University, Singapore

E.E. Alonso, A. Lloret & A. Gens

Department of Geotechnical Engineering and Geosciences, UPC, Barcelona, Spain

ABSTRACT: Shallow slides are often triggered by climate effects. An understanding of the slope failure
conditions and effective remedial measures can be achieved by comprehensive field monitoring of climatic
and hydrologic changes and the consequent changes in slope responses. Two contributions from two different
geographic regions are presented to gain understanding of the complex phenomena involved in slope failure
studies. In the first part Alonso et al., contributes theoretical analysis of a stochastic model for the reliability
of planar slides in a partially saturated soil, subjected to a rainfall history described as a time series and then
presents a case history of shallow mudslides triggered by a Mediterranean climate, analyzed by means of a
coupled hydro-mechanical modeling tool. The joint saturated-unsaturated consideration of the slide is necessary
to understand field data. In the second part Rahardjo et al., contributes field monitored data from three residual
soil slopes in Singapore and demonstrates how field monitored data on climatic, hydrologic, and slope variables
were used to evaluate slope responses under subtropical Singapore climate.

1 INTRODUCTION antecedent rainfall, rainfall intensity and rainfall

duration. However, it is impossible to dissociate
Global warming, rising sea level and climatic changes climate effects from the type of slide. In general,
have become important issues of the world in recent large landslides require specific analysis. Shallow
decades. Climatic changes have affected rainfall pat- slides, however, react to rainfall conditions in a more
terns in many parts of the world, causing occurrences predictable manner.
of numerous landslides. Many tropical areas are prone The objective of this paper is to present strategies
to frequent rainfall-induced slope failures (Poh et al. for monitoring and modelling of slope response to cli-
1985, Pitts & Cy 1987, Tan et al. 1987, Chatterjea mate changes. In this context two contributions from
1994, Toll et al. 1999). The problem is escalated by the two different geographic regions are presented in two
increasing rate of hillside developments for engineered sections (PART-A and B). In the first part (PART-A)
and fill slopes in many regions. Alonso et al., presents a stochastic model for the anal-
Rainfall of both event-based and antecedent, runoff, ysis for the reliability of planar slides in a partially
infiltration, and their contributions to pore-water pres- saturated soil, subjected to a rainfall history described
sure changes in a residual soil slope are the variables as a time series and then presents a case history of shal-
pertinent to slope responses. The dynamic flux bound- low mudslides triggered by a Mediterranean climate
ary conditions are controlled by the physical properties analyzed by means of a coupled hydro-mechanical
of the soil, in particular the unsaturated soil in the modeling tool. The theoretical analysis of the risk
vadose zone above the water table. of slide associated with a given record of rainfall
Attempts to relate climate conditions to the presented by Alonso et al., considers infiltration con-
occurrence or reactivation of landslides are numer- ditions through a partially saturated soil and integrates
ous (Wieczorek (1996) and Corominas (2000) pro- rainfall effects through a random function formula-
vide a detailed account). Most methods combine tion of the concept of risk of failure. Closed form

67
solutions were found for the probability of failure.
The theoretical analysis highlights several aspects of
rain-induced risk of failure. The case history for shal-
low instabilities in a clay formation analyzed by means
of a coupled hydro-mechanical analysis presented by
Alonso et al., is also valid for saturated and unsaturated
slope conditions.
In the second part (PART-B) Rahardjo et al.,
presents field monitored data from three residual soil
slopes in Singapore and demonstrates how field moni-
tored data on climatic, hydrologic, and slope variables
were used to evaluate slope responses in terms of;
Figure 1. Geometry of the solved problem.
(i) pore-water pressure changes; (ii) runoff generation
and infiltration amount. Part-I and II together provides
understanding of the complex phenomena involved in it will be assumed to be constant in the following).
slope failure studies. The safety margin of an infinite slope against failure
along a plane at depth (H − h) (Fig. 1) is given by
  H 
2 PART-A. MODELLING RAINFALL EFFECTS M = γw n Sr dz + γs (1 − n)(H − h)
ON SHALLOW SLIDES (ALONSO ET AL.) h

× cos α(cos α tan φ  − sin α)

2.1 Rain infiltration and risk of sliding—a simple
stochastic model + c + (pa − pω ) tan φb (2)
Soil moisture variation in time within slopes is recog-
Safety margin is defined as the difference between
nized as a major source of uncertainty concerning the
shear strength and shear stress and it provides a linear
safety conditions of any slope. Any wetting process
measure of safety. In Equation 2, Sr , n and γs are the
modifies several parameters which control the over-
degree of saturation, the porosity and the solid specific
all stability. In fact, the natural unit weight increases,
weight respectively.
the available shear strength decreases with a decrease
in soil suction and, when the soil becomes satu-
2.1.2 Infiltration model
rated, the pore water pressure becomes positive thus
The classical theory assumes that the net infiltration,
modifying effective stresses. In addition, the internal
q(t), results from a balance between rainfall, corrected
geometry of different soil strata having varying per-
by means of a runoff coefficient, and evapotranspira-
meabilities, plays a significant role in controlling the
tion. On the other hand the stochastic analysis of a time
hydraulic regime within the slope. An example devel-
series of data regarding rainfall and evapotranspira-
oped later (Villa Blasi slope) illustrates this comment.
tion provides the necessary parameters to identify q(t)
The variation of soil moisture conditions in the slope
from a stochastic point of view. The equations govern-
is controlled by the varying rates of infiltration which
ing the flow of water in a rigid partially saturated soil
in turn depend on rainfall intensity, soil conditions and
are (Bear & Bachmat 1991)
geometry. Except for the case of homogeneous masses
of impervious clay soils or extremely pervious forma- ∂
tions, experience shows that, in a large class of slopes, (Sr n) + div(vω ) = 0 (3)
failures occur during periods of heavy rainfalls. ∂t

pω
vω = −Kgrad z + (4)
2.1.1 Safety margin of slopes in partially γω
saturated soils
Fredlund et al. (1978) proposed the following expres- If a linear expression for the water retention curve
sion for the strength of a partially saturated soil: is adopted,

τ = c + (σ − pa ) tan φ  + ( pa − pω ) tan φb (1) Sr = Sr0 + as ( pω − pω0 ) (5)

where Sr0 is the degree of saturation for a reference

where (σ − pa ) and (pa − pw ) are the net normal stress
suction pω0 and as is a constant, Equations 3–5 become
and suction, (c , φ  ) are the effective cohesion and fric-
tion and tan φb is a suction-related friction coefficient
∂pω ∂ 2 pω K
which takes the value of φ  for very low suctions and = Cs 2 ; Cs = (6)
reduces progressively as suction increases. (However, ∂t ∂z nas γω

68
This equation may be normalized if, Then, the safety margin (2) is found as:

z Cs t N
2πnT
Z= , T = , u = [pω /γω H (q0 /K − 1)] − z/H M (Z, T ) = ω0 + (An 1 + Bn 2 ) cos
H H2
n=1
T0
q0 being a constant under steady state conditions and 2πnT
it results in, + (−An 2 + Bn 1 ) sin (11)
T0

∂u ∂ 2u where the different coefficients have been defined

= (7) in Appendix 1. The mean safety margin, ω0 , is a
∂T ∂Z 2
deterministic function of Z.
Figures 3a, b show the evolution of safety margin
Boundary conditions for an infiltration q(t) at the
at different depths of a slope subjected to sinusoidal
surface and a fixed ground water level at depth H are
infiltrations of widely different frequencies. It can

u = 0; Z =0 (8a)

∂u q(t) − q0 q1 (T )
= I (T ) = = (8b)
∂Z q0 − K q0 − K

If q0 is selected as the mean of the process q(t), I (T )

becomes a stochastic process of zero mean, which, on
account of the periodic nature of hydrologic events,
will be expressed as:


du  N
= I (T ) = An cos λ2n T + Bn sin λ2n T ;
dZ Z=1 n=1
(a)

2π n
λ2n = (9)
T0

where the coefficients An and Bn are non correlated

Gaussian random variables of zero mean and com-
mon variance (σn2 ) and T0 is a dimensionless reference
period.
Figure 2 shows a 6 year record of the process I (t) for
a meteorological station in the vicinity of Barcelona.
Plotted values represent averages of 10 days. Evapora-
tion data was obtained from actual monthly measure-
ments, in a 1.25 m diameter tank corrected to reduce
the free water surface to soil conditions. The solution
of equation (7) with (8) and (9) is


N
2π nT 2π nT
u(Z, T ) = Tn (Z) cos + Wn (Z) sin (10)
T0 T0
n=1

(b)

Figure 3. Evolution of safety margin at different depths

(p = 0 to 4 m) of a slope (α = 32◦ ) subjected to sinu-
soidal infiltration: (a) Low frequency; (b) High frequency.
Figure 2. Infiltration record measured in a meteorological K = 0.7 × 10−7 m/s, φ = 30◦ ; φb = 25◦ ; as = 10−5 m2 /N;
station in Barcelona. c = 1000 zN/m2 .

69
be observed the limited influence of high frequency νa = (σṀ /2πσM ) exp(−r 2 /2) (14)
components at points not very close to the surface. On
the other hand, high period oscillations of infiltration where σṀ is the standard deviation of the derivative
have the capability of penetrating deep into the slope. in time of the safety margin. Knowing the expression
Figure 3a shows also the delaying effect of infiltration (12) for the autocovariance function, both σM2 and σṀ2
into the soil. are found as

2.1.3 The safety margin as a random process 

N

in time—Correlation structure σM2 = σn2 (21 + 22 );

In order to analyze the slope reliability as a level- n=1

crossing problem associated with the evolution in time 

N
of the safety margin, it is first necessary to estab- σṀ2 = σn2 λ2n (21 + 22 ) (15a, b)
lish its correlation structure. It may be shown that n=1
the autocovariance function of the safety margin is
given by All of these expressions are used subsequently to
predict the reliability of an earth slope in a given

N hydrologic environmental condition (Barcelona area).
CM (T ) = σn2 (21 + 22 ) cos λ2n τ (12)
n=1 2.1.5 Application of the developed formulation
to the reliability of slopes in a given
The individual variances associated to each of these hydrologic environment
periodic terms reflect both the variability of the ‘‘exter- In all the cases presented below the soil properties are
nal’’ infiltration terms (through σn2 ) and the process of maintained constant. They correspond approximately
percolation, the modification of strength parameters to a partially saturated silty soil and they are φ  = 30o ;
of the soil and its weight and the (mechanical) defini- φ b = 25o ; as = 10−5 m2 /N. The parameterc of the
tion of stability—(through the term (21 + 22 ), which strength envelope was assumed to increase slightly
is defined in the Appendix). Once the covariance is with depth (1000p N/m2 , where p is the depth below
determined, standard procedures to analyze stochastic the surface) in order to represent non homogeneous
time records (see for instance Tretter (1976)) were used soil profiles, typical, for instance, in residual deposits.
to determine the power spectrum density function of Failure probabilities correspond always to a yearly
the safety margin. The analysis of a slope subjected period.
to the Barcelona infiltration record shows how the Figure 4 shows the variation of probability of failure
influence of high frequencies of hydrologic variation with the slope angle for different positions of the fail-
is increasingly damped at depth. ure plane. It is interesting to note the transition from
deep-seated surfaces, more likely to occur in steep
slopes, to shallow failures in gentle slopes. In fact the
2.1.4 The level crossing problem influence of rainfall induced moisture changes is deter-
For Gaussian processes, assuming that the rate of minant to provoke shallow instabilities. On the other
crossing a certain barrier follows a Poisson distribution
and that the time intervals spent in the ‘‘safe’’ regions
have a common exponential distribution whose mean
is the inverse of the mean rate of crossings, Vanmarcke
(1975) derived the following expression for the prob-
ability of no crossings in the interval 0 to t:
 
−νa t
LB (t) = [1 − φ(r)] exp (13)
1 − φ(r)

where r is the normalized crossing level, φ(r)√is related

to the error function (φ(r) = 1/2[1 − erf (r 2)]) and
νa is the mean rate of crossings above the level r. In
our case, if one works with the zero-mean safety mar-
gin process, M ∗ = M − ω0 , the normalized barrier
indicating slope failure is r = −ω0 /σM , where σM is Figure 4. Probability of failure of an infinite slope in par-
the standard deviation of safety margin. The mean rate tially saturated soils for different slope angles and depths of
of crossings above level r is given by failure plane.

70
 important result is, however, that the slope is especially
prone to shallow failure surfaces for a wide range of
slope angles. In other words, the critical slope angle
which marks the boundary between shallow or deep
failure planes (approximately at 37◦ if H = 5 m in the
example solved, see Figures 5 to 7) increases when H
increases. It is finally noted that the above results were
all obtained for rainfall records averaged over 10 days.
No significant influence was found when the average
period changed from 5 to 20 days.

2.2 A case history: ancona slides

2.2.1 Background
In December 1982, a large flow slide in overcon-
solidated clays destroyed the suburbs of the city of
Ancona (Italy). Several years later, a research project
was launched to investigate the behavior of shallow
Figure 5. Influence of soil permeability. H = 5 m; φ =
30◦ ; φb = 25◦ ; as = 10−5 m2 /N; c = 1000 p N/m2 . slides in clays in a Mediterranean climate. Villa Blasi
slope was selected for the study and a field investiga-
tion was set out. Inclinometers, as well as piezometers,
were installed and specimens were taken for special-
ized testing. Rainfall was recorded. A representative
profile of the slope is shown in Figure 7. Three layers
(α, β, γ) could be distinguished. The surface layer α
(3–4 m thick) is made of remoulded brown Ancona
clay with a significant proportion of organic matter. It
overlies a 4 to 10 m thick layer (β) of stiff brown silty
clay with sandy inclusions (brown Ancona clay). At
depth, the Pliocene substratum consists of stiff blue
silty clay with sandy inclusions (blue Ancona clay).

2.2.2 Soil properties

Dominant clay minerals are montmorillonite, illite and
chlorite. Calcite amounts to 22% of the total min-
eral composition. Brown and blue Ancona clays are

Figure 6. Influence of depth of ground water level in failure

risk for different slope angles and position of failure plane.

hand steep slopes have small (deterministic) safety

margins, decreasing with depth according to the sta-
bility model adopted, and their influence overrides the
strong moisture induced effects close to the surface.
The marked influence of permeability is shown in
Figure 5. Other conditions being equal the increase
in permeability reduces fast the risk of failure. For a
given infiltration history the developed pore pressure
gradients are smaller for larger K values and therefore
changes in soil strength are small.
The depth of groundwater level (Fig. 6) controls
the mean ‘‘dryness’’ of the soil in the sense that both Figure 7. Longitudinal profile of slope and position of the
Sr and pw are reduced. Larger strengths (implied by phreatic level. The position of piezometers (A, B, C, D, E and
higher suctions) result in increased reliabilities. The P) is also shown.

71
Figure 8. Water retention curve of brown Ancona clay.
Figure 9. Variation of apparent cohesion with suction for
Brown Ancona clay.

classified as CH or MH (WL = 52–64%; PI: 25–34%;

clay fraction: 40–55). Water content is low in these phases. Equilibrium conditions were first defined.
formations: 21%–32%. Oedometer tests indicated a An idealized geological sequence was simulated. The
significant overconsolidation stress: 2 to 3 MPa for soil was initially deposited under normally consoli-
layer β and 3 to 4 MPa for layer γ. In drained tri- dated conditions. Then, erosion (unloading) took the
axial tests, the brown Ancona clay exhibited a brittle slope to its actual geometry. For this phase on an elasto-
behaviour (brittleness index, IB = 0.40). plastic hyperbolic model, having parameters c = 95
Suction controlled oedometer tests were performed kPa, ϕ  = 27◦ , was used. This sequence tried to
to determine the effect of suction on volumetric defor- reproduce the actual preconsolidation stresses mea-
mations and on permeability. Experimental data on the sured in oedometer tests. The phreatic surface of the
variation of permeability and degree of saturation with deep hydrogeological regime was also defined at this
suction is shown in Figure 8. stage (see Fig. 7).
A coupled finite element flow-deformation code In the second phase of the analysis, rainfall infil-
(NOSAT) for saturated-unsaturated analysis, devel- tration is simulated. Average uniform monthly flow
oped at the Department of geotechnical Engineering rates were imposed as flow boundary conditions at the
and Geosciences of the UPC, was used in calculations. slope surface. If a positive pore water pressure is cal-
NOSAT solves the balance equations for water and air, culated at the surface (which is an indication that the
and the mechanical equilibrium. rainfall rate was larger than the infiltration capacity of
the soil), the boundary condition is changed to zero
2.2.3 Modelling Villa Blasi slope suction, which is a flooding condition.
The soil, in this case, was characterised by a non-linear The third stage in the analysis concerns deforma-
elastic behaviour. Volumetric deformations induced by tions and safety factors. Local safety factors were
suction changes were introduced by means of state sur- determined by comparing at some depths within the
faces relating void ratio, suction and mean net stress. slope available shear strength and existing shear stress.
Figure 9 shows the variation of apparent cohesion with The rainfall record actually used in computations
suction for the intermediate layer (brown clay). is shown in Figure 10. There was some uncertainty in
A first step in modelling was to approximate the the actual field permeabilities of layers α, β and γ. This
actual hydrogeologic conditions. Two flow regimes is key information to understand stability conditions.
were identified. A deep flow regime was characterised Time records of pore pressure changes measured in
by a phreatic level at nearly constant elevation in piezometers could be used, however, to define a per-
both wet and dry seasons (depth varies between 4 meability profile. The idea was to perform a sensitivity
and 14 m). Figure 7 shows some piezometer read- analysis varying the permeabilities of layers α, β and γ.
ings and the calculated position of the phreatic level. A total of 10 different permeability profiles, shown
In addition, a surface flow regime was directly con- in Table 1, were subjected to the same rainfall record
trolled by rainfall. Positive and negative pore water in the manner outlined above. Then, the most probable
pressures were measured during the year in the upper permeability layering was identified by comparing
few meters. Calculations were performed in three piezometer measurements and calculated values.

72
Figure 10. Rainfall records used in calculations.
Figure 11. Comparison of measured and calculated
piezometer readings (piezometer C). Depth of measuring
Table 1. Combination of soil layer permeability for sensi-
chamber 9.80 m (open tube piezometer in substratum).
tivity analyses.

Saturated water permeability kws (m/s)

Case Layer α Layer β Layer γ

a 10−9 10−9 10−9

b 10−9 10−9 10−9
c 10−8 10−9 10−9
d 10−7 10−9 10−9
e 10−6 10−9 10−9
f 10−8 10−8 10−9
g 10−7 10−8 10−9
h 10−6 10−8 10−9
j 10−7 10−7 10−9
k 10−6 10−7 10−9

2.2.4 Computational results

Figures 11 and 12 show two examples of the com-
parison made between field data and calculated pore
pressures. Figure 11 indicates the readings in an
electrical piezometer located in α layer.
Some combinations of layer permeability lead to
a consistent agreement with field data. (Case h, Figure 12. Computed depth of water level at piezometer
for instance). Suctions were often recorded. There P5699 for time t = 350 days after the beginning of the
is another interesting consideration about the rela- prediction exercise (October 1st, 1992). Depth of sensor:
tionship between failure risk and soil permeability. 2.80 m.
Examining the results of all computations, it became
clear that some combinations of layer permeability and
layering sequence led to the ‘‘strongest’’ response of the computed response of piezometer P5699, 350 days
the slope, in the sense of producing the highest pore after the initial date for the modelling.
pressures for a given sequence of rainfall. The gen- The combination Kα = 10−7 m/s, Kβ = 10−8
eral idea is that, given a climatic record, there are m/s, leads to the maximum value of the (posi-
particular combinations of soil permeability which are tive) pressure at the point considered. There are no
critical in terms of pore water pressure development. easy rules to predict a critical permeability profile
This idea is illustrated in Figure 12 which is a plot of because the computed water pressure integrates many

73
 Figure 14. Profiles of local safety factors (based on FE com-
putations) at a borehole C (Case h, homogeneous strength).

Figure 13. Computed and measured profiles of horizontal

displacements at borehole B.

phenomena: surface flow, infiltration, storage capacity

of the soil, previous wetting history of the soil etc.
The calculated and measured profile of horizontal
displacements is shown in Figure 13. It is clear that
a distinct shear surface has developed at the interface
between α and β layers. Most probably, a previously
existing shear surface was reactivated.

2.2.5 Stability analysis—Local safety factors

Field observations and the results of inclinometer
readings indicate that a simple planar model could Figure 15. Profiles of FE-based local safety factors at
represent the shallow slides. borehole C (Case h, heterogeneous strength).
Since stresses were determined also in a theoret-
ically more accurate analysis, an expression for the
safety factor may be found. It uses stresses and water
pressures calculated in the numerical model:

(σx + σy ) (σx − σy )
c + −
2 2 
cos 2α − τxy sin 2α − pw tan ϕ 
F(FEM ) =
(σx − σy )
sin 2α − τxy cos 2α
2
(17)

Calculated variations of local safety factors with

time and depth are shown in Figure 14. Absolute values Figure 16. Profiles of water pressure for a permanent
of F are high, however. If the soil strength is assumed flooding of the slope. (Case h, heterogeneous soil strength).
to be heterogeneous and remoulded parameters (c =
0; φ  = 24o ) are assigned to the upper α layer, the
calculated profiles of the local safety factor are shown
in Figure 15. Minimum safety factors are now found at The analysis performed so far only includes the
the α−β interface, where shear slips were detected by rainfall recorded in 13 months. Heavier rains cer-
the inclinometer. However, the minimum calculated tainly had occurred in the immediate past of the slope.
value (F = 1.5), is not able to explain the failure. An extreme rainfall event is to assume that the slope

74
Figure 17. Profiles of FE-based local safety factor. Hetero-
geneous strength distribution. Case h.

Figure 18. Location of the study slopes, generalized geo-

logical map of Singapore, schematic diagram of relative
surface is flooded during a certain time period. This position and arrangement of field instruments.
condition is illustrated in Figure 16, which shows the
calculated profiles of water pressure. Three days of Table 2. Description of the slopes.
permanent surface saturation leads to almost hydro-
static pore water pressure conditions in the upper Variable Yishun Mandai NTU-CSE
remoulded α layer. This situation is compatible with
suction values maintained in the lower β layer. This Area (m2 ) 165 180 140
situation brings a different picture of the safety condi- Vegetation cover (%) 100 100 100
tions of the slide (Fig. 17). Now surface instabilities Slope angle (◦ ) 23 31 27
are predicted after 12 hours of surface saturation of Slope height (m) 7 11 7
Slope length (m) 16.5 18.0 14.0
the slope. The entire upper α clay layer becomes soon Aspect West West West
unstable under extreme rainfall events. Slope form Plain Plain Plain

3 PART-B: SLOPE HYDROLOGY

AND RESPONSES TO CLIMATE Service Singapore 1997). The three recorded highest
CHANGES (RAHARDJO ET AL.) rainfalls in Singapore are; 512 mm in 1978, 467 mm
in 1969 and 366 mm in December 2006. The rain-
fall is usually greatest in the months of November to
3.1 Study area
January (the north-easterly monsoon) but rain falls in
Three slopes were selected for instrumentation to pro- all months of the year, with an average of 179 rainy
vide data for rainfall-induced slope failure studies. Out days in a year. Rainstorms are short, intense and gen-
of these three slopes (see Figure 18) one was located erally have a limited spatial extent, with intensities
in Yishun, one in Mandai and two in Nanyang Tech- typically ranging between 20 and 50 mm/h, although
nological University (NTU) campus (hereafter called short duration (5 min) rainfall intensities can exceed
the NTU-CSE slope). The characteristics of the four 100 mm/h (Sherlock et al. 2000). The potential evapo-
slopes are shown in Table 2. These slopes were selected ration rate in Singapore was calculated to vary between
because they were located in two major geological for- 5.16–7.53 mm/day (Gasmo 1997). All four slopes had
mations in Singapore, subjected to frequent shallow grass as the vegetative cover.
landslides. The geology of the study sites consists of residual
The climates at the study sites are hot and humid soils from two major geological formations (Public
equatorial, with no marked dry season. The temper- Works Department 1976); (a) the Bukit Timah Gran-
atures vary little throughout the year with an annual ite (Yishun and Mandai) which occupy the north and
average temperature of 26.6◦ C and a mean relative central-north region of Singapore; and (b) sedimentary
humidity of 84% (Meteorological Service Singapore rocks of the Jurong Formation (NTU-CSE), occu-
1997). The average annual rainfall in Singapore varies pying the west and southwest region of Singapore
between 2000 mm around the fringes of the island to (see Figure 18). These two residual soils comprise
about 2300 mm in the central region (Meteorological two-thirds of Singapore’s land area.

75
3.2 Field instrumentation and data collection of the slopes as obtained from pressure plate tests are
shown in Figure 20. Elaborate discussion on the soil
The slope responses were characterised through mea-
properties of these slopes are presented in Rahardjo
surement of the following variables: (i) rainfall inputs
et al. (2004b).
to the slopes (ii) runoff generation from the slopes
(particularly NTU-CSE slope) and (iii) changes in
pore-water pressures in response to rainfall.
3.3.2 Evaluation of slope responses in terms
Rainfall was recorded at each slope with a tipping-
of pore-water pressure changes due
bucket rain gauge. Runoff was measured using a
to rainfall
perspex flume (Rahardjo et al. 2004a). Corrugated
Slope responses to rainfall were assessed through the
zinc sheets, 300 mm high and driven about 100 mm
pore-water pressure changes in the slope. The ten-
into the ground bordered each instrumented slope
siometer readings from the slope crest (row A) were
(Figure 18). The boundaries guided the runoff into
plotted with the corresponding rainfall readings to
the perspex flume at the lower end of the plot where
indicate the seasonal pattern of pore-water pressures at
the surface runoff was measured using a capacitance
Yishun slope (Figure 21). The tensiometers recorded
water-depth probe installed in the flume. The water-
increases in pore-water pressures at all levels on wet
depth probe was connected to a data logger that stored
days. The deeper soil layers (3 m depth) at the slopes’
runoff data every 10 s during rainfall. Runoff measure-
crest and toe maintained positive pore-water pressures
ments for both simulated and natural rainfalls were
for all the 420 monitored days, while frequent changes
conducted particularly in NTU-CSE slope.
in pore-water pressure from negative to positive were
Pore-water pressure changes in response to climatic
recorded for shallow depths (50 cm) in response to
changes were recorded using jet-fill tensiometers. In
rainfall (Figure 21).
addition to the tensiometers, piezometers and tem-
The soils at shallower depths are in close proxim-
perature sensors were installed at various locations
ity to the atmosphere and slopes’ vegetation. As a
in the slopes. Figure 18 shows the details of field
result, the soils at shallow depths are easily and fre-
installation of the instruments. All sensors within
quently influenced by rainfall and evapotranspiration
each instrumented slope were monitored continuously
compared to deeper soil layers. During the monitor-
and automatically by a field data acquisition system
ing period, negative pore-water pressure development
(DAS). The DAS was programmed to scan the sensors
as low as −57 kPa was observed at shallow depths
at 4 h intervals during periods of no rain and at 10 min
(Yishun slope, Figure 21). Positive pore-water pres-
intervals during rainfall and continuing at the same
sures were also observed at all soil depths after a signif-
rate until 30 min after cessation of rainfall. A rainfall
icant rainfall and appear to be a common phenomenon
event would trigger the rain gauge, and data for sur-
in the monitored slope (Figure 21). Infiltration into the
face runoff were collected during rainfall events only.
slope does not lead to a constantly wet soil condition,
More details on field instrumentation can be found in
as is seen in the rapid matric suction recovery during
Rahardjo et al. (2007).
dry periods (Figure 21). The average daily matric suc-
tion recovery rates were 5 kPa/day for shallow depths
and 1 to 3 kPa/day for greater depths. The instant
3.3 Results and discussion
response of pore-water pressure to the infiltrating rain-
The results of the field tests, laboratory tests and water indicates considerable infiltration on the grass
field monitoring are presented in two forms. First, covered slopes.
field test and laboratory test results used to char- To illustrate the sensitivity of pore-water pressures
acterise the engineering properties of the slopes are to rainfall and to reflect the hydrologic response of
presented in the next section. Second, field moni- the slopes during a wet and a dry period, the record
toring and data analyses results are presented in the of pore-water pressures and rainfall for a period of 6
subsequent sections to show how slope responses were weeks was selected from the time series of the Yishun
evaluated in terms of (i) pore-water pressure changes slope as shown in Figure 22, which illustrates the
and (ii) runoff generation and infiltration amount. decreasing pore-water pressures during the dry period
from February 12 to March 8, 1999. A small storm
on March 8 only affected the pore-water pressures at
3.3.1 Site observation a shallow depth (0.5 m) due to the small amount of
The geotechnical properties of the residual soils at the infiltrated water. After the storm the pore-water pres-
four slopes derived from laboratory tests and field tests sures at 0.5 m depth started to decrease again due to the
are shown in Table 3. The information obtained from evapotranspiration processes. On the other hand, sev-
site investigation and laboratory tests was used to pro- eral rainfall events during the wet period from March
duce simplified soil profiles for the slopes as shown in 9 to March 16, 1999, caused the pore-water pressures
Figure 19. The soil water characteristic curve (SWCC) at all depths (0.5, 1.7, and 2.9 m) to increase.

76
 Table 3. Soil properties of the slopes.

Average* properties

Index Shear strength

Hydraulic
Slope Layer Description USCS w (%) LL (%) PL (%) Fines (%) ρ (Mg/m3 ) Gs c (kPa) φ  (◦ ) Cc ks (m/s)

Yishun 1 Moderate to low

plasticity silt ML 35 45 31 58 1.88 2.688 12 33 0.30 7.2 × 10−7
2 Moderate plasticity silt MH 52 57 40 78 1.76 2.714 13 29 0.64 –
3 Silty sand SM 19 42 26 34 2.05 2.667 35 31 – –

Mandai 1 Silty sand SM 25 55 31 41 2.02 2.684 12 30 0.19 9.5 × 10−6

77
2 Clayey sand SC 26 88 34 40 1.94 2.686 14 31 – –
3 Moderate to low
plasticity silt MH–ML 36 52 34 60 1.89 2.680 10 29 0.24 –
4 Silty sand SM–MM 23 45 27 46 1.95 2.680 0 28 0.20 –

NTU-CSE 1 Moderate to high

plasticity silt and clay CH–MH 30 65 35 94 – – 139 34 0.5 1.9 × 10−7
2 Low plasticity clay CL to SC–CL 19 36 23 74 2.10 2.725 107 36 – 8.0 × 10−7
3 Silty sand SM-ML 12 29 23 27 2.32 – 167 37 – 2.8 × 10−9

USCS = Unified Soil Classification System (ASTM, 1997); w = Water content; LL = Liquid Limit, PL = Plastic Limit; ρ = Total density; c = Effective cohesion; φ  = Angle
of friction; Gs = Specific gravity; Cc = Compression index; and ks = saturated permeability.
* Average of 2 to 9 samples, on some occasions there were only one sample.
– Results were not available due to sample damage or technical difficulty.
 Figure 21. Time series of rainfall and pore-water pressure
at Yishun slope.

Figure 19. Generalized soil profile of the four slopes (a)

Yishun slope, (b) Mandai slope, (c) NTU-CSE slope.

Figure 22. A trace of pore water pressure and rainfall record

in Yishun slope showing tensiometer response to a dry period
and wet periods.

Tensiometer readings at all depths and all loca-

Figure 20. Soil-water characteristic curves of Yishun,
Mandai and NTU-CSE slope. tions (6 rows × 5 depths) in the Yishun slope for
two selected periods (March 8 and March 16, 1999)
were used to produce the pore-water pressure contours
shown in Figures 25 and 26. These figures illustrate
Pore-water pressure records in row A at all depths the magnitude and distribution of pore-water pres-
of the Yishun slope are plotted for four different times sure within the entire slope profile and identify the
during a dry period (Figure 23) and four different movement of water in the subsurface regime.
events during a wet period (Figure 24). These two specific events were chosen to observe
These figures show the pore-water pressure distri- the magnitude and distribution of pore-water pres-
bution across the soil depth and the progressive vertical sures across the entire slope (that is slope response)
movement of water (infiltration and evapotranspira- at the end of a dry period (March 8) and a wet
tion) during these two periods. Figure 23 indicates period (March 16) period (Figure 22). Figure 25 shows
that the soil profile dries from the surface downwards, widespread negative pore-water pressure prevailing
owing to evaporation and transpiration during dry peri- across the entire slope during the dry period. The
ods. Figure 24, in conjunction with the rainfall events groundwater table from the piezometric readings (near
shown in Figure 22, shows the successive increase in slope crest, midslope, and toe) for this period is also
pore-water pressure along the soil depths in response shown in Figure 25. However, the slope experiences
to cumulative rainfall amounts during the wet period. more matric suction development at the crest than

78
 Figure 25. Magnitude and distribution of pore-water pres-
sure along the soil profile in Yishun slope on 8 March 1999
at the end of a dry period.

Figure 23. Pore water pressure profile in row A of Yishun

slope showing the advancement of a drying event at different
depths.

Figure 26. Magnitude and distribution of pore-water pres-

sure along the soil profile in Yishun slope on 16 March 1999,
after a wet period with 98 mm of cumulative rainfall.

saturation to near saturation conditions throughout the

soil profile at higher rainfall rates.
The downslope movement of water becomes more
Figure 24. Pore-water pressure profile in row A of Yishun
obvious from the contours plots of the distribution of
slope showing the advancement of the wetting at different
depths during a wet period. hydraulic heads shown in Figures 27 and 28.
As water flow is caused by the hydraulic head gradi-
ent, a comparison of total hydraulic heads (composed
of the pressure head and elevation head, assuming
at the toe. This reflects the variability in hydrologic velocity head to be negligible) at the slope crest
response within the same slope and could be because, and toe (both during dry condition, Figure 27, and
in addition to evaporation, there is preferential lateral wet condition, Figure 28) shows the existence of a
movement of water downslope. Figure 26 shows the hydraulic gradient during both dry and wet periods.
widespread development of positive pore-water pres- The hydraulic gradient during the wet period is greater
sure across the entire slope after the wet period, with than during the dry period, which suggests there is
about 98 mm of cumulative rainfall (from Figure 22). more downslope flow during a wet condition than
The groundwater table location from the piezometric during a dry condition (assuming hydraulic isotropy).
readings for this period was not conclusive as they did Decreases in matric suction were observed during
not show any significant rise in the water table (instru- the analyses of pore-water pressure response to rainfall
ment malfunction), and therefore the water table is not at almost all magnitudes of rainfall events. Therefore,
shown in Figure 26. The pore-water pressure distri- the correlation between the increase in pore-water
bution during this period is fairly uniform along the pressure uw , (The difference in pore-water pressure
slope face, indicating that significant rainfall brings before and after a rainstorm) and rainfall amount, R,
uniformity of pore-water pressures in the soil profile. was tested. This was done by splitting the rainfall
The widespread development of positive pore-water data into daily totals and taking the algebraic dif-
pressures is a consequence of the increasing degree of ference in pore-water pressure before and after daily

79
Figure 27. Magnitude and distribution of hydraulic head
along the soil profile in Yishun slope on 8 March 1999 at the
end of a dry period.
Figure 29. Relationship of increase in pore-water pressure
at 50 cm depth to daily rainfall amount for all instrumented
slopes.

to assess the infiltration effects on the slope due to

natural rainfall events because natural rainfall inten-
sity and duration vary within a rainfall event and from
event to event. To understand infiltration effects on
the residual soil slope under controlled rainfall condi-
tions 10 simulated rainfall experiments (5 experiments
during February of 1998 and another 5 experiments
between 17 November 1998 and 5 January 1999) were
Figure 28. Magnitude and distribution of hydraulic head conducted on NTU-CSE slope. The runoff data were
along the soil profile in Yishun slope on 16 March 1999, analysed for the 27 natural and 10 simulated rainfalls
after a wet period with 98 mm of cumulative rainfall. to determine total runoff volume, total infiltration,
peak intensity of each rainfall event. The total runoff
amount was calculated from integration of the area
rainfall, in row A at the depth of 50 cm. A regres- under the runoff hydrograph. From the time series of
sion analyses was performed to fit the data sets of the rainfall and pore-water pressures shown in Figures 21
increase in pore-water pressure, uw , and daily rain- and 22 and similar plots for other slopes, rainfall
fall amount, R. A semi-logarithmic equation of the amounts of less than 0.5 mm per day appear to have
form, uw = a2 + b2 log R, where a2 and b2 are coef- no significant effect on the pore-water pressures, even
ficients, best described the data set. Figure 29 shows at shallow depths. Therefore, interception by the slope
the relationships between the increase in pore-water vegetation (grass) was estimated at 0.5 mm from the
pressure at 50 cm depth of row A, and the daily rainfall time series of pore-water pressures. Total infiltration
amount for the three slopes. The equations in Figure 29 was estimated for all runoff measurements by subtract-
indicate that both coefficients a2 and b2 increase with ing total runoff and an interception loss of 0.5 mm from
increasing particle sizes (see Table 2). Mandai, being total rainfall.
rich in coarse particles, shows higher values of uw Figure 30 plots runoff amounts against rainfall
than other slopes. This suggests that the increase in amounts for the 27 monitored natural rainstorms. A
pore-water pressure is not only dependent on the rain- linear regression fitted to the data points shows an
fall amount, but it is also affected by soil properties intercept on the x-axis, indicating that a threshold
and antecedent soil moisture. Although uw values rainfall amount of about 10 mm must be exceeded to
are different for Yishun, Mandai, and the NTU-CSE, produce a significant runoff. Hydrologic responses of
they appear to increase with the rainfall amount, and slopes are locally and geographically variable, causing
the rate of increase tends to decline at a daily rainfall difficulty in comparing results. The limited references
amount greater than about 10 mm. from Southeast Asia further add to this difficulty. The
threshold rainfall observed in this study is compara-
3.3.3 Evaluation of slope responses in terms ble to that reported by Tani (1997) for hilly area in
of runoff generation and infiltration Okayama, Japan, but is about half of that reported by
Rainfall and runoff data were collected for 27 natu- Premchitt et al. (1992) for Hong Kong in a sub-humid
ral rainstorms in the NTU-CSE slope. It is difficult tropical climate.

80
 Figure 32. Rainfall and runoff hydrograph from a simulated
rainfall event on 9 February 1998 in the NTU-CSE slope.
Figure 30. Relationship between storm rainfall (natural
rainfall) and runoff amount at NTU-CSE slope.

Figure 33. Percent infiltration as a function of rainfall

amount.
Figure 31. Rainfall and runoff hydrograph of a natural
storm event on 12 December 1998 recorded at NTU-CSE
slope. Rainfall and runoff hydrograph from a simulated
rainfall event during February 1998 on the NTU-CSE
slope is shown in Figure 32. The infiltration rates
Figure 31 shows the rainfall and runoff hydrographs shown in Figure 32 are derived by subtracting runoff
of a natural composite storm recorded at the NTU-CSE rates from rainfall rates. The hydrograph (Figure 32)
slope on December 12, 1998, with a total rainfall of shows that there was no runoff during the early part
45 mm, and two peak rainfall intensities of 240 and of the simulated rainfall event. During this period all
120 mm/h, resulting 18 mm of total runoff. This com- the rainfall water is lost as infiltration (a fraction may
posite storm is selected because it reflects the slope have also been retained by the slope vegetation as inter-
response to rainfall during dry and wet soil conditions. ception). It is noted that during the rainfall event of
The first runoff hydrograph is characterized by a flash 9 February 1998 the runoff rate did not exceed the
flood, which decreases quickly and represents over- infiltration rate and the rise in runoff rate is slow. This
land flow under dry soil conditions. The first rainfall is due to the relatively dry soil conditions that prevailed
with 28 mm of rainfall produced 10 mm of runoff, before the rainfall event.
which is about 35% of the rainfall. In contrast, the In Figure 33, infiltration amounts (as a percentage
second rainfall event with similar rainfall characteris- of total rainfall) are plotted against rainfall amounts
tics and a total amount of 17 mm, but under wet soil from 27 natural and 10 simulated rainfalls monitored
conditions, produced about 8 mm of runoff, a conse- in the NTU-CSE slope. It appears from Figure 16 that
quence of increasing saturation condition of the slope. rainfall events producing small total amounts of rain-
This runoff is about 47% of the rainfall. fall may contribute fully to infiltration. This again

81
suggests the existence of a threshold rainfall amount. • In relatively gentle slopes shallow failures induced
Any rainfall below this amount will not produce by moisture changes are more likely than deep
any runoff, and the whole rainfall may end up as seated failures.
infiltration. • However, if the slope angle increases deep failures
With reference to Figure 33 (broken line), this become more frequent
threshold appears to be about 10 mm of total rain- • For the cases analyzed here, increasing the perme-
fall. Beyond the threshold rainfall, the percentage ability reduces the risk of failure
of rainfall contributing to infiltration decreases with • In general, the deeper the water tables the lower
an increase in total rainfalls. The infiltration amount the risk of sliding. Shallow failures tend to develop,
could decrease to about 40% of the rainfall (Figure 33) with more probability, if the water table is deep.
for rainfall events that produce a high total amount of
rainfall. This, however, does not mean that the total
4.2 PART-A: The case history
infiltration amount is less during rainfall events with
a higher total rainfall than the total infiltration dur- Two deformation mechanisms were identified in the
ing rainfall events with a smaller total rainfall. For slope: a creep-type displacement, which was detected
example, with 40% of the rainfall contributing to infil- along the full depth of the soil investigated (around
tration, a 100 mm rainfall (higher total amount) would 12 m) and a surface planar slide. The first mechanism
result in a total infiltration of 40 mm. With 100% is interpreted as a deformation associated with volume
of the rainfall contributing to infiltration, a 10 mm changes of the over-consolidated clays as the water
rainfall would produce only 10 mm of total infiltra- pressures change in time as a reaction to rainfall events.
tion. The data suggest that in residual soil slopes total Water pressures at the upper two weathered layers are
infiltration could range between 40% and about 100% controlled by the atmospheric weather.
of the total rainfall depending on the rainfall amount. A fairly good agreement between computed and
The relationship (Figure 33) derived from the rainfall measured water pressures is achieved when the three
records in the residual soil slope has practical signifi- identified layers are characterized by three different
cance. If the rainfall amount is known, Figure 33 could permeabilities.
indicate the fraction of the rainfall that could become For a given climatic record, the critical situation of
infiltration. This may be useful for seepage analysis a given slope (in the sense of reacting with the maxi-
that requires this information as flux boundary condi- mum development of water pressures) is obtained for
tions. More rigorous interpretation on the hydrological a particular combination of layer permeabilities. In
responses of these slopes can be found in Rezaur et al. other words, given a soil profile and geometry and
(2003); Rahardjo et al. (2005). its associated permeabilities and additional water flow
parameters, there exist rainfall records which lead
to a maximum ‘‘reaction’’ of the slope in terms of
pore pressure development. Permeability and water
4 CONCLUSIONS retention are therefore fundamental properties in slope
stability analysis.
4.1 PART-A: The stochastic model As deformation accumulates, peak strengths are
A method to analyze the risk of failure of slopes in par- attained in surface layers and eventually remoulded
tially saturated soils induced by climatic changes has and even residual strength conditions develop.
been presented. Spectral and correlation functions for Weathering mechanisms result also in a change in
the safety margin of the slope were explicitly derived permeability, which is stronger the closer to the sur-
in terms of the spectral representation of the infil- face. Water pressures recorded are consistent with
tration record. Some results of the theory of random a decrease of permeability with depth. (10−6 m/s,
processes were then used to find expressions for the 10−8 m/s, 10−9 m/s).
failure risk in a given period (1 year). The method Permeability transitions lead to peak pressures com-
has been applied to the hydrologic conditions prevail- puted at the interfaces. Strength degradation (accumu-
ing in the mediterranean coast close to Barcelona and lated straining) and peak water pressures (positive or
to slopes in a partially saturated silty soil whose mean negative) result in the development of a sliding surface
strength, permeability and suction characteristics were at the α−β interface, where minimum safety factors
estimated from previous work and past experience. It are consistently found.
was found that, Once a planar sliding surface has developed the con-
ceptual model of slope motion is simple: an upper
• Cyclic infiltration/evaporation records exhibiting a reworked layer slides on top of a critical surface at
large time period of occurrence can penetrate deep residual or near-residual conditions.
into the soil. High frequency components only affect Periods of activity are dictated by suction changes
shallow depths and dampen fast at depth. in the upper few meters which depend critically on

82
the slope geometry, flow boundary conditions, rainfall Corominas, J. 2000. Landslides and climate. Proc. 8th Int.
record, flow parameters (permeability, water retention Conf. Landslides. Cardiff. pp. 1–31.
properties) and its spatial variation. Fredlund, D.G., Morgenstern, N.R., & Widger, R.A.
1978. The shear strength of unsaturated soils. Canadian
Geotechnical Journal 15(3): 313–321.
4.3 PART-B: Slope hydrology Gasmo, J.M. 1997. Stability of Unsaturated Residual Soil
and responses to climate changes Slopes as Affected by Rainfall. Master of Engineer-
ing Thesis. School of Civil and Structural Engineering,
The pore-water pressure profile of the slopes presented Nanyang Technological University, Singapore.
in part-II shows distinct differences in slope hydro- Meteorological Service Singapore. 1997. Summary of
logic response during dry and wet conditions. The Observations (annual publication), Singapore.
relationships between the increase in pore-water pres- Pitts, J., & Cy, S. 1987. In situ soil suction measurements in
relation to slope-stability investigations in Singapore. E.T.
sure and daily rainfall allows for an estimate of the
Hanrahan, T.L.L. Orr, and T.F. Widdis, eds., Proc., 9th
rise in pore-water pressure due to rainfall. However, European Conf. on Soil Mechanics and Foundation Engi-
use of these relationships should be restricted to the neering. Vol. 1, Balkema, Rotterdam, The Netherlands,
field conditions under which measurements have taken 79–82.
place. Poh, K.B., Chuah, H.L. & Tan, S.B. 1985. Residual gran-
The results of natural and simulated rainfall— ite soils of Singapore. Proceedings of the 8th South-
runoff experiments conducted on the test slope sug- east Asian Geotechnical Conference, Kuala Lumpur,
gest that a large proportion of the rainfall contributes Malaysia. 11–15 March, 1985. 1(3):1–9.
to infiltration in the residual soil slope. A rainfall may Premchitt, J., Lam, T.S.K., Shen, J.M. & Lam, H.F. 1992.
Rainstorm runoff on slopes. GEO Rep. 12, Geotechnical
contribute from 40% to about 100% of its total rain-
Engineering Office, Hong Kong.
fall as infiltration (assuming negligible interception Public Works Department. 1976. The geology of the Republic
losses) depending on the rainfall amount. This infor- of Singapore, Singapore.
mation is useful for seepage analyses that require the Rahardjo, H., Aung, K.K., Leong, E.C. & Rezaur, R.B.
total infiltration amount as an input parameter. There 2004b. Characteristics of residual soils in Singapore
appears to be a threshold rainfall of about 10 mm as formed by weathering. Engineering Geology 73:
to generate runoff. The characteristics of infiltration 157–169.
processes, runoff generation, and pore-water pressure Rahardjo, H., Lee, T.T., Leong, E.C. & Rezaur, R.B. 2005.
changes identified in this study have relevance for Response of a residual soil slope to rainfall. Canadian
Geotechnical Journal 42: 340–351.
the assessment of rainfall-induced slope instability in
Rahardjo, H., Lee, T.T., Leong, E.C. & Rezaur, R.B.
residual soil slopes under similar climatic conditions 2004a. A flume for assessing flux boundary characteris-
in different geographic regimes. tics in rainfall-induced slope failure studies. Geotechnical
Testing Journal 27(2): 145–153.
Rahardjo, H., Leong, E.C. & Rezaur, R.B. 2007. Effect
ACKNOWLEDGEMENT of antecedent rainfall on pore-water pressure distribu-
tion characteristics in residual soil slopes under tropical
rainfall, Hydrological Process. 21: (in press).
This work presented in Part-B was funded by a
Rezaur, R.B., Rahardjo, H., Leong, E.C. & Lee, T.T. 2003.
research grant from National Science and Technol- Hydrologic behavior of residual soil slopes in Singapore.
ogy Board, Singapore (Grant: NSTB 17/6/16). The Journal of Hydrologic Engineering, ASCE 8(3): 133–144,
authors gratefully acknowledge the field assistance of Sherlock, M.D., Chappell, N.A. & McDonnell, J.J. 2000.
the Geotechnics Laboratory staff, School of Civil and Effects of experimental uncertainty on the calculation
Environmental Engineering, Nanyang Technological of hillslope flow paths, Hydrological Processes. 14(14):
University, Singapore, during the field instrumen- 2457–2471.
tation, trouble shooting and data collection for this Tan, S.B., Tan, S.L., Lim, T.L. & Yang, K.S. 1987.
study. Landslide problems and their control in Singapore.
Proc., 9th Southeast Asian Geotechnical Conf., Southeast
Asian Geotechnical Society, Bangkok, Thailand. Vol. 1,
25–36.
REFERENCES Tani, M. 1997. Runoff generation processes estimated from
hydrological observations on a steep forested hillslope
ASTM. 1997. Annual book of ASTM standards, Philadel- with a thin soil layer. Journal of Hydrology 200(1–4):
phia, 04.08–04.09. 84–109.
Bear, J. & Bachmat, Y. 1991. Introduction to Model- Toll, D.G., Rahardjo, H. & Leong, E.C. 1999. Land-
ing of Transport Phenomena in Porous Media. Kluwer. slides in Singapore. 2nd Int. Conf. on Landslides, Slope
Dordrecht. Stability and the Safety of Infra-structures, Singapore.
Chatterjea, K. 1994. Dynamics of fluvial and slope processes 269–276.
in the changing geomorphic environment of Singapore. Tretter, S.A. 1976. Introduction to Discrete Time Signal
Earth Surface Processes and Landforms 19: 585–607. Processing. Wiley. NY.

83
  
Vanmarcke, E. 1975. On the Distribution of the First
ω0 = Sr0 − as pω0 − γω (q0 K −1 )
Passage Time for Normal Stationary Random Pro-
cesses. Journal of Applied Mechanics Division. A.S.M.E.

March: 215–220. H +h
Wieczorek, G.F. 1996. Landslide triggering mechanisms, In
γω n + γs (1 − n)
2
A.K. Turner & R.L. Schuster (eds.) Landslides: investiga-
tion and mitigation. TRB Special Report, 247. National × (tan φ  cos α − sin α) cos α (H − h)
Academy Press, Washington, 1996, 76–90.  q 
+ c + tan(φb ) pa − hγω
0
−1
K
5 APPENDIX: COEFFICIENTS 1 = 11 β11 − 12 β12
FOR EQUATIONS (10) AND (11)
2 = 12 β11 + 11 β12
ω1 H (−ω11 + ω12 )
Tn = ω21 Cn + ω22 Dn 11 = − ω2 ω21
2αn
Wn = −ω22 Cn + ω21 Dn ω1 H (ω11 + ω12 )
12 = − ω2 ω22
f1 (Bn − An ) + f2 (Bn + An ) 2αn
Cn =  
2αn (f12 + f22 ) ω11 = cos αn (e−αn + eαn ) − cos αn Z e−αn Z + eαn Z
f2 (Bn − An ) − f1 (Bn + An ) ω12 = sin αn (e−αn − eαn ) − sin αn Z(e−αn Z − eαn Z )
Dn =
2αn (f12 + f22 )
 f2 − f1
β11 =
αn = 1 2λn 2αn (f12 + f22 )
  f2 + f1
f1 = cos αn e−αn + eαn β12 =
  2αn (f12 + f22 )
f2 = sin αn −e−αn + eαn q   
  − 1 cos α cos α tan φ  − sin α
0
ω1 = γω2 nas H
ω21 = cos αn Z e−αn Z − eαn Z K
  q 
0
ω22 = sin αn Z −e−αn Z − eαn Z ω2 = tan φb γω H −1
K

84
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Soil nailing and subsurface drainage for slope stabilization

W.K. Pun
Geotechnical Engineering Office, Civil Engineering Department, Government of the Hong Kong Special
Administrative Region, China

G. Urciuoli
Department of Geotechnical Engineering, Via Claudio, Napoli, Italy

ABSTRACT: A wide range of slope stabilization and protective measures are available. Soil nailing and
subsurface drainage are amongst the very commonly used techniques for slope stabilization. The concept of soil
nailing involves creating a stable block of composite material by strengthening the insitu ground with soil nails.
The interaction between the ground and the soil nail is complex, and the mobilization of forces in the soil nail
is dependent on many factors. The mechanism of subsurface drains in slopes involves a decrease in pore water
pressures in the subsoil, and consequently an increase in effective stresses and soil shear strength in the whole
drained domain. This paper gives an overview of the mechanism of soil-nailed system and subsurface drainage
measures and presents some geotechnological developments related to their applications.

1 INTRODUCTION Zaruba & Mencl 1982, Leventhal & Mostyn 1987,

Schuster 1992, Hausmann 1992, Fell 1994, Holtz &
Landslides have resulted in the loss of human lives Schuster 1996, Perry et al, 2003, Ho 2004). This paper
and properties in many parts of the world. To com- focuses on the developments in geotechnology in the
bat landslide risk, a wide range of risk mitigation two of the common stabilization measures, viz. soil
measures are available. These range from hard engi- nailing and subsurface drainage.
neering measures of slope stabilization and landslide
protective works to soft community means of pub-
lic education. Stabilization works aim at reducing the
likelihood of failure of a slope whereas the other mea- 2 SOIL NAILING TECHNOLOGY
sures reduces the risk by minimising the consequences
of slope failures. 2.1 Introduction
The range of slope stabilization works may be The soil nailing technique was developed in the early
categorized as follows (Ho 2004): 1960s, partly from the techniques for rock bolting and
a. surface protection and drainage, multi-anchorage systems, and partly from reinforced
b. subsurface drainage, fill technique (Clouterre 1991, FHWA 1998). The
c. slope regrading, New Austrian Tunnelling Method introduced in the
d. retaining structures, early 1960s was the premier prototype to use steel bars
e. structural reinforcement, and shotcrete to reinforce the ground. With the increas-
f. strengthening of slope-forming material, ing use of the technique, semi-empirical designs for
g. vegetation and bioengineering, soil nailing began to evolve in the early 1970s. The
h. removal of hazards, and first systematic research on soil nailing, involving
i. special materials and techniques. both model tests and full-scale field tests, was carried
out in Germany in the mid-1970s. Subsequent devel-
Further detail breakdown of the various categories opment work was initiated in France and the United
of slope stabilization works and landslide protective States in the early 1990s. The result of this research
measures are depicted in Figure 1. and development work formed the basis for the for-
The development, characteristics, application and mulation of the design and construction approach for
performance of various types of slope stabilization the soil nailing technique in the subsequent decades.
works have been reported by many researchers and The concept of soil nailing involves creating a sta-
practitioners (e.g. Hutchinson 1977, Veder 1981, ble block of composite material by strengthening the

85
 'impermeable' cover
vegetated cover with geotextile protection
toe erosion protection measures
Surface protection infilling of tension cracks
and drainage surface drainage channels
inverted granular filter over discharge points
short relief drains
trench (counterfort) drains
cut-off drains
raking drains
vertical drainage wells pumped
drainage galleries/adits gravity drainage
Subsurface drainage siphon drains
geosynthetic drains
vertical sand/gravel drains
pressure relief wells
electro-osmosis drainage
vacuum dewatering
blasting

Slope regrading cut back

toe weighting
embedded walls (cantilever or tie-back)
Retaining structures gravity walls
composite walls
reinforced fill materials: metallic, polymeric,
woven/non-woven, fibre glass,
electro-kinetic geo-synthetics
Structural construction: reinforced earth/anchored earth
reinforcement (usually horizontal reinforcement, but may have vertical reinforcement),
hybrid system (e.g. segmental block and gabion)

insitu reinforcement soil nails materials: metallic, polymeric

(usually unstressed but may be prestressed) installation: drill & grout,
jacked, driven,
prestressed anchors (large presumes) ballasted
soil dowels large diameter bored piles
driven piles
reticulated micropiles hand-dug caissons
rock bolts and rock dowels
lime nails/piles
other techniques (e.g. electro-osmotic tie-back)
chemical admixtures aerated cement suspension and mortar
lime
flyash/lime slurry
Calcite In-situ Preciptation System (CIPS)
recompaction replacement of failed material with compacted
fill
surface recompaction of loose fill slopes
Strengthening dynamic compaction
of slope-forming compaction induced by displacement piles
material shear trenches trenches backfilled with compacted gravel or
(exclude structural no-fines concrete
reinforcement and grouting fracture grouting
drainage) compacting grouting
permation grouting
void infilling
others dentition, buttressing of loose rock blocks,
various ground improvement techniques
tree root reinforcement
Vegetation and bioengineering
tree - live staking, live fascines, brushlayering, willow poles,
root reinforcement
bioengineering branchpacking, etc. ('living approach')
reinforced grass system, Biobund system, etc. ('combined living and non-living' approach)

netting and meshing (including concrete grillage)

Defensive measures rock ditches
debris barriers
scaling of loose rock blocks
Removal of h azards removal of boulders
trimming of local over-steepened slope
excavation of local loose fill
Special materials lightweight fill, EPS, Pneusol, Rubber Soil TM, reuse of waste materials
and techniques underdrainage to landslide trail, thermal treatment,
short-circuit conductors, ion-exchange

Figure 1. Classification of slope stabilization and protective measures (after Ho 2004).

86
insitu ground with soil nails. This requires that the • Long soil nails are difficult to install, and thus it
soil nails are installed at close spacing, both horizon- renders the use of the soil nailing technique more
tally and vertically. The following are typical merits of difficult in dealing with deep-seated landslides and
adopting the soil nailing technique in respect of ease sizeable slopes.
of construction, cost and performance (GEO 2008): • As soil nails are not prestressed, mobilisation of
soil-nail forces will be accompanied by ground
• It is suitable for sites with difficult access because deformation. The effects on nearby structures, or
the construction plant required for soil nail installa- services may have to be considered, particularly in
tion is small and mobile. the case of soil-nailed excavations.
• It can more easily cope with site constraints and • Soil nails are not effective in stabilising localised
variations in ground conditions encountered dur- steep slope profiles, back scarps, overhangs or in
ing construction, e.g., by adjusting the location and areas of high erosion potential. Suitable measures,
length of the soil nails to suit the site conditions. e.g., local trimming, should be considered prior to
• During construction, it causes less environmental soil nail installation.
impact than cutting back and retaining wall con-
struction as no major earthworks and tree felling
are needed. 2.2 Mechanisms of soil-nailed system
• There could be time and cost savings compared
to conventional techniques of cutting back and 2.2.1 Load transfer mechanism
retaining wall construction which usually involve The soil nailing technique improves the stability of
substantial earthworks and temporary works. slopes, retaining walls and excavations principally
• It is less sensitive to undetected adverse geologi- through the mobilisation of tension in the soil nails.
cal features, and thus more robust and reliable than The tensile forces are developed in the soil nails pri-
unsupported cuts. In addition, it renders higher sys- marily through the frictional interaction between the
tem redundancy than unsupported cuts or anchored soil nails and the ground as well as the reactions pro-
slopes due to the presence of a large number of soil vided by soil-nail heads and the facing (Fig. 2). The
nails. tensile forces in the soil nails reinforce the ground by
• Its failure mode is likely to be ductile, thus providing directly supporting some of the applied shear loadings
warning signs before failure. and by increasing the normal stresses in the soil on
the potential failure surface, thereby allowing higher
Like every other stabilization technique, soil nailing shearing resistance to be mobilised. Soil-nail heads
has its limitations: and the facing also provide a confinement effect by
• The presence of utilities, underground structures, or limiting the ground deformation close to normal to
other buried obstructions poses restrictions to the the slope surface. As a result, the mean effective stress
length and layout of soil nails. and the shearing resistance of the soil behind the soil-
• The zone occupied by soil nails is sterilised and this nail heads will increase (Fig. 3). Soil-nail heads and
site poses constraint to future development. the facing also help preventing local failures near the
• Permission has to be obtained from the owners of surface of a slope and promote an integral action of
the adjacent land for the installation of soil nails
beyond the lot boundary. This places restrictions on
the layout of soil nails.
• The presence of high groundwater levels may lead to
construction difficulties in hole drilling and grout-
ing, and instability problem of slope surface in the
case of soil-nailed excavations.
• The effectiveness of soil nails may be compromised
at sites with past large landslides involving deep-
seated failure due to disturbance of the ground.
• The presence of permeable ground, such as ground
with many cobbles, boulders, highly fractured
rocks, open joints, or voids, presents construction
difficulties due to potential grout leakage problems.
• The presence of ground with a high content of fines
may lead to problems of creeping to soil nails. Figure 2. Load transfer mechanism of soil nailed structure.

87
 ground, the nail inclination, and the tensile strength,
10kPa shear strength and bending capacity of the soil nail.
Generally speaking, the axial strain will mobilize ten-
30kPa
sile or compressive forces, and the lateral strain will
nail head 50kPa
mobilize shear force and bending moment in the soil
position nail. Due to relatively slender dimensions of soil nails,
(400mm) 70kPa the reinforcing actions from shear and bending are lim-
ited by the small flexural strength, and they are usually
90kPa negligible (FHWA 1998). The effect of inclination and
110kPa
bending stiffness of soil nail are discussed further in
the subsequent sections of this paper.
130kPa Compressive and shear strains are developed in the
150kPa
soil behind a soil-nail head in response to the ground
deformation in the active zone (Fig. 2). If the resul-
170kPa
tant strain is close to the direction perpendicular to
the base of soil-nail head, the head-ground interaction
(a) Contours of mean effective stress in soil nailed slope will be predominantly in the form of a bearing mech-
10kPa
anism. However, if the resultant strain is in a direction
that deviates significantly from the normal to the base
30kPa of the soil-nail head, the head-ground interaction will
be a combination of bearing and sliding mechanisms.
50kPa These interaction, particularly the bearing mechanism,
gives rise to tensile loads at the heads of soil nails. The
70kPa tensile loads at the soil-nail heads are taken up by the
soil-nail reinforcement. The interaction increases as
90kPa
the size of the soil-nail heads or the coverage of fac-
110kPa ing increases, resulting in larger tensile loads. Further
discussion on the effect of soil-nail heads is given in
130kPa
Section 3.5 below.
150kPa The mobilization of pullout resistance along a soil
170kPa
nail in the passive zone depends on many factors.
Theoretically, the bond strength between the soil nail
and the ground depends on the contact stress and the
(b) Contours of mean effective stress in unreinforced slope interface coefficient of friction. Where a soil nail is
installed by the drill-and-grout method, the process of
Figure 3. Contour of mean effective stress in (a) soil-nailed drilling reduces significantly the radial stress at the
slope, and (b) unreinforced slope. circumference of the drillhole. The drillhole remains
stable by soil arching. Subsequent grouting will restore
a certain level of the radial stress in the soil around the
the reinforced ground mass through redistribution of drillhole. The contact pressure at the drillhole face is
forces among soil nails. The resistance against pullout generally small compared to the overburden pressure
failure of the soil nails is provided by the part of soil except where pressure grouting is adopted. This seems
nails that is embedded into the passive zone. to imply small bond strength at the ground/grout inter-
When there is a tendency of ground movement in face. In reality, as the drillhole face may be irregular
the active zone, the soil nail will experience both axial and rough, the mechanical interlocking between the
and lateral strains through two fundamental mecha- cement grout and the ground also contributes a signif-
nisms of nail-ground interaction. They are: (i) the icant portion of the bond strength. Upon pulling of the
nail-ground friction that leads to the development of soil nail, shearing may occur within the ground mass
axial strains in the soil nails, and (ii) the soil bearing in a finite zone surrounding the soil nail. If the soil
on the soil nails and the nail-ground friction on the is dilative, the effect of restrained soil dilatancy will
sides of soil nails that lead to the development of lat- come into play. The effect of this can be significant and
eral stains in the soil nails. In these two mechanisms, can lead to high soil-nail friction (Pun & Shiu 2007).
the interactions between the ground and the soil nails Soil nails are considered to tie the active zone to
are complex and the forces developed in the soil nails the passive zone. It should be noted that the two-zone
are influenced by many factors such as the bearing concept is only an idealisation for design purpose. In
capacity of the ground to resist reaction force from reality there is a complex shearing zone subject to
the soil nail, the relative stiffness of the soil nail and shear distortion, unless the failure is dictated by joint

88
settings where the failure surface is distinct. The effect
of the shear zone on the mobilization of forces in soil
nails is discussed in Section 3.4 below.

2.2.2 Effect of nail inclination

Unlike the reinforcements in reinforced fill structures,
which are placed in horizontal direction, soil nails can
be installed in the ground at various inclinations. In
cramped sites, soil nails are sometimes installed at
large inclinations. Different nail inclinations may pro-
duce different effects on the behaviour of soil-nailed
structures. (a) Principal Strains in Unreinforced Sand after Peak
In this paper, nail inclination, α, is the angle of a soil
nail made with the horizontal; and nail orientation, θ,
is the angle between a soil nail and the normal to the
shearing surface. The typical relationship between α
and θ is presented in Figure 4.
Jewell (1980) investigated the fundamental behaviour
of reinforced soil by carrying out a series of direct
shear box tests on sand samples reinforced with bars
and grid reinforcements. One of the significant find-
ings of his work was that the shear strength of the
reinforced soil is dependent on the orientations of the
reinforcements. Jewell’s investigation shows that rein-
forcement significantly modifies the state of stress and (b) Principal Strains in Sand Reinforced by a Grid at an
strain in soil, and that by varying the orientation of the Orientation = +30º (legend: double arrow represents principal
tensile strain)
reinforcement, the reinforcement can either increase
or decrease the shear strength of the soil. Figure 5
compares the pattern of strain in soil between the Figure 5. Incremental strains at peak shearing resistance in
unreinforced and reinforced sand (after Jewell 1980).
unreinforced and reinforced tests. The presence of the
reinforcement causes a significant reorientation of the
principal axes of strain increment of the soil. The soil
strains close to the reinforcement are small because
the reinforcement inhibits the formation of the fail-
ure plane. When the reinforcement is orientated in the
same direction of the tensile strain increment of the
soil, tensile forces are induced in the reinforcement
through the friction between the soil and the reinforce-
ment. Likewise, compressive forces are induced in the
reinforcement if the reinforcement is placed close to
the compressive strain increment of the soil.
Figure 6 shows the orientations of the reinforce-
ment in which compressive or tensile strain increments

Figure 6. Increase in shear resistance for reinforcement

placed at different orientation (after Jewell & Wroth 1987).

are experienced. The shear strength of the soil starts

to increase when the reinforcement is placed in the
direction of tensile strain increment, and it reaches a
maximum when the orientation of the reinforcement
is close to the direction of the principal tensile strain
increment. When the reinforcing elements are oriented
Figure 4. Relationship between nail inclination and orien- in a direction of a compressive strain increment, there
tation. is a decrease in shear strength of the reinforced soil.

89
Figure 7. Geometry and material parameters of model slope.

This shows that in order to optimise strength improve- ratio of the actual soil shear strength to the minimum
ment of soil, the reinforcement should be placed in shear strength required for equilibrium. As pointed
the directions of principal tensile strain in the soil. out by Duncan (1996), FoS can also be defined as
When the reinforcement deviates from its optimum ‘‘the factor by which the shear strength of the soil
orientation, strength improvements decrease. would have to be divided to bring the slope into a
Results of laboratory investigation of the behaviour state of barely equilibrium’’. FoS can therefore be
of soil reinforced with bars were also reported by Mar- determined simply by reducing the soil shear strength
chal (1986), Hayashi et al (1988), and Palmeira & until failure occurs. This strength reduction approach
Milligan (1989). They showed that the orientation is often used to compute FoS using finite element or
of reinforcements played a significant role in the finite difference programs (Dawson et al 1999, Krahn
improvement of the shear strength of the reinforced 2003).
soil mass. A negative orientation can lead to the devel- Figure 8 shows that for soil nails with a small incli-
opment of compressive force in the reinforcement nation of 20◦ , tensile forces are developed in all the
and consequently loss of shear strength of the soil. soil nails. On the other hand, when the soil nails are
There exists an optimum reinforcement orientation in inclined steeply at an inclination of 55◦ , compressive
terms of strengthening of the soil. These findings are forces are developed in the top four rows of soil nails
consistent with that of Jewell (1980). whereas tensile forces are mobilized only in the bot-
The effects of nail inclination on the safety margin tom three rows of soil nails. Tensile forces in the soil
of a slope was studied by Shiu & Chang (2005) in nails can improve the slope stability whereas compres-
Hong Kong by means of numerical analysis using sive forces can have opposite effect. Increases in FoS
the two-dimensional finite difference code, Fast (FoS) due to the soil nails were calculated for differ-
Lagrangian Analysis of Continua (FLAC) (Itasca ent nail inclinations. Figure 9 shows the relationship
1996). A simulated slope of 20 m in height, stand- between the calculated FoS and nail inclinations (α)
ing at an angle of 55◦ , and with an up-slope of 10◦ in for the model slope. The  FoS is close to 1 with lit-
gradient was adopted for the analysis. Figure 7 shows tle variations for the range of α between 0◦ and 20◦ .
the geometry of the slope and the material parame- The  FoS decreases quickly as α increases beyond
ters used in the numerical analysis. Each soil nail was 20◦ , reflecting that the reinforcing effects of the soil
20 m long with a 40 mm diameter steel bar in a 100 mm nails reduce rapidly with increasing nail inclinations.
grouted hole. A Mohr Coulomb model was assumed At α = 55◦ , the value of  FoS is almost zero.
for the soil. A cable element was used to represent the These studies show that the nail inclination can sig-
soil nail as the bending stiffness of the soil nail was not nificantly affect the reinforcing action of the soil nails.
considered. Developments of the tensile forces in the Increase in nail inclination would decrease the effi-
soil nails were governed either by the tensile strength ciency of the reinforcing action of the soil nails. For
of the nail or the peak shear strength at the soil-grout steeply inclined soil nails, axial compressive forces
interface. may be mobilized in the soil nails. The compressive
Slope stability analysis was first carried out on the forces would reduce the stability of the soil-nailed
unreinforced slope. From the results of the analy- structure.
sis, the unreinforced slope has a minimum factor of
safety (FoS) close to 1.0 for the initial soil strength 2.2.3 Effect of bending stiffness of soil nail
parameters of c = 10 kPa, and φ  = 43◦ . In slope Steel reinforcements can sustain shear forces and
engineering, the FoS is conventionally defined as the bending moments, and thus this ability of steel soil

90
 Figure 9. Variation of increase in factor of safety against
inclination of soil nails.

laboratory and theoretical study reported by Pedley

(1990) and Jewell & Pedley (1990, 1992).
In the study of Pedley (1990), a series of direct
shear tests were carried out in a large-scale direct shear
apparatus (1 m × 1 m × 1 m). Three different types
of circular elements were tested. They were solid steel
bars (16 to 25.4 mm in diameter), metal tubes (15.88 to
25.4 mm in external diameter and 13.19 to 22.36 mm
in internal diameter) and grouted bar (50.8 mm in
diameter with steel bar diameters 6.71 to 16 mm).
Figure 10 shows the distributions of: (a) the reinforce-
ment bending moment (M) normalised by the plastic
moment capacity (Mp ); (b) the reinforcement shear
force (Ps ) normalised by the full plastic axial capacity
(Tp ); and (c) the lateral stress on the reinforcement
(σ l ) normalised by the limiting soil bearing stress
(σ b ). These distributions were the stress conditions
of the reinforcement at a shear displacement of soil of
60 mm. The test results show that even when the mea-
sured bending moments were close to the fully plastic
moment (M/Mp = 1) in all the tests, the maximum
shear force (Ps ) in the reinforcement was less than 6%
of the plastic axial capacity (Pp ).
Figure 8. Axial force distribution in soil nails in model Pedley (1990) confirmed the laboratory test results
slope. by theoretical analysis. He derived elastic and plas-
tic models for determining the maximum shear force
nails may enhance the shear strength of soil. The mobilised in the reinforcement bar. These models were
development of shear force in soil nails involves a also reported in Jewell & Pedley (1990, 1992). The
mechanism which is dependent on the relative stiff- elastic analysis simply defines the stress condition
ness of the soil nail and the ground, the soil bearing before the reinforcement reaches plasticity; it does not
strength, the orientation and shear deformation of represent the failure condition. As such, only the plas-
reinforcement, and the thickness of shear zone. tic analysis is discussed here. The plastic analysis will
The effect of the bending stiffness of a soil nail always give a larger reinforcement shear force than the
on nail forces and displacements has been investi- elastic analysis.
gated by many researchers, e.g. Schlosser (1982), The limiting plastic envelope for a bar of rectangular
Marchal (1986), Gigan & Delmas (1987), Pluemelle cross-section is given by (Calladine 2000):
et al (1990), Pedley (1990), Jewell & Pedley (1990,
1992), Bridle & Davies (1997), Davies & Le Masurier
2
M T
(1997), Smith & Su (1997) and Tan et al (2000). The + =1 (1)
most notable and comprehensive investigation was the Mp TP

91
 Figure 11. Plastic analysis of soil nail interaction (after
Jewell & Pedley 1992).

bar, σy is the yield stress of the bar and σb is the limiting
bearing stress between soil and reinforcement.
The limiting bearing pressure (σb ’) between the soil
and the reinforcement required to achieve the plastic
equilibrium is:
Figure 10. Profiles of bending moment, shear force and
lateral stress in reinforcement (after Pedley 1990).
8Mmax
(σb )max = (4)
ls2 D
Equation 1 is slightly conservative for circular bar.
Since no simple relationship can be derived for circular The theoretical plastic limiting maximum shear
bar, this equation has been adopted by Jewell & Pedley force Ps that can be generated in an ungrouted round
(1990). bar that also supports axial force, P, is:
The relation between maximum shear force Ps and
the maximum moment Mmax depends on the mag-   2 
nitude and distribution of the lateral loading on the Ps 8 T
reinforcement bar. For plastic analysis for a soil nail = 1− (5)
Tp 3π(ls /D) Tp
under the lateral loading shown in Figure 11, the
equations for this distribution of lateral loading are:
Details of explanations and derivation of Equations
4Mmax (1) to (5) can be found in Pedley (1990).
(P s ) = (2) Jewell & Pedley (1992) computed and presented
ls
envelopes of limiting combinations of shear force Ps


ls 4σy and axial force T for a grouted reinforcement bar
= (3) of 25 mm diameter with typical soil parameters and
D 3σb showed that the magnitude of the limiting shear force
in the reinforcement was only a small proportion of
where ls is the distance between the points of max- the axial force capacity. This was the case even when
imum moment on either side of the potential shear the reinforcement was oriented so as to mobilize the
surface, (Fig. 11), D is the diameter of reinforcement maximum shear force.

92
 Pedley (1990) also back-analysed an instrumented to play a role. However, soil nails are not effective in
6 m high soil-nailed wall, which was loaded to failure. providing dowel action. For that purpose, other types
He found that the highest contribution of reinforce- of structural element should be considered, e.g. large
ment shear force to soil strength improvement was less diameter piles.
than 3% to that due to reinforcement axial force. This The above studies show that the contribution of
is in agreement with the theoretical study result that shear force and bending stress of soil nails on enhanc-
only very small amount of shear force can be mobilised ing the shear strength of the soil mass is very small.
in a soil nail. Large soil displacements are required to mobilise shear
The effect of bending stiffness of soil nails on nail and bending forces in the soil nails. When failure con-
forces and displacements and the safety margin of soil- ditions are approached, the contribution of shear and
nailed slopes has also been studied by Shiu & Chang bending action may be more significant but is still
(2005) in Hong Kong by means of numerical simula- small. For these reasons, the soil nail design practice
tions using the finite element code PLAXIS. The slope in most countries, such as the USA (FHWA 1998),
model is the same as that shown in Figure 7. The FoS of the UK (Department of Transport 1994), Japan (JHPC
the soil-nailed slope, the tensile forces, shear stresses 1998) and Germany (Gässler 1997), ignores any ben-
and bending moments developed in the soil nails at eficial effects from the mobilisation of shear force
different inclinations were computed. or bending stress in the soil nails. An exception to
The maximum axial force developed in a soil nail is this is the French design approach (Clouterre 1991) in
Tmax . Figure 12 shows the sum of the maximum tensile which the contributions of shear and bending of the
forces mobilised in all the soil nails (Tmax ) at limit nails are considered. Clouterre (1991) emphases that
equilibrium condition of the slope model. The maxi- shear forces are mobilised in soil nails only when the
mum shear force in a soil nail at the location where the structures are near failure.
shear plane intersects the soil nail is Psmax . The sum of Although the bending stiffness of soil nail has
the maximum shear forces (Psmax ) mobilized in the little contribution to the shear resistance of a soil-
soil nails at limit equilibrium condition of the model nailed system, the beneficial effect of shear ductility
are also plotted in Figure 12. The value of Psmax of soil nail reinforcement should not be ignored. For
rises steadily with increasing nail inclination. The rise example, steel has large shear ductility. As a result
is small, from 31 kN/m at α = 10◦ , to 76 kN/m at of the mobilization of shear and bending ductility at
α = 55◦ . In contrast, the value of Tmax decreases large deformations, a soil-nailed system comprising
rapidly with increasing nail inclination. For small nail steel reinforcement tends to exhibit ductile rather than
inclinations, Tmax is much larger than Psmax . Com- brittle failure.
paring Figures 9 and 12, it can be seen that both FoS
and Tmax decrease with increasing nail inclinations. 2.2.4 Effect of thickness of shear zone
This similarity illustrates that FoS is strongly influ- The reported laboratory investigation and numerical
enced by the nail axial force. The FoS is not sensitive studies generally considered sliding along a shear
to the mobilized shear resistances in the soil nails. plane. To investigate the influence of the thickness of
The modeling results show that small shear forces are shear zone on the mobilisaion of shear force and bend-
mobilized in soil nails and they have little effect on ing moment in soil nails, a study involving numerical
the factor of safety of the slope, except perhaps at simulations of large shear box tests was conducted
very steep nail inclination where dowel action starts by the Geotechnical Engineering Office (GEO) in
Hong Kong. The two-dimensional finite element code
PLAXIS was used. Three cases were considered, rang-
ing from a well-defined shear plane to a wide shear
zone (Fig. 13).
The model shear box was 3 m deep and 6 m
long. A reinforcement was placed in the middle of
the box across the vertical slip surface. The box was
assumed filled with homogeneous sand. The model
and the assumed parameters are shown in Figure 14.
A Mohr-Coulomb model was assumed for the soil.
In the numerical simulation, the shear box was ini-
tially restrained to move. An overburden pressure of
80 kPa was applied on the top of the box to model a 5 m
high earth pressure. In the case of sliding along a shear
Figure 12. Variation of total maximum tensile force (Tmax ) plane, an imposed downward uniform displacement δ
and total maximum shear force (Ps max ) with nail inclina- of 5 mm, 10 mm, 25 mm and 50 mm was applied
tion (α). in sequence at the right-half part of the shear box.

93
 Figure 15. Shear force along steel bar in shear box model.

Figure 13. Cases considered in numeral study of effect of

shear zone.
Figure 16. Bending moment along steel bar in shear box
model.

25 mm and 50 mm was again applied in sequence in

the downward direction beyond the shear zone along
the top and bottom of the right-half part of the box.
Two values of z, 100 mm and 200 mm respectively,
were considered. In the case of wide shearing zone,
the shear zone was extended to cover the entire right-
hand part of the shear box.
The results of the numerical simulations are shown
in Figures 15 and 16. The mobilisation of shear force
and bending moment in soil nails is affected by the
thickness of the failure shear zone. The narrower the
shear zone, the higher is the shear force and bending
moment in the soil nails.

2.2.5 Effect of soil nail head and facing

Our understanding of the magnitude and distribution
of loadings developed at soil-nail head is not as good
as our knowledge of the development of tensile forces
Figure 14. Model for numeral study of effect of shear zone.
in soil nails (FHWA 1998). This is because of the lack
of good quality field monitoring data. The available
data from instrumented soil nails are generally dif-
In the case of a narrow shearing zone, a linearly ficult to interpret in the vicinity of soil-nail heads,
varying downward displacement was applied across where bending effects of soil nail tend to be more sig-
the shear zone of width z during the loading phases. nificant arising from the weight of the soil-nail head
An imposed uniform displacement of 5 mm, 10 mm, (Thompson & Miller 1990). There has been little field

94
monitoring data obtained using load cells at soil-nail base. Result of the tests is shown in Figure 18. It indi-
heads probably because of the difficulties in placing cates that the reinforced slope with facing can sustain
load cells between soil-nail heads and soil (Stocker & a higher load than the reinforced slope with no fac-
Reidinger 1990). ing, and a much higher load than the unreinforced
Despite the lack of good quality field monitor-
ing data, a number of studies including model tests
(e.g. Muramatsu et al 1992, Tei et al 1998), full-
scale field tests (e.g. Gässler & Gudehus 1981,
Plumelle & Schlosser 1990, Gutierrez & Tatsuoka
1988, Muramatsu et al 1992) and numerical simula-
tions (e.g. Ehrlich et al 1996, Babu et al 2002) have
been carried out. The results of these studies provide
useful insight into the role and behaviour of soil-nail
heads. Many of the studies are related to soil-nailed
retaining walls where soil-nail heads are integrated
into a concrete facing.
Gutierrez & Tatsuoka (1988) reported loading tests
performed on three model sand slopes: (i) unrein-
forced slope, (ii) slope reinforced with metal strips
but without a facing, and (iii) slope reinforced with
metal strips and with a facing (Fig. 17). The slopes
were loaded at the crest by a footing with a smooth

Figure 19. Shear stain contour in model tests by Gutierrez &

Tatsuoka (1988).
Figure 17. Model slope reinforced with metal strips and
with a facing tested by Gutierrez & Tatsuoka (1988).

Figure 20. Relationship between factor of safety and soil-

Figure 18. Results of tests of Gutierrez & Tatsuoka (1988). nail head size.

95
Figure 21. Variation of axial nail forces for (a) 800 mm soil head and (b) no nail head.

slope. Figure 19 shows the failure planes of the slopes in nails with heads of 800 mm wide. For the soil nails
when loaded at the crest. Deep and shallow failure without soil-nail heads, no tensile force is developed at
planes were observed in the unreinforced slope. For the front end of the soil-nail whereas for the soil nails
the reinforced slope with no facing, failure took place with soil-nail heads, large tensile forces are mobilised
close to the slope face. This was because the reinforce- in the soil nails at the connections to the soil-nail heads.
ment alone was not effective at retaining the active The maximum tensile forces mobilized along the soil
zone. For the reinforced slope with facing, the failure nails are much larger in the latter case.
was observed at a greater depth. The maximum ten- A series of centrifuge tests has also been con-
sile force generated in the reinforcement was larger ducted in the Geotechnical Centrifuge Facility of the
than that in the reinforced slope with no facing. Fur- Hong Kong University of Science and Technology
thermore, substantial tensile force was induced in the to investigate the reinforcing effect of soil nails and
reinforcement at the connection to the facing. The full- soil-nail heads (Ng et al 2007). Figure 22 shows an
scale field tests by Plumelle & Schlosser (1990) and instrumented model used in one of the nailed slope
Muramatsu et al (1992) also showed that the presence centrifuge tests. The test results support the results
of a facing could enhance the stability of a reinforced of the numerical simulations that soil-nail heads can
slope and helped prevent shallow failures. substantially improve the stability of a soil-nailed
Soil-nail heads used in slope stabilization works slope.
in Hong Kong are usually in the form of isolated Results of the above model tests, field measure-
reinforced concrete pads. To investigate the effect ments and numerical simulations highlight the impor-
of soil-nail heads on stability of nailed slopes, a tance of soil-nail heads in the soil nailing applications.
series of numerical simulations using the two dimen- They show that soil-nail heads, whether in the form
sional finite element code FLAC was conducted of individual concrete pads or as part of concrete
(Shiu & Chang 2004). Figure 7 shows the slope model. facing, greatly enhance the stability of a soil-nailed
Strength reduction technique (Dawson et al 1999) was slope.
employed to compute the factor of safety (FoS) of
the model slope. In the simulations, soil-nail heads
2.3 Modelling and design
of different sizes were considered. The slope without
any soil nails (i.e. unreinforced) had a minimum FoS 2.3.1 Design approach and standard
close to 1. Based on the FLAC analysis, Figure 20 A soil-nailed structure is required to fulfil fundamental
shows the relationship between the calculated FoS of requirements of stability, serviceability and durabil-
the model slope and soil-nail head sizes. The FoS ity during construction and throughout its design life.
increases from 1 for the unreinforced slope to 1.2 for Other issues such as cost and environmental impact
the soil-nailed slope with no soil-nail heads. Substan- are also important design considerations.
tial increases in the FoS are observed when soil-nail The design for stability generally entails the set-
heads of sizes ranging from 400 mm wide to 800 mm ting up of ground and design model, consideration of
wide are provided. The trend of increase in FoS levels potential failure mechanisms, stability analyses, deter-
off for soil-nail head sizes larger than 800 mm wide. It mination of soil-nail design capacity, soil-nail head
shows that soil-nail heads can significant enhance the and facing design, and detailing.
stability of a soil-nailed slope. The failure mechanisms of nailed structures can
Figure 21 compares the axial tensile forces devel- broadly be classified as external failure and internal
oped in soil nails without soil-nail heads with those failure. External failure refers to the development of

96
 LV DT1 LV DT2
PPT
Unit:mm codes. Driven by the Eurocode, all the European coun-
Upstream drainage 50 0
Model box tries under the European Union use the partial factors
280
board 100
approach in soil nail design. In the USA, it appears
PPT13 PPT14
Nail F
20 ¡ã
Downstream
that it is at a transition stage of changing over from
PPT12
PPT10 Nail E PPT11 50 0
drainage board
Downstream
global safety factor approach to partial safety factors
CoarseCDG Nail D
Upstream temporary
reservior
block PPT9 temporary reservior approach. The American design code permits the use
Outlet hole
Inlet hole PPT7 Nail C PPT8
PPT6 65¡ã
50 of either load and resistant factor design approach
Nail B
PPT5
PPT1 PPT2 40 0 PPT4 PPT3
100 (which is similar to partial factors approach) or ser-
1130 vice load design approach. In Hong Kong, the soil nail
design approach is essentially a combination of global
(a) Set-up of a nailed-slope model in centrifuge test.
safety factor approach (permissible stress design) and
partial safety factor approach. Table 1 summarises
the design approaches recommended by the different
design codes.
Calculation methods involving trial wedges (single-
wedge or double-wedge) and limit equilibrium meth-
ods (LEM) of slices on circular, spirial, or other
non-circular slip surfaces are commonly used. While
these methods are good enough for design purpose,
none of them can account for the actual behaviour of
a soil-nailed structure, which is a strain compatibility
problem. It is possible to define a wide variety of nail
length patterns that satisfy stability requirements but
that may not satisfy serviceability requirements.
It is essential to have a good understanding of the
(b) Front view of the slope model in centrifuge test principles behind the calculation methods so that the
appropriate method is used and the results are inter-
Figure 22. Centrifuge test to study soil-nailed slope. preted correctly. For instance, the factor of safety of a
soil-nailed slope computed using the simplified Janbu
method is in sensitive to the location of the applied soil
potential failure surfaces essentially outside the soil- nail force. This is an inherent limitation of the method,
nailed ground mass. The failure can be in the form and it may give rise to an over- or under-estimation of
of sliding, rotation, bearing, or other forms of loss of the true safety margin. In light of this, GEO (2008)
overall stability (see Fig. 23a). Internal failure refers recommends that only stability analysis methods that
to failures within the soil-nailed ground mass. Internal satisfies both moment and force equilibrium should be
failures can occur in the active zone, passive zone, or used in soil nail design.
in both of the two zones of a soil-nailed system. In Hong Kong, more than 3,000 slopes and retain-
In the active zone, internal failure modes could be: ing walls have been stabilized using soil nails. The
vast amount of soil nail designs had allowed the
• failure of the ground mass, i.e., the ground disinte-
development of a prescriptive design approach for
grates and ‘flows’ around the soil nails and soil-nail
the design of soil-nailed soil cut slopes and retain-
heads
ing walls. Prescriptive measures are pre-determined,
• bearing failure underneath soil-nail heads
experience-based and suitably conservative modules
• structural failure of the soil nail under combined
of works prescribed to a slope or retaining wall to
actions of tension, shear and bending
improve its stability or reduce the risk of failure, with-
• structural failure of the soil-nail head or facing, i.e.,
out detailed ground investigation and design analyses.
bending or punching shear failure, or failure at head-
Using prescriptive measures has the technical benefits
reinforcement or facing-reinforcement connection
of enhancing safety and reducing the risk of failure, by
• surface failure between soil-nail heads, i.e.,
incorporating simple, standardised and suitably con-
washout, erosion, or local sliding failure
servative items of works to deal with uncertainties in
In the passive zone, the failure mode is mainly pull- design that are difficult to quantify. There would also
out failure of soil nail along soil-grout interface or be savings on time and human resources, by eliminat-
reinforcement-grout interface. ing detailed ground investigation and design analyses.
The various internal failure modes are illustrated in The prescriptive soil nail design guidelines in Hong
Figure 23b. Kong comprise standard soil nail layouts and a set of
The approach of limit state design incorporating qualifying criteria for the application of the prescrip-
partial factors is adopted in many soil nail design tive measures to ensure that the prescriptive design

97
Figure 23. Principal modes of failure of soil-nailed system.

approach was applied safely within the bound of past reviews, laboratory investigations (e.g. Law et al 1998,
experience. The guidelines were developed based on Ng & Chiu 2003) and numerical modelling (e.g. Cheuk
the findings of review of several hundreds of soil nail 2001), specific guidelines on the use of soil nails in
designs (Pang & Wong 1997, Pun et al 2000, Lui & fill were developed by HKIE (2003). The following
Shiu 2004). Soil nail designs using the prescriptive are salient points of the design recommendations:
approach have been successfully applied to many soil
cut slopes and retaining walls since the promulgation • Large-strain steady-state undrained shear strength
of the design guidelines in Wong et al (1999) and Lui & should be assumed for loose fill in the design.
Shiu (2005). • Global stability should be provided for by bonding
Soil nailing is also a feasible option for the stabiliza- soil nails into a competent stratum.
tion of loose fill slopes. However, as loose granular fill • Local surface stability should be enhanced by the
material exhibits contractive behaviour upon shearing, provision of a concrete grid structure covering not
there is concern that the loose fill may lose strength at less than 50% of the slope surface and connecting
such a rate that the forces mobilised in the soil nails soil nail heads.
will not be able to compensate for the loss of shear • Soil nails should be closely spaced horizontally and
strength in the fill. Following comprehensive literature vertically.

98
Table 1. Summary of approach and method of soil-nailed slopes design.

Place Document Design approach Common method of analysis

China Geotechnical Engineering Partial factors approach Limit equilibrium method using
Handbook 1994 single wedge mechanism
Europe Eurocode 7 : Geotechnical Limit state approach incorporating No specific method defined for
Design, 1995 partial safety factors soil nail design
France Soil Nailing Recommendations Limit state approach incorporating Limit equilibrium method based
Clouterre 1991 partial safety factors on Bishop and two-part wedge
mechanism
Hong Guide to Soil Nail Design and Global safety factor approach Limit equilibrium method of
Kong Construction (Geoguide 7) combined with separate factors slices on non-circular slip
of safety on tensile strength
of steel and pull-out resistance
Japan Japanese Design Guide: Design Global safety factor approach Calculation models based on
and Construction Guidelines combined with safety factors for conventional circular or linear
for Reinforced Cut Slopes tensile strength of steel and pull-out slip surface analyses
resistance
South Lateral Support in Surface Global factor of safety combined with Limit equilibrium method using
Africa Excavations: Code of Practice separate factors of safety on strength trial sliding wedges
1989 of steel and pull-out resistance
UK Design Methods for the Limit state approach incorporating Limit equilibrium method using
Reinforcement of Highway partial safety factors two-part wedge mechanism
Slopes by Reinforced Soil and
Soil (HA 68/94)
UK BS 8006: 1995 Code of Practice Limit state approach incorporating Limit equilibrium method with
for Strengthened/Reinforced partial safety factors reference two-part wedge,
Soils and Other Fills circular slip and log-spiral
methods
USA Manual for Design & Limit state approach incorporating Design models based on Load
Construction Monitoring of Soil partial safety factors or Service and Resistance Factor Design
Nail Walls Load Design and Service Load Design
Limit equilibrium based on
two-part wedge and slip circle
method is adopted

• The grid structure should be designed to withstand zone is to be checked. The UK practice also requires
bending moments and shear forces generated by the the checking of the bearing capacity failure in soil.
loose fill it is retaining; it should be adequately The French, Japanese and German methods use
founded on a competent stratum. empirical earth pressures which are related either to
• The potential of leakage from water-carrying ser- the maximum tension developed in the soil nail (Tmax )
vices should be duly considered. or Coulomb earth pressure. The U.S. and Japanese
practice consider directly the strength of soil-nail head
2.3.2 Soil-nail head design when determining the magnitude and distribution of
Guidelines on the design of soil-nail head are available nail forces along the length of the soil nails. If the
in the design codes of the UK (Department of Trans- beneficial effect of the soil-nail head is not considered,
port 1994), France (Clouterre 1991), USA (FHWA the pull-out resistance of the soil nail at the active zone
1998), Japan (JHPC 1998) and Germany (Stocker & would be significantly reduced. This may lead to more
Riedinger 1990). All these documents recognize the number of soil nails being required.
soil-nail head or facing as a significant component of In all the design methods, the size, thickness and
the overall soil nail system, and they provide specific reinforcement details of soil-nail heads are determined
recommendations for design pressures. They also rec- on the basis of the earth pressure acting on the soil-nail
ognize that the magnitudes of pressures induced in the heads. Two main design aspects are considered: the
soil-nail heads are controlled by many factors such as bearing capacity of the soil beneath the soil-nail head
the density and length of the nails and the stiffness and the structural strength of the soil-nail head itself.
of the soil-nail head. Both the UK and the Japanese Many of the design methods (such as those used in
practice require that the pull-out failure at the active France and U.S.) were developed mainly for soil-nailed

99
walls, where the soil-nail heads form part of the con- Knowing the shear strength parameters of the soil, the
crete facing. In these cases, bearing failure of the soil steel bar diameter and the slope angle, a designer can
beneath soil-nail heads or facing is unlikely to occur determine the size of soil-nail head from these plots.
and as such little guidance has been provided in respect A design table has been derived from the plots for
of soil bearing failure. A method on the design against different combinations of slope angle and angle of
bearing failure of the soil behind isolated nail heads is shearing resistance of soil (Table 2).
given in the UK guidance document HA68/94.
For the development of soil-nail head design guide- 2.3.3 Pullout resistance
lines in Hong Kong, a series of numerical analysis Pull-out capacity is a key parameter for the design
was carried out. In the study, FLAC analysis was per- of soil nails. At present, methods for estimating
formed to examine the bearing failure beneath square pullout capacity are not unified as reflected by the
soil-nail heads. A small slope model of 5 m in height many approaches used in different technical stan-
was used and various slope angles were considered dards and codes of practice, such as effective stress
(Fig. 24). In the analysis, the soil-nail head was pushed method (GEO 2008), empirical correlation with SPT
into the ground by a nail force to simulate the situ- N values (JHPC 1998), correlation with pressureme-
ation of soil moving out from a slope and pressing ter tests (Clouterre 1991), and correlation with soil
against the soil-nail head. The nail forces used are types (FHWA 2003). The merits and limitations of the
determined from the allowable tensile strength of steel various methods are summarized in Table 3.
bars. Figures 25 and 26 show the shear strains and the The effective stress method is adopted in Hong
displacement vectors respectively at the point of bear- Kong. The allowable pullout resistance provided by
ing failure for a 600 mm x 600 mm soil-nail head on the soil-grout bond strength in the passive zone, TP , is
a 45◦ slope. Typical results of the analyses in terms of given by (Schlosser & Guilloux 1981):
c’-φ’ envelope for limit equilibrium (i.e. when bear-
ing failure occurs) are plotted in Figure 27. In this plot,
c Pc Lc + 2Dσv μ∗ Lc
the soil-nail head loads are expressed as diameters of TP = (6)
steel bars. A number of the plots have been devel- FP
oped for different slope angles and soil-nail head sizes.

Figure 24. Slope model for bearing capacity analysis.

Figure 26. Typical displacement vector plot.

14
40 mm diameter bar
12

10
32 mm diameter bar
8

6
25 mm diameter bar
4

0
31 32 33 34 35 36 37 38 39 40 41 42 43 44

Figure 27. Shear strength required for 600 mm × 600 mm

nail heads on a 45◦ slope to mobilise allowable tensile
Figure 25. Typical shear strain plot. strength of nails of specified diameters.

100
Table 2. Recommended sizes of isolated soil-nail heads in Geoguide 7 (GEO, 2008).

45◦ ≤ Slope angle < 55◦ 55◦ ≤ Slope angle < 65◦ Slope angle ≥ 65◦

Soil shear strength Diameter of Diameter of Diameter of

parameter near the soil-nail soil-nail soil-nail
slope surface reinforcement (mm) reinforcement (mm) reinforcement (mm)

ϕ’ c’ (kPa) 25 32 40 25 32 40 25 32 40
2 800 800 800 600 600 800 600 600 800
4 600 800 800 600 600 800 600 600 800
34◦ 6 600 800 800 400 600 800 400 600 600
8 600 600 800 400 600 800 400 600 600
10 400 600 800 400 600 600 400 600 600
2 600 800 800 600 600 800 600 600 800
4 600 800 800 400 600 800 400 600 800
34◦ 6 600 600 800 400 600 800 400 600 600
8 400 600 800 400 600 600 400 600 600
10 400 600 800 400 600 600 400 400 600
2 600 800 800 400 600 800 600 600 600
4 600 600 800 400 600 800 400 600 600
38◦ 6 400 600 800 400 600 600 400 600 600
8 400 600 800 400 600 600 400 400 600
10 400 600 800 400 400 600 400 400 600
2 600 600 800 400 600 800 600 600 600
4 400 600 800 400 600 600 400 400 600
40◦ 6 400 600 800 400 600 600 400 400 600
8 400 600 600 400 400 600 400 400 600
10 400 600 600 400 400 600 400 400 600

Notes: (1) Dimensions are in millimetres unless stated otherwise. (2) Only the width of the square soil-nail head is shown in
the Table.

Table 3. Merits and limitations of the methods for determining ultimate pull-out resistance.

Method Merits Limitations

Empirical Related to field performance data; Need a large number of field data and take a
correlation can better account for influencing factors. long time to establish a reasonable correlation;
a general correlation may not be applicable to all sites.
Pull-out test Related to site-specific Need to carry out a considerable number of
performance data. field pull-out tests during the design stage;
not feasible for small-scale project;
time consuming.
Undrained shear Based on soil mechanics Generally not suitable for Hong Kong;
strength principles; easy to apply. many factors that affect the
pull-out resistance are not accounted for.
Effective stress Based on soil mechanics Many factors that affect
principles; easy to apply. the pull-out resistance
are not accounted for.
Pressuremeter Related to field performance data; Need a large number of field data
can better account for to establish a reasonable correlation;
influencing factors. a general correlation may not be
applicable to all sites; pressuremeter
test is not common in Hong Kong.

101
 where c is effective cohesion of the soil, Pc is outer
perimeter of the cement grout sleeve, Lc is bond length
of the cement grout sleeve in the passive zone, D is
outer diameter of the cement grout sleeve, σν is vertical
effective stress in the soil, μ∗ is coefficient of apparent
friction of soil (μ∗ may be taken to be equal to tan φ  ,
where φ  is the effective angle of shearing resistance of
the soil), FP is factor of safety against pullout failure
at soil-grout interface.
For design purpose, the vertical effective stress in
the soil is calculated from the overburden pressure,
which implies that the contact pressure at the soil-grout
interface is governed by the overburden pressure. This Elevation
assumption is not necessarily true because the normal
stress at the face of the drillhole is reduced to zero after
drilling due to arching effect and the grouting pressure
is generally so low that only a small contact pressure
can be restored. The contact pressure is likely much
less than the overburden pressure.
The effects of hole drilling process, overburden
pressure and grouting pressure on pullout resistance
was investigated by means of laboratory pullout tests
by Yin & Su (2006). The test set up is shown in
Figure 28. Compacted fill of completely decomposed
granite was used in the tests. The study showed that
(a) the drilling process during soil nail installation led
to stress reduction in the soil around the drillhole and End View
the pullout resistances of the nails were not dependent
on the amount of vertical surcharge applied if gravity Figure 28. Schematic set-up of the laboratory pullout test.
grouting was adopted (Fig. 29); and (b) pullout resis-
tances of the soil nails increased with an increase of
grouting pressure (Fig. 30).
Pullout tests are routinely carried out in sacrifi-
cial test nails in Hong Kong for the verification of
design assumptions. In order to examine the signif-
icance of the potential stress reduction due to the
arching effect, the results of about 900 pullout tests
were reviewed. The pullout resistance measured in the
field was compared with the theoretical values esti-
mated by the effective stress method. About 84% of
the tests were conducted in granite or volcanic sapro-
lite. The rest were conducted in other types of material
such as fill, colluvium and moderately decomposed
rock.
Figure 29. Plot of pullout resistance against overburden
Many of the test nails were not loaded to bond fail- pressure.
ure because the ultimate pullout resistance (Tult ) of
the bonded section was higher than the yield strength
of steel. The pull-out tests were stopped when the
test load reached 90% of the yield strength of steel The difference between the measured and the esti-
to avoid tensile failure of the steel reinforcement. mated pullout resistance is due to many factors includ-
Figure 31 shows the plot of the ratio of the field pull- ing soil arching, restrained soil dilatancy, soil suction,
out resistance to that estimated using the effective roughness of drillhole surface, and over-break, which
stress method against the overburden pressure. The are hard to quantify in design. All these factors except
field pull-out resistances were generally several times soil arching tend to result in higher pullout resis-
higher than those estimated, but the safety margin tance than the design value. The finding of the review
(i.e. Tult(field) /Tult(estimate) ) gradually decreases when gives assurance on the adequacy of the effective stress
overburden pressure increases. method.

102
Figure 30. Plot of pullout resistance against grouting
pressure.

20
Tp (Fill) Dry
18
Tult (Colluvium) Dry
16 Tp (Colluvium) Dry
estimate

Tult (C/HDG) Dry

14
Tult (C/HDG) Wet
12 Tp (C/HDG) Dry
Tp (C/HDG) Wet
10 Tult (C/HDV) Dry
Tult (C/HDV) Wet
resistance

8
Tp (C/HDV) Dry
6 Tp (C/HDV) Wet
4 Mean
2 field/ estimate = 1
0
0 100 200 300 400 500 600

Figure 31. Plot of field (Tp + Tult ) to estimated pull-out

resistance against overburden pressure.

2.3.4 Potential effect of blockage of subsurface

drainage by soil nailing works
Soil nails installed in the ground may impede ground-
water flow and as a result dam up the water level.
To study the significance of this effect, a number of
numerical models were set up in both 2-D and 3-D for
various geological settings, subjected to infiltration
(Halcrow China Limited 2007). Typical nail spacings Figure 32. Flow Patterns in a slope (a) without soil nails,
of 1 m to 2 m were adopted in the models. Figure 32 (b) soil nails with excessive grout loss, (c) soil nails with no
illustrates an example of computed flow nets and water grout loss (after Halcrow, 2005).
table distributions for a slope under three conditions:
(a) without soil nails; (b) soil nails with excessive grout
loss, and (c) soil nails with no grout loss. head and to take action to drain the ground upstream
Results of the numerical modeling show that under of the nails.
typical conditions where there is little grout loss during
the grouting operation, there should be no significant
blockage of the drainage paths. It is also found that 2.3.5 Long-term durability of soil nails
the influence of soil nails on groundwater flow can Durability is an important aspect of soil nailing system.
be significant if excessive grout escapes laterally to The long-term performance of soil nails depends on
affect large volumes of the country rock. Therefore, their ability to withstand corrosion attack from the sur-
measures should be taken to avoid excessive grout loss. rounding ground. Soil nails of different ages exhumed
Where excessive grout loss occurs during installation from the ground in Hong Kong revealed that localized
of soil nails, the cause should be investigated and, if corrosion could occur even if hot dip galvanization
necessary, measures taken to monitor rises in hydraulic was provided, particularly in areas where voids existed

103
 soil. Due to the alkalinity of cement grout, a tight oxide
film is also formed on the surface of the steel bar. This
further protects the steel from corrosion. However,
micro-cracks will occur in the cement grout when the
soil nail is subject to tensile stress. Shrinkage cracks
may also be formed during the setting of the cement
grout. Once cracked, the function of the cement grout
in corrosion protection is not reliable.
The provision of sacrificial steel thickness is a sim-
ple and widely used method. It allows for corrosion of
the steel by over-sizing the cross-section of the steel
bar. Products of corrosion also form a protective coat-
Figure 33. Localized corrosion in exhumed soil nail. ing between the steel and its surrounding. Whilst this
coating offers no physical protection to the steel, it
may slow down the rate of corrosion by changing the
in the cement grout (Shiu & Chang 2003). Figure 33 kinetics of the chemical reactions.
shows the corrosion on an exhumed soil nail. Zinc is the most common type of metal used to pro-
The design for durability of a steel soil nail entails vide corrosion protection to steel bars. The galvanizing
the assessment of the corrosivity of soil at the site and zinc coating is strongly resistant to most corrosive
the provision of corrosion protection measures. Dif- environments. It provides a barrier protection and a
ferent soil corrosivity assessment schemes are avail- cathodic protection function to the steel.
able in the world, and different countries may adopt Non-metallic coatings in the form of fusion-bonded
different corrosion protection measures. epoxy have been used in the USA to protect steel bars
Eyre & Lewis (1987) developed two soil corrosiv- from corrosion. The epoxy coatings do not conduct
ity assessment schemes, one for general assessment electricity and they isolate the steel bars from the sur-
purposes and another for reinforced earth and culvert rounding environment. To be effective, the coating has
application. Two types of assessment scheme were to be impermeable to gases and moisture and free from
subsequently developed by others in the UK. The cracks. The interface between the steel and the coating
first was that of Murray (1993), who incorporated the has to be tight.
scheme for general assessment purposes into the spec- When a high level of corrosion protection is needed,
ifications for soil corrosivity assessment in respect of corrugated plastic sheaths can be used in conjunc-
soil nailing in the UK. Tests are conducted on soil sam- tion with cement grout. The sheath prevents ingress
ples only. The soil aggresivity classification scheme in of water and corrosive substances even if the cement
Hong Kong (GEO 2008) was developed based on that grout is cracked.
of Murray (1993). To overcome the problem of corrosion of metallic
The UK Department of Transport made reference reinforcement, non-metallic soil nails may be used. An
to the scheme for reinforced earth and culvert applica- alternative to steel reinforcement is composite mate-
tion and the recommendations of Brady & McMahon rial made of fibres embedded in a polymeric resin.
(1993) and developed a corrosivity assessment scheme It is generally known as fibre-reinforced polymers
for corrugated steel buried structures under roadways (FRP). FRP is highly corrosion resistant. The common
(Department of Transport 2001). In this scheme, both types of fibre used in composites for civil engineering
soil and water samples are collected and tested. CIRIA works are carbon FRP (CFRP), glass FRP (GFRP) and
(2005) has recommended a corrosivity assessment aramid FRP (AFRP). Shanmuganathan (2003) gave a
scheme based on that developed by the Department state-of-the-art review of the development and applica-
of Transport (2001). tion of FRP composites in civil and building structures.
Depending on the soil aggresivity of a site, the There are reported cases of using CFRP reinforce-
required design life and the intended degree of protec- ments to slope stabilization works in the USA, the UK,
tion, different measures may be adopted to protect steel Spain, Greece, Japan and Korea (e.g. Unwin 2001,
bars against corrosion. The common corrosion protec- Ground Engineering 2004).
tion measures are cement grout, sacrificial thickness Carbon fibres are the primary load-carrying com-
to the steel, sacrificial metallic coating to the steel (e.g. ponent in CFRP reinforcement, which are character-
hot-dip galvanizing with zinc coating), sacrificial non- ized by low weight, high strength and high stiffness.
metallic coating to the steel (e.g. epoxy coating), and The primary function of the resin is to provide a
corrugated plastic sheathing. continuous protection medium to the fibres and to
Cement grout can prevent corrosion by forming a transfer stresses among fibres. CFRP reinforcement
physical and a chemical barrier. The cement grout is anisotropic in nature and is characterized by high
physically separates the steel from the surrounding tensile strength in the direction of the fibres. It is

104
non-corrosive and has a much better strength-to- 2,280 MPa. This average tensile strength is about five
weight ratio than steel reinforcement. However, CFRP times that of high yield steel.
reinforcement does not exhibit yield behaviour. The The shear strength of CFRP reinforcement is gener-
lack of ductility necessitates special consideration in ally much lower than its tensile strength. Benmokrane
its application as a soil nail. et al (1997) reported that the shear strength of some
Figure 34 illustrates schematically the stress-strain CFRP reinforcements is only about 11% of its tensile
behaviour of CFRP reinforcement in comparison to strength.
steel. According to ACI (2001), the tensile strength While there are a number of national design and
of CFRP reinforcement ranges from 600 MPa to construction guides on the use of CFRP reinforcement
3,690 MPa (c.f. 460 MPa for high yield steel bar). in concrete structures (e.g. Japanese Ministry of Con-
Specific test results on tensile testing of CFRP rein- struction 1997, JSCE 1997, IStructE 1999, ACI 2001),
forcement strips (Fig. 35) which were used in a field international standard on the use of CFRP reinforce-
trial by the GEO in Hong Kong are summarized in ment as soil nails is lacking. An interim design and
Table 4. The tests gave tensile strength ranging from construction guideline for CFRP soil nails has been
1,990 MPa to 2,550 MPa, with an average value of developed by the GEO for use in its slope upgrading
programme. The following are salient points of the
GEO guideline:
• A suitably conservative estimate of the design ten-
sile strength is made using a partial material safety
factor of 3.3 to cater for the uncertainty in material
properties and to compensate for the lack of ductility
of CFRP.
• A partial safety factor of 1.4 on bond strength
between CFRP reinforcement and grout is adopted
following the recommendation of IStructE (1999).
• The inclination of the CFRP soil nails to limited
to within 15◦ from the horizontal so as to optimize
the reinforcing efficiency of the soil nails and to
limit the slope movement for the mobilization o the
tensile force in the soil nails.
• Only CFRP reinforcements with a shear strength
equal to or greater than that of steel are used.
• CFRP reinforcements with polyester resins should
not be used because of their relatively ease of
degradation in highly alkaline environment.
Figure 34. Stress/strain curves of typical high yield steel 2.3.6 Aesthetic considerations
bar and CFRP bar. A good soil nail design should give due attention to
the aesthetic aspects in addition to safety and service-
ability considerations. GEO Publication No. 1/2000
(GEO 2000) provides comprehensive guidance on the
landscape treatment and bio-engineering for slopes
and retaining walls. The publication contains general
principles and good practice for enhancing the appear-
ance of engineered slopes and illustrates these by a
large number of case examples. While the principles
of aesthetics and landscape treatment are given for
unreinforced slopes, they are generally applicable to
soil-nailed systems. The following design principles
are worth considering for enhancing the appearance
of soil-nailed slopes and retaining walls:
• Make the appearance of soil-nailed systems compat-
ible with and minimize visual impact to the existing
Loading Failure environment.
• Identify and preserve, wherever practical, mature
Figure 35. Tensile testing of CFRP strips. trees on slopes and near their crests and toes.

105
Table 4. Results of tensile test on CFRP reinforcement bars.

Specimen Width Thickness Rupture Tensile stress at Modulus of Rupture

no. (mm) (mm) load (kN) rupture (MPa) elasticity (GPa) strain (%)

1 29.0 4.3 318 2,548 120.2 2.12

2 29.0 4.5 330 2,522 119.4 2.11
3 28.9 4.5 262 1,996 111.6 1.79
4 29.0 4.4 253 1,985 114.9 1.73
5 28.7 4.6 314 2,363 103.5 2.28
Average 2,283 113.9 2.01

granitic or volcanic saprolite of a gradient up to 60◦

have been successfully vegetated using this technique.
Where the provision of vegetated surface cover on a
slope is practically not feasible, hard landscape treat-
ment can be provided to improve its appearance. Possi-
ble methods are masonry block facing, ribbed or other
patterned concrete finishes, toe planters, colouring
and planter holes. More fancy techniques such as dec-
orative artwork and artificial rock may also be used.

2.4 Construction technology

2.4.1 Construction method
The choice of installation method depends on a number
of factors such as cost, site access, working space, and
ground and groundwater conditions. A brief descrip-
Figure 36. Fixing details of steel wire mesh and erosion tion of the commonly available soil-nail installation
control mat on slope face with soil nail heads. methods is summarised below.
a. Drill-and-Grout. This is the most common instal-
• Locate soil nails and other engineering features lation method. In this method, a soil-nail reinforce-
away from tree trunks and roots. ment is inserted into a pre-drilled hole, which is
• Pay attention to the design and location of man- then cement-grouted under gravity or low pressure.
made features such as surface drainage channels, Various drilling techniques, e.g., rotary, rotary per-
stairways and catchpits in order to minimise their cussive and down-the-hole hammer, are available
visual impact, e.g. concrete aprons on either side of to suit different ground conditions. The advantage
drainage channels can be designed using geotextiles of this method is that it can overcome underground
or other bioengineering techniques. obstructions, e.g., corestones, and the drilling spoil
• Route maintenance stairways with care to minimise can provide information about the ground. The size
visual impact and paint railings in sympathetic un- and alignment of the drillholes can be checked
obtrusive colours. before the insertion of reinforcement, if needed.
• Place soil nails in a regular, rather than a random Potential construction difficulties are hole collapse
pattern. and excess grout loss. The drilling and grouting
• Recess isolated soil-nail heads and treated them process may also cause disturbance and settlement
with a matt paint of a suitable colour to give a less to the adjacent ground.
intrusive visual appearance. b. Self-Drilling. This is a relatively new method when
compared with the drill-and-grout method. The
A large variety of greening techniques are avail- soil-nail reinforcement is directly drilled into the
able in the market. The Hong Kong experience shows ground using a sacrificial drill bit. The reinforce-
that a simple method which involves the use of an ment, which is hollow, serves as both the drill rod
erosion control mat in conjunction with a steel wire and grout pipe. The installation process is rapid
mesh allows the provision of vegetation covers to steep as the drilling and grouting are carried out simul-
slopes (Fig. 36). Many steep soil-nailed cut slopes in taneously. Instead of using air or water, cement

106
 grout is used as the flushing medium, which has • drainage lines intersecting slopes, within which
the benefit of maintaining hole stability. No cen- colluvium may be present, erosion pipes may be
tralisers nor grout pipes are needed and casing is developed, and preferred groundwater through-
usually not required. However, self-drilling soil flow indicated by seepage locations/horizons
nails may not be suitable for the ground containing may also occur.
corestones as they cannot penetrate through rock
efficiently. It may be hard to ensure the alignment 2.4.2 Use of non-destructive testing method
of long soil nails due to the flexibility of reinforce- for quality control
ment. Durability may also be a concern if it relies Like other buried works, it is difficult to verify the
on the provision of grout cover and corrosion pro- quality of an installed soil nail. In the context of this
tective coatings to steel reinforcement as corrosion paper, the quality of an installed soil nail refers to
protection measures. the as-built length and the integrity of cement grout.
c. Driven. In this method, soil-nail reinforcement In order to enhance the quality control of soil nailing
is directly driven into the ground by the ballis- works, non-destructive testing (NDT) methods could
tic method using a compressed air launcher, by be carried out on installed soil nails. With the help
the percussive method using a hammering equip- of NDT, the overall picture of the quality of installed
ment, or by the vibratory method using a vibrator. soil nails can be built up, which facilitates the iden-
During the driving process, the ground around the tification of the areas for follow-up actions. A num-
reinforcement will be displaced and compressed. ber of NDT methods including sonic echo method,
The installation process is rapid and it causes mini- Mise-a-la-Masse method, magnetometry, electromag-
mal ground disruption. However, due to the limited netic induction method and time domain reflectometry
power of the equipment, this method can only be (TDR) have been examined in Hong Kong. Amongst
used to install soil nails of relatively short length. these, the TDR method was found to be reliable, sim-
Moreover, the soil-nail reinforcement may be dam- ple and not expensive (Cheung 2003, Lee & OAP
aged by the excessive buckling stress induced 2007).
during the installation process, and hence it is not The principle of TDR technique was derived in
suitable for sites that contain stiff soil or corestones. 1950s from that of radar. Instead of transmitting a
Since the soil-nail reinforcement is in direct contact 3-D wave front in radar, the electromagnetic wave
with the ground, it is susceptible to corrosion unless in the TDR technique is confined in a waveguide
non-corrodible reinforcement is used. (O’Connor & Dowding 1999). TDR is commonly used
The experience in Hong Kong shows that soil in the telecommunications industry for identification
nails can generally be constructed by means of the of discontinuities in transmission lines. In the 1980s,
drill-and-grout method without many difficulties. the application of the technique was extended to many
However, under some unfavourable ground condi- other areas such as geotechnology, hydrology, mate-
tions, construction problems may be encountered. rial testing, etc (Dowding & Huang 1994, Siddiqui et
The following geological conditions are susceptible al 2000, Liu et al 2002, Lin & Tang 2005). TDR is
to excessive grout leak during soil nail installation: based on transmitting electromagnetic pulses through
f a transmission line, which is in the form of coaxial
or twin-conductor configuration, and receiving reflec-
• fill, containing a significant proportion of coarse tions at the locations of discontinuities. By measuring
materials, i.e., boulders, cobbles, gravel, and the time for the pulses to travel from the pulse genera-
sand; tor to the point of discontinuity, one can determine its
• colluvium and fluvial deposits with a high pro- location using Equation (7).
portion of coarser material;
• erosion pipes which may be partly infilled by
porous and permeable material; L = vp t (7)
• material boundaries within colluvium, and
between colluvium and in-situ material, and where L is the distance between the pulse generator
within corestone-bearing saprolite, especially at and the point of discontinuity, and t is the respective
the margins of corestones, open joints, faults pulse travel time. The pulse propagation velocity, vp ,
and shear zones, and other discontinuities (e.g., is related to the electrical properties of the material in
zones of hydrothermal alternation, etc.) that are the close proximity to the pair of conductors by the
weathered and eroded, and so are open; following expression (Topp et al 1980):
• landslide scars, tension cracks, and other fea-
tures related to slope deformation, as these
vc
may include voids within transported and in-situ vp = √ (8)
materials; and ε

107
 According to Equation (9), a positive reflection will be
returned at the discontinuity when there is an increase
in electrical impedance (e.g. reflection 1 at the inter-
face of grout/void and reflection 3 at the end of the
pair), whereas a negative reflection will be returned
otherwise (e.g. reflection 2 at the void/grout interface).
Moreover, the pulse travel time is less than that in the
fully grouted pair. In other words, one can in-principle
determine the quality of an installed soil based on a
Figure 37. Analogy of a soil nail with pre-installed wire as TDR waveform.
a twin-conductor transmission line. In order to investigate the feasibility of applying
TDR technique in the estimation of soil-nail length,
TDR tests were conducted on prefabricated soil nails
where vc is the speed of light in vacuum (3×108 m/s) of various known lengths. Figure 39 shows the TDR
and ε is the dielectric constant which measures how a test results where reflections are returned from the
material reacts under a steady-state electric field (for respective soil-nail ends and the time of pulse prop-
air ε ≈ 1, for cement grout ε ≈ 10, for water ε ≈ 80). agation is found to be proportional to the length of the
If a wire is pre-installed alongside a soil-nail rein- soil nail.
forcement, which is generally a steel bar, as shown Based on the contrast in pulse propagation veloc-
in Figure 37, the configuration becomes analogous to ity in air and grout and the occurrence of reflections
a twin-conductor transmission line and the end of the where these is a change in impedance along the
reinforcement-wire pair becomes a discontinuity. This reinforcement-wire pair, TDR results in-principle can
suggests that TDR can be used to determine the length be interpreted to infer the grout integrity of a soil nail.
of installed steel soil nails. To examine this, TDR tests were conducted on prefab-
As indicated in Equation (1), the two key parameters ricated soil nails with built-in grout defects of varying
that have to be known for the estimation of soil-nail void sizes at different locations along the soil nails as
length are (i) the time for a pulse to travel from the shown in Figures 40 and 41. These model test results
reinforcement head to its end, t, and (ii) the pulse indicate that soil nails with significant grout defects
propagation velocity, vp . Equation (8) further sug- will result in shorter TDR-deduced length with some
gests that the pulse propagation velocity, vp , along characteristic patterns in the TDR waveform. These
a reinforcement-wire pair in air will be much greater patterns depend on the location as well as size of the
(2 to 3 times) than that in cement grout. Hence, the defects.
pulse travel time along a soil nail with voids in grout Up to the end of 2007, over 10,000 soil nails at
will be less than that in a fully grouted soil nail of the about 850 sites have been tested using TDR in Hong
same length. Kong (Fig. 42). In general, the percentage difference
Apart from the effect on pulse propagation veloci- between the TDR-deduced length and design length of
ties, a reflection will be induced whenever an electrical the soil nails is small and lies within the uncertainty
pulse reaches the location of discontinuity in the limit of the test. There were a small number of soil nails
grouted reinforcement-wire pair (e.g. the end of a soil (less than 1%) with such difference exceeding an alert
nail or a void). The magnitude and polarity of the limit and displaying anomalous TDR wave forms. Fur-
reflection depend on the amount of changes in electri- ther investigation was conducted where needed. The
cal impedance at the location of discontinuity, which anomalies encountered so far were found to be due to
can be expressed in terms of the reflection coefficient, grout defect in the soil nails.
(Hewlett Packard 1998): The experience shows that the TDR technique can
be an effective tool to supplement site supervision in
Vr Z − Zo the quality control of soil nailing works, which cannot
= = (9)
Vi Z + Zo be checked easily after construction. While TDR, like
any other NDTs, does not give definitive answer to
where Vr is the peak voltage of the reflected pulse, the cause of anomalous test results, it flags up soil
Vi is the peak voltage of the incident pulse, Z is the nails that warrant further examination and coupled
electrical impedance at the point of reflection and Zo is with appropriate NDTs, the conditions of the soil nails
the characteristic electrical impedance of the grouted can be ascertained (Pun et al 2007).
reinforcement-wire pair.
Figure 38 shows a theoretical TDR waveform of
2.5 Performance of soil-nailed system
a cement grouted reinforcement-wire pair with void
section in the middle. There will be reflections at the Since the introduction of the soil nailing technology to
location of the void as well as the end of the pair. Hong Kong in late-1980’s (Watkins & Powells 1992),

108
Figure 38. Theoretical TDR waveform of a soil nail with defect in grout sleeve.

Figure 39. TDR results on soil nails of various known

length.

Figure 40. TDR waveforms for soil nails with a void

section between grouted sections.
more than 3,700 soil cut slopes have been upgraded by
means of soil nailing in Hong Kong. The performance
of these soil-nailed slopes gives an indication of the
reliability of the soil nailing technique. of a similar order of magnitude as that of engineered,
In the period between 1993 and 2006, a total of unsupported soil cut slopes in Hong Kong.
25 landslide incidents on permanent soil-nailed slopes Amongst the 25 reported landslide incidents, 15
were reported to the GEO. The landslides were gen- cases were reviewed in detail. All the 15 landslide
erally of small scale, involving local shallow sliding incidents involved active zone failures. There was
failure or washout, with failure volumes ranging from no report of external failure or passive zone failure.
less than 1 m3 to a maximum of about 35 m3 . The The slopes were steep and had a vegetation cover
average annual failure rate of such relatively minor before failure, four of which had a gradient equal
landslides on soil-nailed slopes between 1997 and to or exceeding 45◦ and the other 11 exceeding 50◦ .
2006 is 0.078%. This average annual failure rate is Twelve of the 15 cases were associated with surface

109
Figure 41. TDR waveforms for soil nails with two void Figure 43. Shallow failure at a soil-nailed slope.
sections between grouted sections.

Figure 42. Conducting TDR test at a soil nail.

erosion and detachments from the near-surface mate-

rials between the nail heads (see example in Fig. 43).
In two cases, the soil-nail heads were partially exposed
but the nail reinforcements and grout sleeves remained
intact. In the remaining case, the soil-nail heads were
undermined and the soil nail reinforcement bars were
bent (see Fig. 44). The common factors contribut-
ing to the landslides were inadequate slope protection, Figure 44. Failure of a soil-nailed slope undermining soil-
inadequate drainage provisions or presence of adverse nail heads.
geological or hydrogeological conditions.
The review shows that steep soil-nailed slopes with
vegetated covers are fairly vulnerable to minor failures, inclination in relation to the steep slip surfaces with
as the potential for shallow small detachments between regard to the mobilization of tension forces in the soil
soil-nail heads within the active zone of the soil nail nails. In case of shallow failures on a steep vegetated
system cannot be prevented effectively by means of the soil-nailed slope, there is little horizontal displace-
soil nails. The efficiency of the soil nails, which are ment to mobilize the tensile forces in the nail bars
typically shallowly inclined for the upgrading of sub- and hence their effectiveness in stabilizing potential
standard soil cuts in Hong Kong, is not high in so far as vertical/subvertical failure surfaces may be limited.
prevention of shallow detachment is concerned. This No failure with a volume larger than 50 m3 on soil-
is because the nails are not orientated at an optimal nailed slopes has been reported in Hong Kong so far.

110
As a comparison, the average failure rate of land-
slides with a volume larger than 50 m3 for engineered,
unsupported soil cuts in Hong Kong is 0.018%. Soil
nails appear to be effective in preventing large-scale
failures.
It is also worth noting that no landslides have
been reported on soil-nailed slopes with a hard sur-
face cover in Hong Kong. A hard surface cover is
effective in minimizing surface infiltration and pro-
vides a better protection against surface erosion than
a vegetated cover. However, a hard surface cover Figure 45. Approach to design drains to stabilize a slope.
may not be acceptable from the environmental and
aesthetic points of view. A variety of flexible struc-
tural facing such as tensioned wire mesh may be domain, and (iii) the amount of lowering of the water
used to enhance the stability of vegetated slopes. table. Problems related to excessive ground settle-
This would however increase the construction cost. ments are expected when the drained soil is very thick,
A proper design has to balance between the risk of as in the case of deep drains.
possible minor failure, cost and environmental con- As regards the transient phase, two aspects have to
siderations. In this regard, risk mitigation measures, be evaluated in the design:
such as debris traps, toe barriers or buffer zones, may
be considered as an integral part of the slope design to a. whether the delay until the drains are completely
cater for possible minor detachments from vegetated effective is affordable,
slopes. b. whether settlements associated with de-watering
will damage buildings and infrastructures at the
ground surface.
3 SUBSURFACE DRAINAGE
The steady-state condition is usually analyzed by
3.1 General aspects assuming continuous infiltration of water at the ground
surface to recharge the water table. In the literature,
In saturated soils, subsurface drains are widely used results of steady-state analysis are often presented
as control measures against slope instability, as they in non-dimensional design charts, that practitioners
are less costly than other types of stabilization works generally use to design drainage systems.
and suitable for a large number of cases, even when The water flow captured and discharged by drains
the landslide is very deep and structural measures are depends largely on the permeability of the drained
not effective. soils. In steady-state condition the permeability of
The mechanism of drains inside slopes involves a the ground does not affect the amount of lowering
decrease in pore pressures in the subsoil and conse- of the pore pressures in the subsoil, which depends
quently an increase in effective stresses and soil shear on the hydraulic conditions at the boundaries of the
strength in the whole drained domain. In particular, examined domain and the geometry of the drainage
the increase in soil shear strength along the potential system. Thus the quantity of discharge is not an indi-
sliding surface of the landslide body, due to the func- cator of the performance of drains, which has to be
tion of drains, is responsible for the slope stability investigated by means of piezometers to measure the
improvement. Therefore the first step in the design change in the level of water table by drains. Indeed,
of a drainage system is the determination of the pore pore pressure changes are the most direct and useful
pressure change that is required to increase the factor indicators of drains being in good working condition.
of safety of the slope to the design value (Fig. 45). Measurements of surface and deep displacements are
The next step is to design the geometric configuration good indicators of overall slope stability. These mea-
of drains that will result in the required pore pressure surements complete the instrumentation framework
change. (Fig. 46).
The effect of the drainage system is usually ana-
lyzed for the steady-state condition, which is attained
some time after drainage construction (i.e. in the long
3.2 Drain types
term). After drain installation, a transient phenomenon
of equalization of pore pressures occurs, provoking Among the measures for slope stabilization, drains are
subsidence of the ground surface. The magnitude probably the most commonly used. Because of their
of subsidence depends on (i) the compressibility of widespread application in very different ground condi-
the soils concerned, (ii) the thickness of the drained tions and geomorphology, the technology in this field

111
Figure 46. Measurements to evaluate that drains are in good
working condition. Figure 48. Superficial and deep trenches, with main and
secondary branches: a) map, b) longitudinal section.

Figure 47. Superficial and deep trenches, with only main

branches: a) map, b) cross section. Figure 49. Horizontal drains inserted from the ground
surface: a) map, b) longitudinal section, c) cross section.

is advancing continuously. Nowadays the most com-

mon types of drains used in geotechnical engineering
applications are as follows:
• Superficial and deep trenches, with main (Fig. 47)
and possibly secondary (Fig. 48) branches,
• Horizontal drains (Fig. 49), executed from the
ground surface,
• Wells, with or without horizontal drains (Fig. 50),
• Tunnels, with or without horizontal drains.
Drain trenches should be excavated deep enough
to intercept the regions of positive pore pressures.
Superficial trenches can be excavated by means of an
excavator up to a depth of approximately 5 m from the
ground surface. The width of the trench is dependent
on the type of excavator being used and may vary from
0.5 to 1.0 m.
In open areas, trenches can have sloping sides, the
gradient of which is based on stability consideration.
Where there is not enough space, trench sides have
to be formed to vertical and should be properly sup-
ported. Guidelines on the design of lateral support to
excavation are given in many publications, e.g. BS
6031:1981 (BSI 1981). Problems of trench instabil- Figure 50. Drain wells with and without horizontal drains:
ity can be reduced by opening up trenches in short a) map, b) longitudinal section.

112
Figure 51. Scheme of a superficial trench.
Figure 52. Scheme of a horizontal drain.

lengths and backfilling the trench within a short time To reduce precipitation of calcite it is good prac-
after excavation. tice to drill the hole at an inclination slightly above
Deep trenches can be excavated by means of grab horizontal, such that the pipe is not continuously
shells. The sides of the trenches, being vertical, should submerged. Conversely, there are other chemical phe-
be supported by slurry, e.g. polymeric mud. nomena, favored by bacterial activity, that are due to
Trenches need to have a high discharge capacity to aeration (Walker & Mohen 1987).
avoid the saturation of the backfilling material or of At the portion of a horizontal drain near to the
the lower portion of it. This can be achieved providing slope surface, it is recommended to use a 3–6 m long
a drainage layer of gravel materials or installing at the un-perforated pipe, grouted all around with cement,
bottom of the trench a perforated pipe (with slots on to prevent the penetration of tree roots into the pipe,
the upper part). The perforated pipe should be wrapped which could block the water flow.
with a geotextile to prevent the clogging of the slots by Wells and tunnels are costly and complex to con-
fine soil particles (Fig. 51). A compacted clay cover struct; for this reason they tend to be used in special
should be placed on the top of the trench to prevent circumstances, e.g. in deep landslides where other
ingress of surface water, which should be drained by types of drain are unable to reach the sliding surface.
means of a system of surface drainage network. The Drain trenches and horizontal drains are the most
impermeable cover should have a minimum thickness commonly used drain systems in slope stabilization.
of 0.5 m and should be compacted in layers. Therefore these are treated in detail in the next part of
Trenches should be constructed starting from the this paper.
lowest point in the area to be drained, so that they can
drain water during construction. Inspection wells that
intercept the trenches should be installed to allow: 3.3 Influence of groundwater on landslides
in saturated soils
• monitoring of the working condition of the drainage
system, possibly by measuring the flow; Investigation carried out in many parts of the world
• maintenance, possibly flushing of the perforated (Kenney & Lau 1984, Urciuoli 1998) showed that
pipe. pore pressures at shallow depths are strongly influ-
enced by seasonal atmospheric conditions. Urciuoli
Horizontal drains involve the drilling of holes in (1998) showed, on the basis of piezometer measure-
the ground. The diameter of the hole is usually 100- ments, that a critical line can be drawn inside clay
120 mm, and it is drilled with a tricone or drag bit. formations, separating the zone in which groundwa-
A PVC slotted pipe, protected by a geo-textile, is ter regime is transient, due to variation in atmospheric
inserted in the hole (Fig. 52). The maximum length conditions, from the deep zone, in which pore pres-
of horizontal pipes is around 100 m, but in some cases sures remain essentially constant throughout the year
it has been possible to reach 300 m. (Fig. 53). The division of the saturated soil domain
Deposits of calcium salts and iron oxide can block into two zones is useful to interpret the mechanism of
horizontal drains; regular maintenance by flushing the landslides and its relation with pore pressures, as is
pipes with a high pressure water jet, should therefore described below.
be programmed. In the absence of maintenance, drain Active shallow landslides are usually characterized
pipes cannot remain functional for a long time. by the presence of a sliding surface in the zone affected

113
 analyzed by considering the groundwater regime as
a steady-state phenomenon.

3.4 Groundwater regime

3.4.1 Equation
If the subsoil is assumed to be a saturated porous
medium, characterized by isotropic permeability K,
the equation governing transient flow can be obtained
by imposing the conservation of the liquid mass.
Darcy’s Law is inserted therein. In this way the
subsequent equation is obtained:

−K · (hxx + hyy + hzz ) = ε̇v (10)

Figure 53. Pore pressure in clayey formation: a) measure-
ments at Casagrande piezometers installed at different depths
(Basento Valley, Italy), b) scheme of pore pressure regime in where h is the piezometric head:
the subsoil.
u
h=ξ+ (11)
γw
by a transient groundwater regime. In this case, the
safety factor of the landslide body varies with time ξ is the geometric head, u is the pore pressure and
along with pore pressure fluctuations. Pore pressures γw is the water unit weight; ε̇v is the volumetric strain
attain their maximum values during wet seasons: shal- rate of the soil.
low landslides can reactivate, as a direct consequence Assuming the solid skeleton is an isotropic linearly
of reduction in shear strength along the sliding sur- elastic medium, with Young’s modulus E and Poisson
face. The rate of slope movement is characterized by index ν, volumetric strain is expressed as:
a seasonal trend (Pellegrino et al. 2004b).
In active deep landslides, slope movement may well 3(1 − 2ν)
be due to intense plastic strains, occurring in some εv = dp · (12)
E
regions of the subsoil where shear stresses are close
to the failure envelope of the soil. In other words, dis- where dp is the variation in the mean effective stress
placement points of the landslide body may not be during the transient phenomenon.
localized on a sliding surface. The rate of movement Under the assumption that mean total stresses
is generally very small and constant (i.e. not charac- p( p = p + u) remain constant in all the volume
terized by a seasonal trend), according to the steady subjected to the analysis we obtain:
groundwater regime.
In both cases drains can play an important role in
dp = −du = −γw dh (13)
contributing to slope stability. In potential or active
shallow landslides drains might prevent the rise of the
water table, which is a consequence of atmospheric Substituting eq. (13) in eq. (12) and eq. (12) in
changes, up to the critical level that endangers slope eq. (10), we obtain:
stability. Nonetheless, methods of analyzing the sta-  
bilization effect of drains commonly available in the ht − cv hxx + hyy + hzz = 0
literature (e.g., Hutchinson 1977, Desideri et al. 1997)
KE
model the groundwater regime as a steady-state phe- cv3D = (x, y, z) ∈  0 < t ≤ T (14)
nomenon (seepage) and assume the presence of a film 3(1 − 2ν)γw
of water on the ground surface. In areas where the
weather is not very rainy, such as in southern Europe, where cv3D is the coefficient of consolidation in 3D
this assumption underestimates the effects of drains on condition;  and T are respectively the spatial and
slope stability. time integration domains.
In potential or active deep landslides drains reduce The first of equations (14) can be simply specialized
the steady-state pore pressures, which in turn in- to 2D and 1D cases; contextually the expression of
crease the difference between mobilized shear stresses the coefficient of consolidation has to be modified,
and the soil shear strength. In this way the plas- according to table d5.
tic strain rate in the overstressed zone is drastically Once the hydraulic boundary conditions are fixed,
reduced. The effectiveness of drains is correctly the transient solution tends to a steady distribution

114
Table 5. Consolidation equation and coefficient of consolidation for 2D and 1D conditions.

Conditions 2D 1D

Equation ht − cv2D (hxx + hzz ) = 0 ht − cv1D hzz = 0

(σx + σz )(1 + ν) σz (1 + ν)
p (mean effective stress)
3 3(1 − ν)
(1 + ν)(1 − 2ν) 1 − ν − 2ν 2 
εv (volume strain) d(σx + σz ) dσz
E (1 − ν)E
assumption d(σx + σz ) = 0 dσz = 0

KE KE · (1 − ν)
cv (coefficient of consolidation) cv2D = cv1D =
2 · γw (1 + ν)(1 − 2ν) γw (1 − ν − 2ν 2 )

of pore pressures u(∞, x,y,z), that can be obtained

directly by integrating equation (15):

(hxx + hyy + hzz ) = 0

(15)
(x, y, z) ∈  0 < t ≤ T

From eq. (15), it is clear that the steady solution

does not depend on the properties of the soil.
At equations (10) or (15), hydraulic conditions at
boundaries must be added. Among them, water flow
through the ground surface affects pore pressures in
subsoil more than others. To take this aspect into
account, two different hydraulic conditions at ground
surface are considered in this paper:
Figure 54. Usual geometry of drain systems: a) drain
• a film of water continuously present, trenches, b) horizontal drains. The cells analysed in this paper
• a flux of water varying with a seasonal trend. are delimited with bold lines.

When, in practice, drainage works is analysed by

means of numerical codes (DEM or FEM), the prob- water fixed at ground surface. From this configuration
lem may be solved by taking soil stratigraphy and 2D solutions are derived as particular cases.
heterogeneity into account. Pore pressures can be cal- Cells are considered, each containing a single drain
culated all along the critical sliding surface; then they and confined on lateral boundaries with imperme-
can be used in slope stability analysis. able planes (Fig. 54). The scheme represents a case
Practitioners very often prefer to estimate pore pres- in which an infinite distribution of drains is work-
sure, lowered by drains, by means of non-dimensional ing, so that it is possible to isolate the soils drained
charts obtained for homogeneous soil and very sim- by a single drain, between two consecutive planes
ple geometric schemes. Design charts are a general of symmetry, that behave as impermeable surfaces.
tool: they cannot consider hydraulic conditions at Hydraulic conditions along cell boundaries are indi-
ground surface with a seasonal trend, which necessar- cated in Figure 55. At ground surface the following
ily depends on typical climatic features of the region condition, which represents the film of water at the
being considered. Hence the design charts presented ground surface and is able to recharge the water table,
in this paper are obtained under the most generic is imposed:
assumption of a water film at the ground surface.
h=z (16)
3.4.2 First solution (film of water at ground surface)
In this section steady-state solutions are presented for Under this assumption all the examined domain is
drains operating in 3D conditions, assuming a film of submerged.

115
 Groundwater regime in natural conditions, before
the construction of drains, can be very often schema-
tized through the model of infinite slope (1D), which is
a very simple case to analyse, the pore pressure regime
being described by the relationship below:

γw · (D − Hw )
u= (17)
1 + tan α · tan β

where D and Hw are the depths of a generic point

and the water table surface respectively, both measured
along the vertical from the ground surface (Fig. 59); α
and β are the slopes of the ground surface and of the
water flux respectively.
Supposing the water flux is parallel to the ground
Figure 55. Analysed cells, with indication of hydraulic level and using the quantity zw , which expresses the
boundary conditions. depth below the water level measured along the normal
to the slope, eq. (17) is changed into eq. (18):
Condition (16) is imposed along the drain bound- u = γw zw cos α (18)
aries1 ; it physically means that inside the drains the
pore air pressure is atmospheric. As has been shown above, distribution of pore
The presence of drains inside soils modifies the pressure caused by drains is complex. Therefore sim-
water level in a complex way. In the analysed cell plification is required to handle the problem more
represented in Figure 56 this situation is shown for manageably. Accordingly, 3D pore pressure distri-
the case of drain trenches, by means of pressure head bution resulting from the action of drains can be
envelopes on three planes parallel to the ground sur- schematized as a 1D distribution, equivalent to 3D
face at three different depths. The lowering of the water distribution as regards its influence on slope stabil-
table caused by drains is not homogeneous in : it ity. 1D distribution can be obtained by replacing the
depends upon the distances of the examined point from pressure head envelope on each plane parallel to the
the drain boundaries and from the ground surface. The ground surface with a uniform one. In Figure 60, on
drainage effect is weaker in the deepest zone of the the longitudinal section of the slope, is represented
slope. the equivalent pore pressure regime that reconnects
Drains cause a non-hydrostatic pore pressure dis- the problem to that of an infinite slope.
tribution, involving a flux downwards. The vertical Assuming 1D equivalent condition in groundwater
component of head gradients are larger near the drain regime, pore pressure is easily calculable. On a pla-
boundaries where pore pressures can be much lower nar sliding surface parallel to the ground level, pore
than those of hydrostatic ones. pressure is uniform and is calculable by means of
The effect of drain surface on pore pressure dis- eq. (17). Along a circular sliding surface , pore pres-
tribution is also evident in Figure 57, where the case sure u (t, x) is a function of the depth of the point of
of three trenches all around the cell is compared to . To simplify the analysis mean pore pressure on ,
the case with two trenches positioned in front of one ū (t, ), can be defined as follows:
another. 
The action of horizontal drains is represented in u (t, x)dx

Figure 58, through the pore pressure distribution and ū(t, ) = (19)
l
its evolution in time, along three vertical axes and three
horizontal ones. where l is the length of the unstable mass on the
To use design charts it is useful to refer the problem ground level, between the two ends of the sliding sur-
to a simple framework that describes pore pressure face. Because a film of water is assumed at ground
regime in slopes, as indicated in the following. surface, sliding surface is completely submerged, and
the average pore pressure ū (t, ) along it can be eval-
uated at the depth 2/3z , where z is the maximum
1 For horizontal drains, Marino (2007) has shown that the depth below the water level in direction z, normal to
hole can be schematized as a segment coincident with its the slope.
axis. With this simplification she obtained results very The effect of drainage is often evaluated by means
close to the more rigorous ones in which the hole is of the efficiency E(t, x, y, z), which in each point of 
correctly schematized as a cylinder. expresses the difference between the initial and current

116
Figure 56. Pressure head envelope between trenches on three planes at different depths: a) D/H0 = 0.5, b) D/H0 = 1, c)
D/H0 = 1.5.

Figure 57. Pressure head envelope on the plane at the base

of trenches for the case of 3 drain surfaces (a) and 2 drain
surfaces (b), adjacent to the analysed cell.

value of pore pressure (at a generic time t), normalized

to the initial value: Figure 58. Pore pressure distribution determined by hori-
zontal drains, along 3 vertical axes (a, b, c) and 3 horizontal
u(0, x, y, z) − u(t, x, y, z) axes (e, f, g).
E(t, x, y, z) = (20)
u(0, x, y, z)

When design charts are used, the function E, The function Ē∞ plays a key role in designing slope
defined in each point of , is not particularly use- stabilization by drains, because it represents the final
ful for evaluating the overall stability of slopes. We distribution of pore pressure [ū(∞, )], used in the
need a function that expresses the global evolution of calculation to obtain the desired improvement in slope
pore pressure along the critical sliding surface . stability.
By means of ū(t, ) we can define the average In practice, Ē∞ () is calculated, after determining
efficiency Ē along the sliding surface: the function ū(∞, ) from slope stability analysis, as
the pore pressure distribution that guarantees the safety
factor chosen by the designer. From Ē∞ (), by means
ū(0, ) − ū(t, )
Ē(t, ) = (21) of non-dimensional charts, the designer can determine
ū(0, ) the geometric characteristics of the drain system.
On the basis of the analysis presented in this paper,
Finally, with a view to representing the steady-state the value of Ē∞ (D) has been calculated and repre-
solution, the function Ē∞ can be used: sented in design charts (see Appendix A) as a function
of the following parameters:
ū(0, ) − ū(∞, ) H = depth of analysed volume ,
Ē∞ () = (22)
ū(0, ) H0 = depth of drain,

117
Figure 59. Infinite slope with a generic water flux.

Figure 61. Water flux recharging water table during the

year, obtained from monthly rainfall, from which runoff
and evapotranspiration are subtracted (from D’Acunto and
Urciuoli 2006, modified).

ground surface and an amount of rain infiltrates. If

the climate is cool, evapotranspiration is low and most
of the infiltrated water recharges the water table. If
the climate is warm, rain evaporates and the quantity
that infiltrates is removed by evapotranspiration. This
case is solved only for a system of drain trenches anal-
ysed in 2D conditions so as to quantify the influence of
hydraulic conditions at the ground surface on drainage
efficiency.
Figure 60. Transformation of the real pressure regime Having fixed a flux at the ground surface, instead of
(represented in the cross section) in a equivalent one (in the a head, the position of the water table is unknown and
longitudinal section). is obtainable from analysis. In the cases studied, the
ground water level is always located a few metres from
the ground level. In this case, the sliding surface is par-
D = depth of the plane on which efficiency is tially submerged. In order to calculate efficiency (E),
evaluated (correspondent to sliding surface), in all points above the water table it is assumed u = 0
Ly = longitudinal length of the analysed volume (neglecting suction); thus efficiency varies from 0 to
 (in the case of trenches it is the spacing between 1 during the phenomenon of pore pressure equaliza-
principal branches of drain trenches), tion. As regards the depth at which efficiency has to be
S = spacing between secondary branches of drain calculated if the sliding surface is circular, mean pore
trenches, pressure is first calculated along the submerged part of
i = spacing between horizontal drains, the sliding surface. It corresponds to the pore pressure
l2 = length of secondary branches of drain trenches, at an equivalent depth zw = 2/3z . Hence this value
l1 = Ly − l2 . has to be adjusted to take into account the part of the
sliding surface above the water level, along which pore
3.4.3 Second solution (variable water flux pressure is zero. To this aim, in Figure 62 a coefficient
at ground surface) [η] is proposed to make this adjustment, based on sim-
The solution presented in this section regards the effect ple geometric considerations on the examined problem
of an unsteady hydraulic condition at the ground sur- that are not illustrated here to save space. By means of
face, where a function of time, representing water flux η the equivalent depth must be calculated as 2/3·η ·z .
which recharges the water table, is applied (Fig. 61). While at work, drain trenches are partially sub-
Water recharging the water table is calculated by sub- merged. This situation influences hydraulic conditions
tracting runoff and evapotranspiration from rainfall, on the contacts between drains and soils. Given that
using a study on groundwater regimes in relation to cli- backfilling material has a degree of saturation more or
matic features in southern Italy (D’Acunto & Urciuoli less equal to zero, water can flow only from the subsoil
2006). The adopted function expresses the concept that to the trenches and not vice versa. In this case, a double
when rainfall is abundant a film of water forms at the boundary condition is necessary on the trench:

118
Figure 62. Corrective coefficient to evaluate the equivalent
depth at which efficiency must be calculated in the case of
partially submerged circular sliding surface.

hn = 0, if u < 0 in  (the trench boundary must be

considered impermeable),
h = z, if u > 0 in ,

where n is normal to the trench boundary.

In Figure 63 the solution of a case with a flux
applied at the ground surface, taken from D’Acunto
et al. (2007), is reported. The piezometric surface is
compared to the pressure head envelope on the plane
through the bases of trenches at different times: at the
initial condition, one year after, eight years after the
beginning of trench work (data on soil and geome-
try are reported in the figure). In each diagram two
axes are reported: on the left there is the height above
the z = 0 plane (positioned at 10 m from the ground
surface), which is useful to locate the water table; on
the right there is the pore pressure scale, to represent
the envelope of pore pressure on the plane through the
base of the trenches. The position of the water table
allows us to determine the height of the trench which is
working at the time considered (it is the height below
the water table). The envelope of pore pressure on a
deep plane allows the efficiency of the trench to be
Figure 63. Evolution of water table (WT) and pore pressure
assessed on that plane. It can be seen that the water (PP) on the plane at the base of trenches in the case of a
table (whose position is essential to fix the bound- water flux applied at the ground surface: a) initial condition,
ary condition on the trench wall) is well above the b) 1 year after the beginning of the drainage process, c) 8
pore pressure envelope. This means that much of the years after (from D’Acunto et al, 2007). In figure a) the water
height of the trench works long after the beginning of table and pore pressure are coincident; the difference is due
drainage, and trenches are always ready to discharge to different scales used for the representation of WT and PP.
the water flow of heavy rains. This explains the abil-
ity of drain trenches to avoid pore pressure peaks (as
3.5 Case history
is shown in Figure 53 for natural regimes) during wet
seasons. Measurements carried out on instrumented 3.5.1 The site
sites where drains were constructed show that the In 1990, in southern Italy, during highway works along
water table is not subject to seasonal fluctuations where the Sele river some ancient landslides were reacti-
drains are in good working condition. This is clearly vated as a consequence of trench excavations. After
shown in the next section. these events further works was needed to assure the

119
safety of the highway. In particular, in the urban district
of Contursi, the highway crosses an extensive unsta-
ble zone affected by several landslides. The unstable
area was stabilized by drains, consisting of trenches in
some zones and wells with horizontal drainage pipes
in others (Fig. 64).
The groundwater regime was investigated before
and after drain construction, using Casagrande pie-
zometers installed inside boreholes in the whole
area. The investigation program was fairly extensive,
aiming to measure pore pressure fluctuations inside
and around the landslide body and in the substra-
tum, in order to determine the hydraulic conditions
at the boundary of the unstable area. Piezometer
measurements were carried out continuously from
1991 onwards (Fig. 65). Data are grouped accord-
ing to zones. As regards zones far away from drains
(Fig. 65a), where the groundwater regime is affected
by natural conditions, pore pressure undergoes sea-
sonal fluctuations of the order of some metres, more
significant in the upper cells but noticeable also in
the lower cells. As regards the drained zone, measure-
ments at piezometers installed between trenches show
that pore pressures are well below those measured far
from the drained area. Moreover, seasonal fluctuations
more or less vanish. This is more evident in zone B,
which is the closest to the drain system, while in zone
A the behaviour is intermediate.

Figure 65. Pore pressure measurements at Contursi in

zones: a) far from drainage, b) intermediate, c) very close
to drainage (from Pellegrino et al. 2004a).

These observations show the greater stabilizing

effect of drains, since the maximum pore pressure,
attained during winter and early spring, is the cause of
reactivation of seasonal landslides. This aspect of drain
action consisting in the ‘‘lamination’’ of the seasonal
trend has not been stressed in the literature despite its
great importance in slope stability.
At the Contursi site, soils were investigated exten-
sively both in the laboratory and in the field: the
most reliable value of permeability k was obtained
from fallen head tests carried out inside piezometers.
Results were very scattered, with an average close to
10−8 m/s. As regards the coefficient of consolidation
cv , a value of 3.8 · 10−5 m2 /s was calculated from the
permeability obtained by in situ tests and oedometer
modulus Eoed , by laboratory tests (cv = k·E γw , γ w =
ed

unit weight of water).

3.5.2 Analysis of groundwater regime in the site

Figure 64. The slope uphill of the highway along the Sele of Contursi
river in Contursi; stabilized with drain trenches and wells: a) Referring to the described site the slope was modelled
map, b) stratigraphic section (from Pellegrino et al. 2004a). as a trapezium (Fig. 66), constituted by homogeneous

120
Figure 66. Analysed domain with hydraulic conditions
adopted at the boundaries (D’Acunto and Urciuoli 2006).

soil, whose bases are 105 m (H ) and 70 m, and length

280 m(L). The slope α is 7◦ 1 . Numerous parametric
analyses were performed in order to study the response
of the water table to different conditions at the bound-
aries. It emerged that water flow through the ground
surface was the hydraulic condition that most influ-
enced pore pressures in subsoil around the slip surface.
This result coincides with the fact that the slip sur-
face is relatively shallow (its depth is always less than
9 m). Hence major efforts were made to examine
the hydraulic condition at the ground surface, which
allowed the function in Figure 61 to be obtained.
The vertical boundaries on the uphill side were con-
sidered a section of an almost infinite slope, since
there was no perturbation either of the geometry or
soil stratigraphy nearby. It was then supposed that
around the boundary the pore pressure was constant
along the direction, parallel to the ground surface. All
along the base of the trapezium, the piezometric head
was assigned according to a linear function, starting
at the downhill side where the water level was at 65 m
(in the Sele River). As regards the hydraulic condition
on the downhill vertical side of the examined volume it
should be noted that this boundary is very close to the
alluvial soils of the Sele river. Hence it is considered
to have a greater capability of draining water from the
analysed subsoil. Figure 67. Groundwater regime at Contursi. Compari-
The analysis was developed in transient conditions, son between measurements and numerical results (from
regulated by the function in Figure 61. Thus results D’Acunto and Urciuoli, 2006): a) calculated pore pres-
were obtained that represent the pore pressure fluc- sure without and with drains, b) position of piezometers
tuations without and with drains as a consequence of with respect to drain trenches, c) comparison between
the variable hydraulic conditions at ground surface. measurements and numerical results for different boundary
When drains are working, fluctuations are drastically conditions.
reduced: Figure 67a shows the seasonal trend of pres-
sure head in a point at a depth of 6.6 m from the ground
3.5.3 Influence on efficiency of the hydraulic
surface. Numerical results are compared (Fig. 67c)
condition at the ground surface
with measurements at piezometers inside two sec-
By considering the imposed flux at the ground surface,
ondary branches of trenches (piezometers TC3, Sc and
instead of the water film, the following results were
SF in Fig. 67b). It can be seen that the solution obtained
obtained:
for a permanent film of water at the ground surface
overestimates pore pressure in the subsoil. Measure- • the water table is strongly depressed compared with
ments are well fitted by the analytical solution with the conditions without drains,
flux imposed at the ground surface. • seasonal fluctuations are lowered by drains.

121
 To evaluate efficiency, reference should be made Geotechnical Engineering in Naples. The related stud-
to maximum values assumed by functions uo and u∝ ies was mainly led by the second author with contribu-
during the year. It is thus possible to calculate that effi- tions made by Prof. Berardino D’Acunto, Ms. Nunzia
ciency is up to 40% greater than that corresponding to a D’Esposito (Eng.) and Ms. Roberta Marino (Eng.),
film of water fixed at the ground surface (D’Acunto & that are gratefully acknowledged.
Urciuoli 2006).

REFERENCES
4 CONCLUSION
ACI 2001. Guide for the Design and Construction of Concrete
Soil nailing is a robust slope stabilization mea- Reinforced with FRP Bars (ACI 440.1R-01). American
sures and is generally applied to stabilize man-made Concrete Institute. USA.
slopes, retaining walls or excavations. The interaction Babu, G.L.S., Murthy, B.R.S. & Srinivas, A. 2002. Analysis
of construction factors influencing the behaviour of soil-
between the ground and the soil nails is complex. Nev- nailed earth retaining walls. Ground Improvement 6(3):
ertheless, the findings of numerous researches have 137–143.
provided some understanding of the mechanism and Benmokrane, B., Xu, H. & Nishizaki, I. 1997. Aramid
behaviour of soil-nailed systems, particularly on the and carbon fibre-reinforced plastic prestressed ground
effect of nail inclination, soil-nail heads and bend- anchors and their field applications. Canadian Journal
ing stiffness of reinforcement bars. Design practice of Civil Engineering 24: 968–985.
is well established. In contrast, there is room for Brady, K.C. & McMahon, W. 1993. The Durability of Cor-
advancement of the construction technology and the rugated Steel Buried Structures. Transport and Road
application of innovative materials as soil nail rein- Research Laboratory, Department of Transport.
Bridle, R.J. & Davies, M.C.R. 1997. Analysis of soil nail-
forcement. The use of non-destructive technique can ing using tension and shear: experimental observations
greatly enhance the confidence of quality control of and assessment. Geotechnical Engineering, Proceedings
soil nail construction. of the Institution of Civil Engineers, July, 155–167.
Subsurface drainage stabilizes the ground by pre- BS: 6031 1981. Code of Practice for Earthworks. British
venting significant rise in porewater pressure. Com- Standard Institution.
pared with other types of engineering works, it is a Calladine, C.R. 2000. Plasticity for Engineers: Theory and
very cost-effective stabilization measures, especially Applications. Horwood Publishing Limited.
at large site. Different types of drains have been devel- Cheuk, C.T., Ng, C.W.W. & Sun, H.W. 2001. Numerical
oped and they serve different purposes. Trench drains analysis of soil nails in loose fill slopes. Proc. 14th
Southeast Asian Geot. Conf., Hong Kong, 1: 725–730.
and horizontal drains are more commonly used than Cheung, W.M. 2003. Non-Destructive Tests for Determin-
other types of drains, and simple design charts have ing the Lengths of Installed Steel Soil Nail (GEO Report
been developed to facilitate designers to assess the No. 133). Geotechnical Engineering Office, Hong Kong.
effects of the drains on pore pressures. CIRIA 2005. Soil Nailing—Best Practice Guidance (CIRIA
C637). CIRIA, UK.
Clouterre 1991. French National Research Project
Clouterre—Recommendations Clouterre. (English Trans-
ACKNOWLEDGEMENT lation 1993). Federal Highway Administration, FHWA-
SA-93-026, Washington, USA.
This paper is published with the permission of the D’Acunto, B. & Urciuoli, G. 2006. Groundwater regime in
a slope stabilised by drain trenches. Mathematical and
Head of the Geotechnical Engineering Office and the Computer Modelling. Pergamon-Elsevier Science LTD,
Director of Civil Engineering and Development, Gov- 43(7–8): 754–765.
ernment of the Hong Kong Special Administrative D’Acunto, B., Parente, F. & Urciuoli, G. 2007. Numerical
Region. models for 2D free boundary analysis of groundwater
The part of this paper on soil nailing technology in slopes stabilized by drain trenches. Computers &
was based on the review and research work carried Mathematics with applications, 53(4): 1615–1626.
out in the Geotechnical Engineering Office in Hong Davies, M.C.R. & Le Masurier, J.W. 1997. Soil/nail interac-
Kong. The soil-nail related studies was mainly led by tion mechanism from large direct shear tests. Proc. Third
the first author with contributions made by many col- Int. Conf. on Ground Improvement GeoSystems, London,
493–499.
leagues, particularly Mr Y.K. Shiu, Dr W.M. Cheung, Dawson, E.M., Roth, W.H. & Drescher, A. 1999. Slope sta-
Dr G.W.K. Chang and Dr D.O.K. Lo. All contributions bility analysis by strength reduction. Géotechnique 49(6):
are gratefully acknowledged. 835–840.
The part of this paper on drainage was based on Department of Transport 1994. Design Manual for Roads
the research work carried out in the Department of and Bridges: Design Methods for the Reinforcement of

122
 Highway Slopes by Reinforced Soil and Soil Nailing Halcrow China Limited 2007. Study on the Potential Effect
Techniques, HA68/94, Department of Transport, UK. of Blockage of Subsurface Drainage by Soil Nailing
Department of Transport 2001. BD 12/01—Design of Cor- Works—Study Report (GEO Report No. 218). Geotech-
rugated Steel Buried Structures with Spans Greater than nical Engineering Office, Hong Kong.
0.9 Metres and up to 8 Metres. (Design Manual for Road Hausmann, M.R. 1992. Slope remediation. (Invited lecture).
and Bridges). Department of Transport, UK. Proc. Specialty Conf. on Stability and Performance of
Desideri, A., Miliziano, S. & Rampello, S. 1997. Drenaggi Slopes and Embankments—II, Berkeley, Geotechnical
a Gravità per la Stabilizzazione dei Pendii. Hevelius Special Publication No. 31, American Society of Civil
Edizioni, Benevento. Engineers, 2: 1274–1317.
D’Esposito, N. 2007. Analisi 3D Dell’efficienza diTtrincee Hayashi, S., Ochiai, H., Yoshimoto, A., Sato, K. &
Drenanti Utilizzate per la Stabilizzazione dei Pendii. Kitamura, T. 1988. Functions and effects of reinforcing
Graduate thesis, University of Naples Federico II. materials in earth reinforcement. Proc. Int. Geot. Sym. on
Dowding, C.H. & Huang, F.C. 1994. Early detection of rock Theory and Practice of Earth Reinforcement, (IS Kyushu
movement with time domain reflectometry. Journal of ’88), Fukuoka, 99–104. Balkema, Rotterdam.
Geotechnical Engineering 120: 1413–1427. Hewlett Packard 1998. Time Domain Reflectometry The-
Duncan, J.M. 1996. State of the art: limit equilibrium and ory (Application Note 1304-2). Hewlett Packard Com-
finite element analysis of slopes. Journal of Geotechnical pany, USA.
Engineering ASCE 122(1): 577–596. HKIE 2003. Soil Nails in Loose Fill Slopes—A Preliminary
Ehrlich, M., Almeida, M.S.S. & Lima, A.M. 1996. Para- Study. Hong Kong Institution of Engineers Geotechnical
metric Numerical Analyses of Soil Nailing Systems. Division, Hong Kong.
In H. Ochiai, N. Yasufuku & K. Omine (eds.), Earth Ho, K.K.S. 2004. Keynote paper: Recent advances in
Reinforcement 747–752. Balkema, Rotterdam. geotechnology for slope stabilization and landslide mit-
Eisbacher, G.H. & Clague, J.J. 1984. Destructive mass move- igation—perspective from Hong Kong. Proc. Ninth Int.
ments in high mountains: hazard and management. Geol. Sym. on Landslides, Rio de Janeiro, 2: 1507–1560.
Surv. Can., Pap. 84–16. 230 p. Holtz, R.D. & Schuster, R.L. 1996. Stabilization of soil
Eyre, D. & Lewis, D.A. 1987. Soil Corrosivity Assess- slopes. In A.K. Turner & R.L. Schuster (eds.), Landslides
ment. Contractor Report 54. Transport and Road Research Investigation and Mitigation 439–473. Transportation
Laboratory, Department of Transport, UK. Research Board, Special Report 247, National Research
Fell, R. 1994. Stabilization of soil and rock slopes. Proc. of Council, National Academy Press, Washington.
East Asia Sym. and Field Workshop on Landslides and Hungr, O., Morgan, G.C. & Kellerhals, R. 1984. Quantitative
Debris Flows, Seoul, 1: 7–74. analysis of debris torrent hazards for design of remedial
FHWA 1998. Manual for Design & Construction monitor- measures. Can. Geotech. J. 21: 663–677.
ing of Soil Nail Walls. Federal Highway Publication No. Hungr, O., Morgan, G.C., VanDine, D.F. & Lister, D.R. 1987.
SA-96-069R. U.S. Department of Transportation, Federal Debris flow defences in British Columbia. In J.E. Costa &
Highway Administration, Washington, D.C. G.F. Wieczorek (eds), Debris flows/avalanches: pro-
FHWA 2003. Geotechnical Engineering Circular No. 7: cess, recognition and mitigation. Reviews in Engineering
Soil Nail Walls, Report No. FHWA0-IF-03-017, Federal Geology. Geol. Soc. Am., 7: 201–222.
Highway Administration, Washington, USA. Hutchinson, J.N. 1977. Assessment of the effectiveness of
Gässler, G. 1997. Design of reinforced excavations and nat- corrective measures in relation to geological conditions
ural slopes using new European Codes. In H. Ochiai, and types of slope movement (General Report). Bulletin of
N. Yasufuku & K. Omine (eds.), Earth reinforcement the Int. Association of Engineering Geology 16: 131–155.
943–961. Balkema, Rotterdam. Itasca 1996. Fast Lagrangian Analysis of Continua (FLAC)
Gässler, G. & Gudehus, G. 1981. Soil nailing—some aspects Manual, Version 4.0. Itasca Consulting Group, Inc.,
of a new technique. Proc. Tenth Int. Conf. Soil Mechanics Minnesota.
and Foundation Engineering, Stockholm, 3: 665–670. IStructE 1999. Interim Guidance on the Design of Reinforced
GEO 2000. Technical Guidelines on Landscape Treatment Concrete Structures Using Fibre Composite Reinforce-
and Bio-engineering for Man-made Slopes and Retain- ment. The Institution of Structural Engineers, UK.
ing Walls (GEO Publication No. 1/2000). Geotechnical Japan Highway Public Corporation (JHPC) 1998. Design &
Engineering Office, Hong Kong. Works Outlines on the Soil-Cutting Reinforcement Soil-
GEO 2008. Guide to Soil Nail Design and Construc- works. (English Translation). Japan Highway Public
tion (Geoguide 7). Geotechnical Engineering Office, Corporation.
Hong Kong. Jewell, R.A. 1980. Some Effects of Reinforcement on the
Gigan, J.P. & Delams, P. 1987. Mobilisation of stresses in Mechanical Behaviour of Soils. PhD thesis, University
nailed structures. English translation. Transport and Road of Cambridge.
Research Laboratory, Contractor Report 25. Jewell, R.A. & Pedley, M.J. 1990. Soil nailing design: the role
Ground Engineering 2004. Soil-nailing—called to the bar. of bending stiffness. Ground Engineering, March, 30–36.
Ground Engineering, May, 14. Jewell, R.A. & Pedley, M.J. 1992. Analysis for soil rein-
Gutierrez, V. & Tatsuoka, F. 1988. Roles of facing in reinforc- forcement with bending stiffness. Journal of Geotechnical
ing cohesionless soil slopes by means of metal strips. Proc. Engineering, ASCE, 118(4): 1505–1528.
Int. Geot. Sym. on Theory and Practice of Earth Rein- Japanese Ministry of Construction 1997. Design guidelines
forcement, (IS Kyushu ’88), Fukuoka, 289–294. Balkema, of FRP reinforced concrete building structures. Journal
Rotterdam. of Composites for Construction, August, 90–115.

123
JSCE 1997. Recommendations for Design and Construction Ng, C.W.W. & Chiu, A.C.F. 2003. Laboratory study of
of Concrete Structures Using Continuous Fiber Reinforc- loose saturated and unsaturated decomposed granitic
ing Materials. Concrete Engineering Series, Japan society soil. Journal of Geotechnical and Geoenvironmental
of Civil Engineers, No. 23. Engineering, ASCE 129(6): 550–559.
Kellerhas, R. 1970. Runoff routing through steep natu- Ng, C.W.W., Pun, W.K., Kwok, S.S.K., Cheuk, C.Y. &
ral channels. ASCE Journal of Hydraulics Division, 96: Lee, D.D.M. 2007. Centrifuge modelling in engineer-
2201–2217. ing practice in Hong Kong. Geotechnical Advancements
Kenney, T.C. & Lau, K.C. 1984. Temporal changes of in Hong Kong since 1970s, Proc. HKIE Geotechnical
groundwater pressure in a natural slope of non fissured Division Annual Seminar, Hong Kong, 55–68.
clay. Can. Geotech. J. 21(1): 138–146. O’Connor, K.M. & Dowding, C.H. 1999. GeoMeasurements
Kenney, T.C., Pazin, M. & Choi, W.S. 1977. Design of hori- by Pulsing TDR Cables and Probes. CRC Press.
zontal drains for soil slopes. ASCE Journal of Geotechni- Palmeira, M. & Milligan, G.W.E. 1989. Large scale direct
cal Engineering Division, 103(GT11): 1311–1323. shear tests on reinforced soil. Soils and Foundations,
Krahn, J. 2003. The 2001 R.M. Hardy Lecture: The lim- 29(1): 18–30.
its of limit equilibrium analyses. Canadian Geotechnical Pang, L.S. & Wong, H.N. 1997. Prescriptive design of soil
Journal 40: 643–660. nails to upgrade soil cut slopes. In K.S. Li, J.N. Kay &
Law, K.T., Shen, J.M. & Lee, C.F. 1998. Strength of a loose K.K.S. Ho (eds.), Slope Engineering in Hong Kong, Proc.
remoulded granitic soil. In K.S. Li, J.N. Kay & K.K.S. Ho HKIE Annual Seminar, Hong Kong, 259–266.
(eds.), Slope Engineering in Hong Kong, Proc. HKIE Pedley, M.J. 1990. The Performance of Soil Reinforcement in
Annual Seminar, Hong Kong, 169–176. Balkema. Bending and Shear. PhD thesis, University of Oxford.
Lee, C.F. & OAP 2007. Review of Use of Non-destructive Pellegrino, A., Ramondini, M. & Urciuoli, G. 2004a. Regime
Testing in Quality Control in Soil Nailing Works (GEO delle pressioni neutre in un pendio in Argille Varicolori
Report No. 219). Geotechnical Engineering Office, stabilizzato con trincee drenanti. International Workshop:
Hong Kong. ‘‘Living with landslides: effects on structures and urban
Lin, C.P. & Tang, S.H. 2005. Development and calibra- settlements. Strategies for risk reduction’’, 141–149.
tion of a TDR extensometer for geotechnical monitoring. Anacapri.
Geotechnical Testing Journal 28(5): 464–471. Pellegrino, A., Ramondini, M. & Urciuoli, G. 2004b.
Liu, W., Hunsperger, R., Chajes, M., Folliard, K. & Interplay between the morphology and mechanics of
Kunz, E. 2002. Corrosion detection of steel cables using mudslides: field experiences from Southern Italy. IX Inter-
time domain reflectometry. Journal of Materials in Civil national Symposium on Landslides, 2: 1403–1409. Rio de
Engineering, ASCE, 14(3): 217–223. Janeiro.
Leventhal, A.R. & Mostyn, G.R. 1987. Slope stabilization Perry, J., Pedley, M.J. & Reid, M. 2003. Infrastruc-
techniques and their application. In Slope Stability and ture embankments- condition appraisal and remedial
Stabilisation, 183–230. Balkema, Rotterdam. treatment. (Second edition). CIRIA Report C592.
Lui, B.L.S. & Shiu, Y.K. 2004. Prescriptive soil nail design CIRIA, UK.
for concrete and masonry retaining walls. Proc. the HKIE Plumelle, C., Schlosser, F., Delage, P. & Knochenmus, G.
Geotechnical Division Annual Seminar 2004—Recent 1990. French National Research Project on Soil Nailing.
Advances in Geotechnical Engineering, Hong Kong, Proc. Conf. of Design and Performance of Earth Retaining
185–197. Structures, ASCE Geotechnical Special Publication No.
Lui, B.L.S. & Shiu, Y.K. 2005. Prescriptive Soil Nail 25, 18–21 June, 660–675.
Design for Concrete and Masonry Retaining Walls (GEO Plumelle, C. & Schlosser, F. 1990. A French National
Report No. 165). Geotechnical Engineering Office, Research Project on Soil Nailing: Clouterre. In
Hong Kong. A. McGown, K.C. Yeo & K.Z. Andrawes (eds.), Per-
Marchal, J. 1986. Soil Nail—Experimental Laboratory Study formance of Reinforced Soil Structures 219–223. British
of Soil Nail Interaction. (English Translation). Transport Geotechnical Society.
and Road Research Laboratory, Department of Transport, Pun, W.K., Cheung, W.M., Lo, D.O.K. & Cheng, P.F.K. 2007.
Contractor Report No. 239. Application of time domain reflectometry for quality con-
Marino, R. 2007. Analisi 3D Dell’efficienza di Aste Drenanti trol of soil nailing works. Proc. Int. Forum on Landslide
per la Stabilizzazione dei Pendii. Graduate thesis, Univer- Disaster Management, Hong Kong (in press).
sity of Naples Federico II. Pun, W.K., Pang, P.L.R. & Li, K.S. 2000. Recent devel-
Mears, A.I. 1981. Design Criteria for Avalanche Control opment in prescriptive measures for slope improvement
Structures in the Runout Zone. United States Department works. Proc. Sym. on Slope Hazards and Their Preven-
of Agricolture, Forest Service General Technical Report tion, Hong Kong, 303–308.
RM-84. Pun, W.K. & Shiu, Y.K. 2007. Design practice and tech-
Muramatsu, M., Nagura, K., Sueoka, T., Suami, K. & nical developments of soil nailing in Hong Kong. Proc.
Kitamura, T. 1992. Stability analysis for reinforced cut HKIE Geotechnical Division Annual Seminar, May, Hong
slopes with facing. Proc. Int. Sym. on Earth Reinforce- Kong, ??.
ment Practice, Fukuoka, Kyushu, Japan, 503–508. Shanmuganathan, S. 2003. Fibre reinforced polymer com-
Murray, R.T. 1993. The Development of Specifications for posite materials for civil and building structures—review
Soil Nailing. Research Report 280, Transport Research of the state-of-the-art. The Structural Engineer, July,
Laboratory, Department of Transport, UK. 26–33.

124
Schlosser, F. 1982. Behaviour and design of soil nail- Topp, G.C., Davis, J.L. & Annan, A.P. 1980. Electromag-
ing. Proc. of Sym. on Recent Developments in Ground netic determination of soil water content: measurement
Improvements, Bangkok, 399–413. in coaxial transmission lines. Water Resources Research
Schlosser, F. & Guilloux, A. 1981. Le forttement dans 16(3): 574–582.
les sols. Revue Francaise de Gèotechnique, 16: 65–77. Thurber Consultants Ltd. 1983. Debris Torrent and Flooding
(in French) Hazards, Highway 99, Howe Sound. Report to B.C. Min.
Schuster, R.L. 1992. Recent advances in slope stabilization. Transportation and Highways, Victoria, B.C.
(Keynote paper). Proc. of Sixth Int. Sym. on Landslides, Unwin, H. 2001. Carbon fibre soil nailing for railway
Christchurch, 3: 1715–1745. embankments. Proc. Int. Exhibition and Sym. of Under-
Shiu, Y.K. & Cheung, W.M. 2003. Long-term Durability ground Construction, London, 697–706.
of Steel Soil Nails (GEO Report No. 135). Geotechnical Urciuoli, G .1998. Pore pressures in unstable slopes con-
Engineering Office, Hong Kong. stituted by fissured clay shales. 2nd Int. Symp. on The
Shiu, Y.K. & Chang, G..W.K. 2004. Soil Nail Head Review Geotechnics of Hard Soils—Soft Rocks, 2: 1177–1185.
(GEO Report No. 175). Geotechnical Engineering Office, Napoli.
Hong Kong. Van Dine D.F. 1996. Debris Flow Control Structures and
Shiu, Y.K. & Chang, G.W.K. 2005. Effects of Inclination, Forest Engineering. Working paper, Ministry of Forests,
Length Pattern and Bending Stiffness of Soil Nails on British Columbia.
Behaviour of Nailed Structures (GEO Report No. 197). Veder, C. 1981. Landslides and their stabilization. Springer-
Geotechnical Engineering Office, Hong Kong. Verlag, New York.
Siddiqui, S.I., Drnevich, V.P. & Deschamps, R.J. 2000. Time Walker, B.F. & Mohen, F.J. 1987. Ground water prediction
domain reflectometry for use in geotechnical engineering. and control and negative pore water pressure. In Walker
Geotechnical Testing Journal, ASTM, 23(1): 9–20. & Fell (eds.), Soil Slope Instability and Stabilisation,
Smith, I.M. & Su, N. 1997. Three-dimensional FE analysis 121–181.
of a nailed wall curved in plan. International Journal for Watkins, A.T. & Powell, G.E. 1992. Soil nailing to existing
Numerical and Analytical Methods in Geomechanics 21: slopes as landslip preventive works. Hong Kong Engineer,
583–597. March, 20–27.
Stocker, M.F. & Riedinger, G. 1990. The bearing behaviour Wong, H.N., Pang, L.S., A.C.W., Pun, W.K. & Yu, Y.F.
of nailed retaining wall. Proceedings of Design and 1999. Application of Prescriptive Measures to Slopes and
Performance of Earth Structure, Geotechnical Special Retaining Walls (GEO Report No. 56). (Second Edition).
Publication No. 25, ASCE, 613–628. Geotechnical Engineering Office, Hong Kong.
Tan, S.A., Luo, S.Q. & Yong, K.Y. 2000. Simplified mod- Yin, J.H. & Su, L.J. 2006. An innovative laboratory box
els for soil-nail lateral interaction. Ground Improvement, for testing nail pull-out resistance in soil. Geotechnical
Thomas Telford, 4(4): 141–152. Testing Journal 29(6): 451–461.
Tei, K., Tayor, R.N. & Milligan, G.W.E. 1998. Centrifuge Zaruba, Q. & Mencl, V. 1982. Landslides and their control.
model tests of nailed soil slopes. Soils and Foundations New York, Elsevier.
38(2): 165–177.
Thompson, S.R. & Miller, R. 1990. Design, construction and
performance of a soil nailed wall in Seattle, Washington.
Proc. Conf. on Design and Performance of Earth Retain-
ing Structures, Geotechnical Special Publication No. 25,
June 18–21, 1990, New York, 629–643.

125
Appendix A. Example of design charts.

126
Special lectures
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Loess in China and landslides in loess slopes

Z.G. Lin & Z.J. Xu

Northwest Research Institute of Engineering Investigations and Design, Xian, China

M.S. Zhang
Xian Geological and Mineral Resources Research Institute, Xian, China

ABSTRACT: Loess in china is outstanding for its stratigraphical intactness, huge thickness, vast expanse of
distribution and metastable geotechnical properties. During the Quaternary Period the arid, semi-arid climatic
environment provided the stage on which material sources came into being, grain particles were wind-transported
and deposited and loess strata were thus formed. The paper puts emphasis on the granulometry and engineering
properties peculiar to loessial deposits as well as the role they might play in the initiation and evolution of loess
landslides. From this, the paper proceeds to such problems as geomorphological zonation, landslide distribution,
morphological and structural features of loess landslides as well as landslide classifications, giving due attention
to seismic landslides. This is followed by a summing-up of the methods of landslide prevention and remediation.
Finally, the issues that need more attention and further in-depth research are raised.

1 INTRODUCTION experiences accumulated up to now so as to contribute

to the development of new research programs and new
Loess covers large and wide parts of Northwest, North prevention and remediation methods.
and Northeast China, its distribution being concen-
trated on the so-called Central Loess Plateau which
comprises Shaanxi, Gansu and Shanxi provinces as 2 DISTRIBUTION OF LOESS IN CHINA
well as most of Ningxia Autonomous Region. Here AND ITS ENGINEERING PROPERTIES
the loess deposit formed in the unique Quaternary
climatic environment is outstanding for its stratigraph- 2.1 Distribution of loess in China
ical intactness, sustained distribution and huge total Loess in China is mainly distributed in Gansu,
thickness. The granulometric composition of loess is Shaanxi, Shanxi, Henan and Qinghai provinces, the
remarkable for stableness, maintaining in the mean- Ningxia Autonomous Region and the Inner Mongolian
while a well-oriented gradual change from Northwest Autonomous Region, blanketing the Central Loess
to Southeast. In engineering properties the soil is char- Plateau and its neighboring areas. Next in distribu-
acterized by its state of consolidation (underconsoli- tion are Hebei, Shandong, Liaoning and Heilongjiang
dated in the upper part and slightly overconsolidated in provinces, and the Xinjiang Autonomous Region .The
its lower part). These properties are likewise remark- land surface covered by loess in China is estimated to
able for their regular and oriented change in the general be 6.3 × 105 km2 which is equal to about 6.6% of the
direction from Northwest to Southeast all across the total land area of the country. See Figure 1.
Loess Plateau.
In such climatic, sedimentation and stratigraphical
2.2 Stratigraphical features
environment conditions, geomorphological evolution
has followed correspondingly unique paths resulting in Stratigraphical intactness and huge total thickness are
the formation of loess landslides peculiar to such con- prominent features of loess in China, owing to which
ditions. In the push for the development of the West, it almost complete records of the geographical, climatic,
is important to have a correct understanding of loess depositional and biological changes and evolution over
landslides and use proper methods of investigation, the entire Quaternary Period have been preserved.
prevention and remediation. This paper tries to give a What is shown in Figure 2 is a typical stratigraphical
brief summing-up of the research results achieved and profile of the loess in China.

129
Figure 1. Distribution of loess in China.

Figure 3. Granulometric zoning of the loess plateau.

Figure 2. A typical sketch profile of loess in China. Figure 4. Engineering geological zonation of the loess of
China.
2.3 Granulometry
Throughout the Quaternary Period the granulomet- from its overburden weight when saturated and that
ric composition kept stable. However against this which collapses under the combined pressure from its
background a gradual change from coarse to fine in own overburden and additional loads (fills, founda-
the general direction from NW to SE is discernible tions, embankments etc) when saturated. In Figure 4
(Fig. 3). This stableness in composition is reflective of is shown the engineering geological zonation of the
the geographical, climatic and depositional environ- loess in China and in Table 1 are given the soils labo-
ment conditions of that period and evidences the eolian ratory test results (mean values of physico-mechanical
origin of loessial deposits. Of course the existence of indices) for Zone I, Zone II and Zones I, II, III, IV,
the numerous reddish brown paleo-soil layers bears V and VII as a whole. Figure 5 is a generalization
evidence of the climatic fluctuations in the generally of the afore-mentioned regularity of oriented grad-
stable environment. ual change (amelioration) in engineering properties
of loess in the direction from NW to SE all across
the Central Loess Plateau and its neighboring regions.
2.4 Engineering properties
And Figure 6 is a photo which tells of the grave con-
Of the engineering properties of loess four deserve sequences that could be brought about by flooding
special attention. They either play important roles in overburden-collapsible loessial terrains (Qian et al.
in the initiation and evolution of landslides or are 1988, Lin 1994, Wang et al. 1990).
contributive to the behavior of man-made slopes. In the meanwhile, it should be noted that in step
with increase of water content the soil sample invari-
2.4.1 Collapsibility ably tends to increase in compressibility with cor-
Loessial terrains (or construction sites) fall into two responding decrease in collapsibility, as shown in
types, i.e. the type which collapses under the pressure Figure 7.

130
 Table 1. Soils laboratory test results (mean values of physico-mechanical indices) for different zones.

Coef.of
Nat.wat. Void Liquid Plast. overburd. Init. Precons. Precons. Sand Slit Clay
cont. Unit wt. rat. lim. index collaps. pressure press.(nat.) press.(sat.) grains cont. cont.
Zone Age w(%) γ (kN·m−3 ) e wL (%) Ip /(%) δzs Psh (MPa) Pc (MPa) Pc (MPa) (%) (%) (%)

Gansu Q3 8.98 14.58 1.032 27.60 9.37 0.052 0.135 0.508 0.137 11.20 78.75 10.05
Prov. Q22 6.79 15.37 0.868 26.87 7.34 0.054 0.270 2.047 0.287 10.00 76.00 14.00
(Zone I) Q12 4.90 16.20 0.742 26.40 8.90 0.051 1.000 3.800 1.040 9.00 77.00 14.00
North. Q3 13.38 14.70 1.105 29.9 11.2 0.023 0.092 0.835 0.097 13.20 65.80 21.00
Shaanxi and Q22 15.30 16.40 0.924 30.9 12.2 0.014 0.416 1.224 0.473 15.00 68.10 16.90
Longdong Q12 18.10 18.30 0.760 30.6 12.3 0.008 0.684 1.680 12.00 64.00 24.00
(Zone II)) Q1 20.10 18.50 0.773 31.1 11.7 0.003 1.856 1.615 13.00 70.00 17.00
Six Zones Q3 13.21 15.26 1.016 28.62 10.86 0.015 0.137 0.580 0.158 10.96 70.70 18.34
(Zone I-V) Q22 12.85 16.26 0.884 29.11 10.79 0.018 0.419 1.380 0.453 15.19 68.62 16.19
and Zone Q12 13.09 17.23 0.783 29.01 11.25 0.026 0.981 2.148 1.036 12.25 69.87 17.88
VII) Q1 20.10 18.55 0.773 31.05 11.65 0.003 1.855 1.615 13.00 70.00 17.00

131
Figure 6.
Figure 5.
lapse pressure-overburden pressure relationships.

in and around test pit caused by flooding (Q4 and Q3 loess).

Characteristic generalization of the initial col-

General view and close-up of ground subsidence

 Figure 9. Normalization is feasible under condition σ3 ≥
Figure 7. Interrelationship between the coefficient of col- 75 kPa.
lapsibility, δs , and the coefficient of compressibility, a, as a
function of the degree of saturation, Sr .

Figure 10. Stress-strain curves of soil samples from Xian

(depth ≤ 10 m).

Figure 8. K0 becomes constant when consolidation pressure

σ3 ≥ σ3cr .

2.4.2 Structurality and structural strength

Loess is outstanding for its structurality which is due
to its environment of deposition and the diagenetic
process it underwent, its micro-structural features, the
existence of soluble salts and the soil matrix suction.
As shown in Figure 8 the coefficient of lateral pressure
K0 turns constant only after the consolidation pressure
reaches and exceeds σ3cr . Figure 9 shows that nor-
malization by P(=(σ1 + 2σ3 )/2) of the (σ1 − σ3 )/P
vs axial strain ε1 curves of the soil sample is fea-
sible only after the consolidation pressure exceeds
75 kPa. Figure 10 and Figure 11 respectively depict
the q(=(σ1 − σ3 )/2) vs ε1 relationships of loess sam-
ples from two different depths and consolidated under Figure 11. Stress-strain curves of soil samples from Xian
different pressures σ3 or, in other words, the soil (depth: 13.2–17.4 m).

132
 suction-dependent strength becomes very low or even
vanishes when the soil water content is over 25%
or degree of saturation exceeds 65%. Figure 12
(Xing 2001) gives an example depicting this soil-water
characteristic.

2.4.4 Shear strength of sliding bed soil types

In addition to its suction-dependency and like many
other soils, the shear strength of loess is also remark-
able for the large drop from peak to residual parameter
values. Table 2 gives the classification of sliding
bed soil types in loess slides and in Table 3 are
Figure 12. Soil-water characteristic curve of a loess (Q3 ) given the geotechnical properties of those soil types
sample from Yangling, Xi’an, Shaanxi Province. (Wang 2005).

Table 2. Classification of sliding bed soil types in loess

slides. 3 DISTRIBUTION AND TYPES OF LOESS
LANDSLIDES
Physical properties
Sliding bed soil type and age of bed soil
3.1 Loess landslides as a peculiar
Malan Loess (Q3 ) geomorphological element of the Loess Plateau
Silty bed soil
Figure 13 (Derbyshire et al 1994) is a schematic repre-
Loess Lishi Loess (Q2 ) sentation of the geomorphological divisions (zoning)
of the Central Loess Plateau. Each division has its
Wucheng Loess (Q1 ) Clayey bed soil characteristic and representative landform element or
elements such as Yuan (platforms or residual plat-
Mixtures of loess and forms) as shown in Figure 14, Liang (ridges) as shown
underlying bedrock material (Q4 ) in Figure 15, Mao (mounds, hills) also shown in
Figure 15, fluvial plains and fluvial terraces as shown
Weak intercalations in bedrock Weak rock-turned in Figure 16. Loess landslides as peculiar landform
bed soil elements are scattered among them.

3.2 Distribution of loess landslides

structurality vs soil age relationships (Qian et al. 1988, Figure 17 (Jin et al. 1996) is a schematic map of
Lin 1994, Wang et al. 1990). landslide distribution in the Loess Plateau and its
Structural strength of loess invariably weakens as contiguous areas, on which the clusterly nature of
water content increases until the state of saturation is landslide distribution is discernible. Each large cluster
reached. In this state the structural strength of loess has its combination of natural conditions and slide-
can be termed the ‘‘residual structural strength’’ or, causative factors. For examples, the NS trending
usually in subsoil treatment and foundation design in lengthy cluster near the right side margin of the map is
China, the ‘‘initial (collapse) pressure’’. It is to be just where the Yellow River gorge and the west slope of
pointed out that for loess the terms ‘‘residual structural the Lüliang mountain range are; the linear assemblage
strength’’, initial (collapse) pressure’’ and ‘‘preconsol- at the lower middle of the map is representative of the
idation pressure’’ are technical synonyms, having the 180 or more old landslides which are closely spaced
same numerical value. one after the other and extend eastward from Baoji
along the north shore the Wei River; and the clusters
2.4.3 Suction or matrix suction of landslides west of the Lüliang Range on the map
As a kind of unsaturated clayey soils and like swelling came into being during the 1920 M = 8.5 Haiyuan
clays, loess distinguishes itself by particularly high Earthquake.
suction when it is in a state of low moisture con- Landslides and their distribution can be located and
tent. This suction in loess manifests itself as a kind delineated with precision by means of satellite images
of negative pore pressure effect and enables the soil and air photos. The successful use of infrared color
to manifest high strength which however decreases imagery for Xi-Yu expressway is an example, as shown
in step with increase of water content. Usually, this in Figure 18.

133
 Table 3. Geotechnical properties of different sliding bed soil.

Geotechnical properties

Strength parameters

Cohesion Angle of int.

Coef. of c(kPa) friction ϕ(◦ )
Permeability collaps.
Sliding bed soil type k(cm · s−1 ) δs Peak Residual Peak Residual

Silty bed soil (3–7) × 10−4 0.01–0.08 15–30 3–15 12–22 8–12
Clayey bed soil (5–6) × 10−6 0 20–130 5–25 10–20 2.5–8
Weak rock-turned bed soil <6 × 10−6 0 50–200 5–25 10–20 2.5–8

Figure 13. Geomorphological divisions of the Loess

Plateau of China.

Figure 15. Liang (ridges) and Mao (hilly mounds), Yanan,

Shaanxi Province.

Figure 16. Loess terraces, Malianhe Valley, Gansu

Figure 14. Part of Xifengyuan, Ningxian, Gansu Province. Province.

the sliding plane confines itself within the loessial

3.3 Structural types of loess landslides
soil mass and sliding takes place following an internal
Based on the strata involved in sliding and the struc- weak zone or weak layer; loess-bedrock surface con-
tural features of the slide, loess slides can be viewed as tact slides in which the sliding bed and the soil-bedrock
falling into three types, i.e. intra-loess slides in which interface coincide; and bedrock slides which again

134
Figure 17. Sketch map of loess slide distribution in the Loess Plateau.

Figure 18. Infrared color converted to color image of Xi-Yu expressway.

can be divided into two subtypes, i.e. slides whose

sliding bed is bedrock-conformative and slides whose
sliding bed is bedrock-nonconformative.

3.3.1 Intra-loess landslides

What Figure 19 (Wang et al. 2005) shows is an intra-
loess slide very common in the Loess Plateau. As
shown, the sliding plane (or bed) is at the top of a
paleo-soil layer where perched water existed. The non-
conformative contact between the ‘‘new loess’’ (Q3 )
and ‘‘old loess’’ (Q2 ) shown in Figure 20 is also a
causative factor for intra-loess slides. It is to be pointed
out that the volume of slide débris involved is usually Figure 19. Profile of Changqingqiao landslide, Xifeng,
less than 100,000 m3 . Gansu Province.

135
 four kinds of causative factors can be summed-up as
follows:

4.1.1 Steeping of loess banks by downward erosion

in gullies and intensified lateral erosion of
loess platforms and terraces by river streams
This kind of causative factors is very common. They
are related to regional periodical deposition-erosion
changes. See Figure 24.

4.1.2 Sustained rainfalls, rainstorms and frequent

irrigation involving large areas
of loessial terrain
Here, the former two are natural processes, while
Figure 20. Loess profile, Zhangzigou, Qinghuabian, the latter is human. These processes inevitably lead
Yan’an, Shaanxi Province. to increased water content in the loess strata and
lowered stability of loess slopes and are therefore
slide-causative. Figure 25 shows the Jiangliu landslide
on the south bank of the Jing River as an exam-
ple. That slide was an old one. It was reactivated
on a large scale at the end of the rainy season of
1984. During that year rainfalls were much more
abundant than usual and the rainy season was also
much longer. In consequence, more landslides took
place in Shaanxi Province during that year and the
following one.
The phenomena mentioned above can also be
expounded to a certain degree in the context of the
relationship between the strength and matrix suction
Figure 21. A loess landslide in Tongchuan, Shaanxi of unsaturated soils. In Figure 26 is shown the progres-
Province. sive failure of a ‘‘typical loess cut slope’’ (Wang et al.
2005). In pace with water content increase, the matrix
suction decreases. The corresponding effect is lowered
3.3.2 Loess-bedrock surface contact landsides strength and weakened stability. If this process goes on
Shown in Figure 21 and Figure 22 are loess-bedrock unremedied, the slope will fail and eventually turn into
surface contact slides, also common in the Loess a slide, as shown in Figure 27.
Plateau, The sliding plane (or bed) is at the surface of
contact between loess and the underlying bedrocks of
4.1.3 Human activities
the Tertiary, Jurassic or Triassic periods. In such slides
In step with economic development in the Loess
the volume of slide débris involved is mostly larger
Plateau and the contiguous regions, landslides caused
than 100,000 m3 , exceeding 1,000,000 m3 sometimes.
or triggered by irrigation, highway and railway con-
3.3.3 Loess-bedrock landslides struction, mining and underground space development
What is shown in Figure 23 is a loess-bedrock slide. activities are becoming more frequent. This has raised
It can be seen that the sliding plane passed all the way serious concern.
through the loess strata, incised at a certain dip angle
into the underlying Tertiary strata and finally levelled 4.1.4 Seismic landslides
off in conformative, near-horizontal sliding. This problem will be discussed in Section 5.

4 EVOLUTION MECHANISM OF LOESS 4.2 Evolution track of loess landslides

LANDSLIDES
Except for slides caused by earthquakes, it usually
takes time, sometimes long time for slides to evolve
4.1 Causative factors of loess landslides
from initiation to general sliding. In other words, the
From the above-mentioned pattern of landslide distri- process is progressive. An incisive summing-up is
bution and geologic structural features of loess slides, given in Figure 28.

136
Figure 22. Profile of a landslide at K10 Wubu, Northern Shaanxi.

Figure 23. Longitudinal section of Pojishan landslide, Baoji, Shaanxi Province.

Figure 25. Jiangliu landslide on south bank of the Jing

River, Guanzhong Basin.
Figure 24. Wangjiaao landslide, Tongchuan, Shaanxi
Province.
highly seismic regions where at least six earthquakes
of magnitude 8 and larger took place in history.

5 LOESS LANDSLIDES IN HIGHLY

5.1 Distribution of earthquake landslides
SEISMIC REGIONS
Figure 29 (Lin et al 2000) shows the distribu-
The east, south and west parts of the Loess Plateau tion of landslides caused by historical earthquakes
and some neighboring areas covered by loess are (1500∼1949). Please note the lower middle left of

137
 are characterized by very small sliding bed angles and
unusually long slide distances like Huihuichuan slide
shown in Figure 30. A lot of such slides are very large
and characterized by panoramic wavy terrain as can be
seen in Figure 31. And what Figure 32 shows is part of
the results obtained in studying the stratigraphic struc-
tures before and after sliding uncovered by trenching.
These results bear witness to the presence of soil (sand)
liquefaction.
It is to be noted that soil (sand) liquefaction dur-
ing strong earthquakes is not exclusive. Often, sudden
Figure 26. Progressive failure mode for a ‘‘typical loess heavy loading on alluvial deposits by a falling or down-
slope’’. sliding soil mass can cause instantaneous excessive
pore water pressure build-up in the alluvial deposits
and result in liquefaction of the underlying alluvium
or the soil (sand) layers in the vicinity of the slid-
ing bed as demonstrated in Figure 33 and Figure 34.
Because of greatly reduced resistance the slide-mass
can not but drift a much longer distance before full
energy dissipation. The 1983 Sale mountain landslide
in Gansu Province (Fig. 35) and the old Pojishan land-
slide in Shaanxi Province (Fig. 23) are examples of
such events.

6 LANDSLIDE PREVENTION
AND REMEDIATION

6.1 Measures and methods commonly used

for landslide prevention and remediation
Figure 27. In the slope the plastic state zone enlarges in in loessial regions
pace with water content increase (at Jiangzhang).
6.1.1 Modification of the slide’s geometrical
configuration
Examples of this approach are: improvement of slope
configuration like partial removal of slide débris with
a view to reducing the pushing force; back-loading
where the slope of the sliding bed turns reversed so as
to increase anti-sliding resistance.

6.1.2 Drainage
Examples of this approach are: use of a system of
open or covered ditches to drain rainwater out of the
slide area and the contiguous ground surfaces; use of
underground drainage ditches and underground slide-
resistant drainage ditches; drainage of the débris mass
Figure 28. Landslide evolution (after Wang Gongxian). by means of large diameter boreholes; drainage of the
débris mass by means of inclined boreholes; use of
Figure 29 where single large slides and slide clusters underground galleries inside as well as outside of the
that came into being during the 1920 M = 8.5 Haiyuan landslide.
earthquake are shown.
6.1.3 Retaining structures
Examples of this approach are: gravity retaining
5.2 Landslides in the course of whose initiation
walls including gabion baskets; anchored retaining
and evolution soil liquefaction played
structures; anti-sliding piles, pretensioned cable-tied
an inalienable part
anti-sliding piles; soil-nailed walls.
During the 1920 Haiyuan earthquake (M = 8.5) nu- The measures and methods enumerated above are
merous slides and slide clusters came into being. They often used in a joint manner.

138
Figure 29. Location and distribution of historical earthquake-induced landslides and slump failures (1500–1949).

Figure 30. Longitudinal section of Huihuichuan slide.

Figure 32. Original stratigraphic structure of the slide area

and evidence of thrust faulting and shearing as well as traces
of liquefaction.

6.2 Examples of remediation projects

6.2.1 Phoenix mountain slide, Yan’an
This slide is an old yet reactivated one. The reme-
dial efforts include drainage of rainwater, backloading
Figure 31. Wavy terrain characterizing low-angle long at the slide toe and anti-sliding piles with evident,
distance slide. satisfactory effects. See Figure 36 and Figure 37.

139
 6.2.2 Fengjiahe loess landslide,
Tongchuan-Huangling 1st class highway
Being a large, very thick old landslide, it is 300–440 m
long, 1300 m wide and 25–45 m thick, reaching 60 m
where the slide mass is most thick. Road cutting led to
the appearance of two 150–220 m long tensile cracks in
the upper reach of the slope which kept on widening.
The main design approach adopted for remediation
was drainage of the slide-mass, including 29 inclined
Figure 33. Schematic diagram of undrained loading upon drainage boreholes each 32.5 m long and inclined at
alluvial deposits by a moving landslide mass. an angle of elevation of 5◦ , an underground drain
320 m long and positioned along the up-slope side
of the road cut, unloading by soil removal at the top
of the slope with a rainwater interception ditch on the
cut platform, and a 5 m high anti-sliding wall at the
foot of the cut slope. See Figure 38 and Figure 39.
According to the results of surface movement moni-
toring and down-hole inclinometer measurements over
the past 6 years, the slide has proved to be stable (Wang
et al. 2002).

6.2.3 Mt. Lishan landslide, Lintong, Xi’an

Figure 40 gives a plane view of Mt. Lishan landslide.
On the right of the Figure is the landslide No. 2 which
Figure 34. Rapid loading test on a saturated alluvial deposit. is a set of retrogressive intra-loess or loess-bedrock
slides. A deep gully named Lao Ya gully separates
landslide No. 2 from landslide No.1 The latter falls
into 3 blocks of which Block I has been the focus of
research, monitoring and remediation.
One of the distinguishing features of the research
study of Mt. Lishan landslide is the completeness of
the set of instruments used and the extended time
span of monitoring. It should be pointed out that
the ‘‘full-length’’ extensometers and extensometers
of various lengths played important roles in delimi-
ting and delineating different deforming parts of the
Figure 35. Sale mountain slide, 1983 (location: Dongxiang, slide. Down-hole inclinometering was of great help in
Gansu Province; slide volume: 3100×104 m3 ; slide distance:
in excess of 1000 m; casualties: 220 deaths) (Wu et al. 2006).
determining the sliding bed. All such results formed
a reliable data base on which subsequent specific
remedial measures were worked out.
The target of remediation of the first phase was
Block I of slide NO. 1. The structural features of
Block I and measures taken as shown in Figure 41 (Lin
1997) are representative of that block. As can be seen,
Block I was taken as composed of three diffentiated
slides, i.e. slides A, B and C. For them the measures
taken were anti-sliding piles, cable-tied slide-resistant
walls and cable-tied anti-sliding piles respectively.

6.2.4 Pojishan landslide, Baoji, Shaanxi

It is a very large old landslide situated close to the
downtown district of the city of Baoji. To the rear of
the slide is the high fourth terrace of the Wei River
and to the front is the flood plain of the latter. See
Figure 23. In the 80’s of the last century, several
Figure 36. Mount Phoenix slide, Yan’an, Shaanxi Province. east-west extending long cracks appeared above the

140
Figure 37. Engineering geological section 2–2 of Mt. Phoenix landslide, Yan’an Shaanxi Province (Zhang et al 2004).

Figure 38. Longitudinal section of Fengjiahe landslide and remedial works.

Figure 39. Low angle inclined draining boreholes at Fengji- Figure 40. Lishan landslide and layout of extensometers
ahe landslide. and inclinometering hole BZ4.

141
Figure 41. Longitudinal geological section of Block I of Lishan landslide and remedial works.

waist of the slide, widening year by year. As there select testing methods, especially the loading-drainage
is a large irrigation canal which runs from west to path to be followed in the course of test, but also pay
east along the down-slope side of the slide waist and attention to the important factor of matrix suction. In
also as the landslide overhangs the densely populated other words, loess strength is very sensitive to change
downtown district of the city, a decision was made in soil water content (or degree of saturation), so it
to remedy the situation and the main approach was can be said that in the geomorphological evolution of
interception of water feeding the slide from outside the Loess Plateau and initiation of loess landslides soil
and drainage of the slide-mass. To that end a system suction has been playing an important part.
of underground galleries was adopted and constructed
with satisfactory results.
7.2 Stability assessment and computation
of the ‘‘slide push’’
7 PROBLEMS IN NEED OF FURTHER In domestic engineering practice it is usual to use either
IN-DEPTH STUDY of the following two methods:
For a segmented sliding bed, the stability factor can
This paper is intended as a concise summing-up of be computed by means of the following Equation. See
the research results and engineering experiences with also Figure 42 (GB50330–2002).
respect to loess and loess landslides in China. It is obvi- n−1  n−1 
ous that much more needs to be done. Here, the authors i=1 Ri j=i j + Rn
would like to mention four issues. Advancement in Ks =     (1)
n−1 n−1
these respects will be of much help to the develop- i=1 Ti j=i j + Tn
ment of the ‘‘West’’, especially in disaster prediction
and disaster prevention and remediation. Where ψj is the coefficient of force transfer
expressed as follows:
7.1 Strength variability of loess ψj = cos(θi − θi+l ) − sin(θi − θi+l ) tan ψi+1
In loess slope stability study, one should not only know Although it is not without deficiencies, most rele-
the peak and residual strengths of the soil and carefully vant codes currently in effect in China are for use of

142
 For loess-bedrock surface contact landslides

L = 14.72 − 5.55 h + 2.49 θ + 20.05 d + 2.57 l (5)

Where L is the forecast slide distance.

These models are results of tentative efforts only
and thus have limitations in practical application.
Especially as it is difficult to estimate the slide
length (l) previous to sliding and the previous thick-
ness (d) of soil mass involved in sliding with fair
Figure 42. Stability factor computation for a non-circular accuracy.
slide plane.

this method in engineering practice, with some points 7.4 High cut slopes in railway, highway
of notice and advised care, of course. and deep excavation design
For slides in a relatively homogeneous soil medium,
the sliding bed can be regarded as circular in shape, Efforts to develop the ‘‘West’’ call for large-scale con-
and either the total stress or effective stress method struction of railways and highways and utilization of
can be used to compute the stability factor as follows underground space in urban areas. In loess regions this
(TB 10027–2001, J 125–2001): inevitably entrains the problem of high cut slopes. In
recent years slopes as high as 60 m and more are no
 longer rare in highway building. In face of a deficient
N tan ϕ + cL
Ks =  (2) theoreretical basis and insufficient experiences, the
T
approach people presently take is that of a coordinated
or combination which comprises a safe general slope
  rate, an appropriate number of steps and platforms,
(N − u) tan ϕ  + c L a system of drainage ditches on slope top and along
Ks =  (3) platforms including torrent troughs, and proper slope
T
(step) face framing, planting and slope foot protection
Besides these equations, Bishop’s Simplified (Li et al. 2005).
Method is often used in China.
It is to be reiterated that the foregoing methods are
far from being perfect and much more research work 8 CONCLUSIONS
remains to be done, including sustained systematic
and instrumented monitoring on landslides of various This paper is a concise description and analysis of
types, structures and scales, as well as development of a natural phenomenon that has been coming into
new methods of data analysis. being, evolving and often threatening to cause havoc in
regions of unique environmental conditions, i.e. loess
7.3 Forecast of disaster extent landslides in provinces of loessial terrain. Northwest
Over the past years, few research efforts have been and North China is typical in this respect.
made to forecast the extent of destruction of a landslide Description and analysis of the behavior peculiar
disaster. At present, the advised and feasible approach to loess landslides is made not only in view of the
is data collection and reduction, back-computation and climatic, depositional, stratigraphical, geomorpholog-
selection of relevant parameters for regression analy- ical and in many cases seismic features characteristic
sis. for example, Wang & Zhang (2005) started from of lands of loessial terrain, but also in the light of
the Factual data of ‘‘typical’’ loess landslides, selected the inalienable granulometry and engineering proper-
thickness of loess (h), slope angle (θ), length of slide ties of the soil, collapsibility and matrix suction in
previous to sliding (l) and previous thickness (d) of soil particular.
mass involved in sliding as parameters for regression Methods of landslide prevention and remediation
analysis and proposed the following models to be used currently in use in China are enumerated and actual
in forecasting the extent of a ‘‘typical ’’loess landslide representative examples given.
disaster. The four issues mentioned in Part 7 of the paper
For intra-loess landslides are only a few of the problems that need in-depth
study. Typical and urgent examples are disaster extent
L = −124.96 − 1.86 h + 3.13 θ + 3.24 d + 2.87 l forecast and stability of high railway and highway
cuts which are no longer rare issues as China’s West
(4) quickens its pace of development.

143
REFERENCES Professional Standards Compilation Group, PRC. 2001.
Code for unfavorable geological condition investigation
Derbyshire, E., Meng, X.M. & Wang, J.T. 1994. Collapsi- of railway engineering (TB 10027–2001, 125–2001) (in
ble loess on the loess plateau of China. In Derbyshire, E., Chinese). Beijing: China Railway Press.
Dijkstra, T.A. & Smally, I.J. (eds), Genesis and Proper- National standards compilation group, PRC. 2002. Technical
ties of Collapsible loess; Proc. of the NATO Advanced code for building-and construction-related slope engi-
Research Workshop, Loughborough, UK 1994: 267–293. neering (GB50330–2002) (in Chinese). Beijing: China
Nether lands: Kluwer. Architecture and Building Press.
Jin, Z.X. 1996. Factors governing the distribution of landslide Wang, Y.Y. & Lin, Z.G. 1990. Structural features and
hazards (in Chinese). In Gansu Science Bulletin, 1996 physico-mechanical properties of loess in China (in Chi-
(Supp.): 123–128. nese). Beijing: Science Press.
Lin, Z.G. 1994. Variation in collapsibility and strength of Wang, X.Y., Zhao, Z.S. & Zhao, J.B. 2002. Analysis
loess with age. In Genesis and properties of collapsible of landslides along Tongchuan-Huangling Highway and
soils; Proc. NATO Advanced Research Workshop, Lough- research on their multi-approach remediation. Xi’an:
borough, UK 1994: 247–263. Nether lands, Kluwer. Highway Investigation and Design Institute, Shaanxi
Lin, Z.G. 1997 Huaqing Palace, Xian, China and landslide Province.
hazard. In Natural disaster prediction and mitigation; Wang, N.Q & Zhang, Z.Y. 2005. Study on loess landslide
Proc. intern. Symp., Kyoto, 1997: 299–307. Kyoto: Kyoto disasters (in Chinese). Lanzhou: Lanzhou University
Univ. Press. Press.
Lin, Z.G., Sassa, K. & Bai M.X. 2000. Undrained loading Wu, W.J. & Wang, N.Q. 2006. Landslide hazards in Gansu
and soil liquefaction in the evolution of low angle long Province (in Chinese). Lanzhou: Lanzhou University
travel distance landslides in regions of loessial terrain. Press.
In Proc. of the Third Multi-lateral Workshop on Devel- Xing, Y.C. 2001. Research on effective stress and
opment of Earthquake and Tsunami Disaster Mitigation deformation-strength characteristics unsaturated soils (in
Technologies and their Integration for the Asia-Pacific Chinese). Xi’an: Xi’an University of Science and Tech-
Region, Manila, Philippines, 2000: 39–43. nology.
Li, Z.Q. & Zhao Z.S. 2005. Research on highway high Zhang, Y.Z. & Xu, Z.J. 2004. Investigation and remediation
slope protection technology in loess regions (in Chi- of Mt. Phoenix landslide (in Chinese). In Proc. of the 6th
nese). Xi’an: Highway Investigation and Design Institute, National Symp. on Geotechnical Engineering case histo-
Shaanxi Province. ries, Beijing, 2004: 284–290. Beijing: Weaponry Industry
Qian, H.J. & Lin, Z.G. 1988. Loess and its engineering prob- Press.
lems in China. In Engineering Problems of Regional Soils;
Proc. Intern. Conf., Beijing, 1988: 136–153. Beijing:
International Academic Publishers.

144
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Advances in landslide continuum dynamic modelling

S. McDougall
BGC Engineering Inc., Vancouver, Canada

M. Pirulli
Politecnico di Torino, Torino, Italy

O. Hungr
University of British Columbia, Vancouver, Canada

C. Scavia
Politecnico di Torino, Torino, Italy

ABSTRACT: Landslide continuum dynamic models have improved considerably in the past three decades,
especially within the past few years, but a consensus on the best method of determining the input resistance
parameter values for predictive runs has not yet emerged. A calibration-based approach is one possible method,
although different calibrated values are often obtained using different models to back-analyze the same event.
These differences may be amplified by the different input geometrical assumptions made in each case. With the
authors’ own models, consistent ranges of calibrated parameter values have been found for specific classes of
events, and these ranges can be used for parametric forward-analyses. A long-term goal of this work is to build
a calibration database large enough to permit probabilistic input parameter selection.

1 INTRODUCTION and is based mainly on the authors’ experiences with

our own models.
Continuum dynamic modelling has emerged as a
useful tool for landslide runout analysis and risk
assessment. With increasing attention and coinciding 2 REVIEW OF EXISTING MODELS
advances in computational capabilities, a large num-
ber of models have been developed or are currently in Landslide continuum dynamic models evolved from
development. Several of these models have included established lumped mass and hydrodynamic methods.
innovations that have significantly advanced both our Subsequent innovations have included various meth-
ability to simulate real events and our fundamental ods to account for the behaviour of real landslides,
understanding of rapid landslide processes. Contribu- including the effects of 3D terrain, internal strength,
tions have been made by a number of researchers with mass and momentum changes due to entrainment and
a wide variety of perspectives and goals, making this spatial and temporal rheology variations.
topic truly multidisciplinary. The following review of existing models is not
Still, some significant challenges remain as we exhaustive. Instead, an emphasis has been put on the
move towards more accurate and objective runout pre- most important innovations with time. Models that
diction using continuum dynamic models. One of the are closely related in this context have been grouped
most important challenges is how to select the input together and, within these groupings, a chronolog-
parameters for forward-analysis, and in the case of ical order has been attempted. Although the focus
calibration-based parameter selection, transferability of this review is on landslide-specific models, some
of parameters between different models is a key issue. innovative snow avalanche models have also been
This paper begins with a detailed review of pre- included.
vious developments in landslide continuum dynamic In the following review, a distinction has been
modelling. The latter part of the paper addresses the made between 2D and 3D models. Although there
issues of input parameter selection and transferability are some inconsistencies in terminology related to

145
depth-averaging (which essentially eliminates one resistances. The model (FLO-2D) was demonstrated
dimension from the governing equations), for the using a full-scale mudflow case study. FLO-2D is
purposes of this paper, 2D and 3D simply denote notable because it has been commercially available for
models that simulate motion across 2D and 3D paths, several years and has been used extensively in practice.
respectively. Laigle & Coussot (1997) presented a 2D model
for mud flows based on the Herschel-Bulkley rheol-
ogy (e.g., Coussot 1994). The input parameters were
2.1 Extension of hydrodynamic methods
measured independently using standard rheometric
The earliest continuum dynamic models for landslides techniques. A 3D version of the model was presented
and snow avalanches were essentially dam-break or by Laigle (1997).
flood routing models with modified basal rheologies.
Lang et al. (1979) modified an existing 2D hydrody-
2.2 Incorporation of path-dependent rheology
namic model to include frictional resistance in addition
to classical kinematic viscosity. The frictional com- Landslide mobility is related to the volume and charac-
ponent in the model (AVALNCH) was increased at ter of the source material, but often more importantly
low speeds to simulate the ‘‘fast-stops’’ that had been to the extent, depth and character of the surficial mate-
observed during snow avalanche deposition. Appro- rial encountered along the path. This material can
priate values for the resistance parameters were inves- strongly influence the shear resistance at the basal
tigated by back-analyses of actual events (Lang & interface (Figure 1).
Martinelli 1979). With noted exceptions (i.e., Lang et al. 1979,
Dent & Lang (1980) modified another 2D hydro- Sousa & Voight 1991, Voight & Sousa 1994), the
dynamic model, based on the Simplified Marker- aforementioned models used constant bulk rheolog-
and-Cell numerical method, to include various ical parameters and neglected the influence of internal
combinations of frictional, viscous and turbulent resis- strength. Sassa (1988) presented a 3D model that
tances. Dent & Lang (1983) modified the same included a method to account explicitly for changes in
model to include a biviscous resistance, similar to resistance along the path using a frictional resistance
the Bingham rheology (e.g., Johnson 1970). The model. Instead of using a constant friction angle for
model (BVSMAC) has also been used to back-analyze the duration of motion, Sassa (1988) used a spatially-
full-scale rock and debris avalanches (Trunk et al. variable effective basal friction angle, which accounts
1986, Sousa and Voight 1991, Voight & Sousa 1994). for the influence of pore water pressure implicitly.
Sousa & Voight (1991) and Voight and Sousa (1994)
used progressively-decreasing resistance parameters
to simulate changes along the path.
The Bingham rheology has been implemented by
other workers. Jeyapalan (1981) and Jeyapalan et al.
(1983a) presented a 2D model (TFLOW) based on the
Bingham rheology, which was used to back-analyze
mine waste flow slides (Jeyapalan 1981, Jeyapalan
et al. 1983b). Schamber & MacArthur (1985) and
MacArthur & Schamber (1986) developed 2D and 3D
models for routing mud flows which were also based
on the Bingham rheology. Pastor et al. (2002) pre-
sented a 3D model for the simulation of flow slides.
The Bingham rheology was used and the consolidation
model proposed by Hutchinson (1986) was imple-
mented to simulate pore pressure dissipation during
the course of motion. Several cases of tailings dam
and mine waste dump failures were back-analyzed.
Fread (1988) developed a dam break model based Figure 1. Oblique aerial view of the May/June 2002
on a power law rheology. The model (DAMBRK) was McAuley Creek rock avalanche near Vernon, British
used by Costa (1997) to back-analyze historic and Columbia. Most of the rock debris (about 7 Mm3 total)
prehistoric lahars in the Cascade Volcanic Belt. deposited immediately below the source area (A), but about
1 Mm3 continued another 1.6 km down the narrow valley
O’Brien et al. (1993) presented a 3D model (B). The mobility of this part of the landslide may have been
for routing floods, mud flows and debris flows enhanced by the presence of fine-grained, relatively weak and
using a generalized quadratic rheology, which is possibly saturated glaciolacustrine deposits in the path. (Pho-
capable of accounting for various combinations of tograph taken on August 26, 2002, courtesy of Dr. Réjean
plastic/frictional, viscous and turbulent/dispersive Couture, Geological Survey of Canada, Ottawa.)

146
He argued that the basal resistance is governed by the 2006). Savage & Hutter (1991) modified their original
properties of the bed material, which can fail under model to account for centripetal acceleration due to
rapid loading (Hutchinson & Bhandari 1971, Sassa curvature of the path, which influences the magnitudes
1985), and proposed that the effective basal friction of the internal and basal stresses. They tested the model
angle at different locations could be measured using using their own series of laboratory flume experiments
high speed ring shear tests on samples taken from involving dry plastic beads. The best simulations were
the path. achieved by incorporating velocity-dependent resis-
tance into the model, which they justified on the basis
of previous tests on glass beads. Hutter & Koch (1991)
and Greve & Hutter (1993) compared model results to
2.3 Incorporation of earth pressure theory
additional confined chute experiments involving the
The concept of strain-dependency of stresses within motion of various materials through both concave and
a deforming granular mass is well established in convex path segments. Velocity-dependent resistance
geotechnical earth pressure theory (e.g., Rankine was not implemented in these tests and differences
1857). Although the internal stresses in an extremely between measured and calibrated basal friction angles
rapid landslide have never actually been measured, were observed.
strain-dependency can be inferred from the observed Hutter et al. (1993) extended the Savage-Hutter
behaviour of experimental granular flows as well as theory for the simulation of 3D motion down an
real landslides. In particular, granular flows and land- inclined plane. To maintain compatibility of the multi-
slides do not appear to spread out or contract as readily dimensional internal stresses, they assumed that the
as fluids (Gray et al. 1999, Hungr 1995). stresses in the downslope direction dominate, consis-
Adopting an established soil mechanics technique, tent with the original Savage and Hutter (1989) model.
Sassa (1988) normalized the internal horizontal nor- Instantaneous stress response was still assumed and, in
mal stresses by the total vertical normal stress using contrast to Sassa’s (1988) 3D model, anisotropic stress
pressure coefficients (depth-averaging was performed states were possible. Greve et al. (1994) and Koch
in the vertical direction). He suggested that the stress et al. (1994) developed an improved 3D extension of
coefficients could range between 0 and 1 (represent- the Savage-Hutter theory, which they solved using a
ing solid and fluid extremes, respectively) and used a Lagrangian finite difference scheme that employed a
value based on Jaky’s (1944) equation for the static triangular mesh.
or ‘‘at rest’’ (Terzaghi 1920) earth pressure coeffi- Gray et al. (1999) extended the Savage-Hutter
cient, which he assumed to be valid during motion and theory further for the simulation of motion across
horizontally-isotropic. The model was demonstrated irregular 3D terrain. The assumption of stress domi-
using a back-analysis of a real rock avalanche. nance in the general downslope direction, as measured
Hutter & Savage (1988) presented a 2D model for in a global reference coordinate system, was main-
the simulation of dry granular flows that incorpo- tained and a similar Lagrangian method was employed.
rated a more advanced method to account for internal The model was tested using partially confined chute
strength. A detailed description of the model was experiments involving deposition on a flat surface.
presented subsequently by Savage and Hutter (1989). Additional tests with various materials were presented
Significantly, they recognized that the internal stresses by Wieland et al. (1999). Tai and Gray (1998) proposed
in a deforming granular mass are strain-dependent and a modification to the model to account for gradual tran-
coupled with the basal shear strength. They assumed sitions between limiting stress states, which reduced
that both the internal and basal rheologies were fric- artificial shocks and instabilities that were caused
tional but could be governed by distinct friction angles. by the original numerical method. Various alter-
Using classical Rankine earth pressure theory, they native shock-capturing methods, including Eulerian
suggested that the internal stresses in a deforming schemes, have been investigated by Gray et al. (2003),
granular mass tend to either active or passive states Wang et al. (2004) and Chiou et al. (2005). Pudasaini
as the material diverges or converges, respectively. and Hutter (2003) and Pudasaini et al. (2005) proposed
In contrast to Sassa’s (1988) assumptions, this implied additional modifications for the simulation of motion
that the internal stresses could be greater than hydro- down highly irregular channels.
static in directions of converging motion. They used
only the theoretical limiting values for the stress coef-
ficients in their model, independent of spreading rate
2.4 Incorporation of entrainment capabilities
or magnitude, an approach that assumes instantaneous
stress response and therefore implies infinite stiffness Surficial deposits (e.g., colluvium, till, residual soil,
of the granular mass. alluvium, organics) in the path of a landslide may
The so-called Savage-Hutter Theory has been the fail under rapid loading. Entrainment of this material
basis for a large body of work (c.f., Pudasaini & Hutter increases the volume and alters the composition of a

147
landslide. Momentum transfer accompanies volume An alternative to Takahashi’s (1991) erosion/
change, as the initially stationary path material is deposition rate formulas was proposed by Egashira
accelerated to the landslide velocity by a combination and Ashida (1997) and implemented in the 2D mod-
of solid collisions and fluid thrust. This process results els of Brufau et al. (2000) and Egashira et al. (2001).
in a velocity-dependent inertial resistance, which is Brufau et al. (2000) separated the mass balances of
additional to the basal shear resistance (e.g., Perla the solid and fluid constituents in order to simulate
et al. 1980). Entrainment can be an important char- spatial and temporal variations in flow density. This
acteristic of rapid landslides at any scale, including two-phase model was extended to 3D by Ghilardi
debris avalanches, debris flows, flow slides and rock et al. (2001) and was used to back-analyze two
avalanches (Figure 2). real debris flows. Egashira et al. (2001) pointed out
Entrainment capabilities within a continuum frame- that entrainment is affected by the grain size of the
work were introduced by Takahashi (1991), who bed material and that large boulders often cannot be
developed 2D and 3D models for simulating debris entrained, even in a reach where erosion dominates.
flows in which the solid-fluid mixture was modelled Papa et al. (2004) described entrainment characteris-
as a dilatant fluid (Bagnold 1954). The internal stresses tics for cases in which bed material is different from
were assumed to be hydrostatic and isotropic. Mass the solid material in a debris flow. Other modifications
and momentum transfer due to erosion and deposi- to Takahashi’s (1991) methods were proposed by Lo
tion of material were accounted for explicitly using & Chau (2003) and Chau & Lo (2004).
semi-empirical relationships for erosion and depo- Entrainment capabilities have also been incorpo-
sition rates. Simulation results were compared with rated into snow avalanche models. Entrainment of a
laboratory experiments and observations of real debris finite depth of snow cover by plowing at the flow front
flow deposits. was included in the 3D model (SAMOS) presented by
Sailer et al. (2002). Sovilla and Bartelt (2002) used a
2D model to investigate plowing at the front as well
as erosion at the base of dense snow avalanches, and
their simulations were compared with field mass bal-
ance measurements. Entrainment of snow cover was
also enabled in simulations presented by Turnbull &
Bartelt (2003) using a 2D model. Naim et al. (2003)
presented a 3D model that accounted for erosion and
deposition explicitly using kinematic criteria similar
to the semi-empirical erosion/deposition rate formulas
proposed by Takahashi (1991) and Egashira & Ashida
(1997).

2.5 Towards comprehensive models

Hungr (1995) presented a 2D Lagrangian model
(DAN) that synthesized much of the earlier work by
including features to account for internal strength,
entrainment and rheology variations. He formalized
the concept of ‘‘equivalent fluid’’, used tacitly by
previous workers, by modelling the landslide as a
hypothetical material governed by simple internal and
basal rheological laws. As in the Savage-Hutter theory,
the bulk of the material was assumed to be frictional
and therefore capable of sustaining strain-dependent
internal stresses that could range between active and
passive states. The influence of internal stiffness was
accounted for by incrementing the stresses in propor-
tion to the magnitude of the prevailing strain, the same
concept proposed later by Tai & Gray (1998). Compat-
Figure 2. The Tsing Shan debris flow, which started as a
ibility of the internal and basal stresses was neglected
relatively small slide of 400 m3 but grew to a total volume of in the original model, but the Savage-Hutter coupled
20,000 m3 through entrainment. (Photograph taken on stress approach has since been implemented. Instead
September 14, 1990, courtesy of J. King, Geotechnical of imposing a single basal rheological law, Hungr
Engineering Office, Hong Kong.) (1995) included a selection of rheologies, which could

148
be assigned to different zones within the landslide Iverson (1997) challenged the use of bulk rheologi-
and/or different segments along the path. The user cal relationships, especially in the dynamic modelling
could also specify the depth of erodible path material of debris flows, whose dynamics are strongly influ-
in each segment, and mass and momentum transfer enced by the interaction of relatively distinct solid and
during entrainment were accounted for explicitly. The fluid components. He proposed a generalization of the
model was used to back-analyze several cases of mine Savage-Hutter theory, based on grain-fluid mixture
waste flow slides. DAN has since been used by many theory, to account for viscous pore fluid effects explic-
workers to back-analyze a variety of landslides (e.g., itly, and incorporated the theory into a 2D Lagrangian
Hungr & Evans 1996, Hungr et al. 1998, Ayotte & model. The original Savage-Hutter assumptions of
Hungr 2000, Jakob et al. 2000, Evans et al. 2001, coupled internal and basal stresses and instantaneous
Hungr et al. 2002, Pirulli et al. 2003, Hürlimann et al. stress response to deformation were retained. Iverson
2003, Revellino et al. 2004, Pirulli 2005). It has also (1997) imposed a longitudinal pore pressure distribu-
been used to demonstrate conceptual theories on the tion, based on experimental evidence, and used the
dynamics of rapid landslides, such as the influence of model to simulate a large-scale flume experiment. Vis-
longitudinal sorting on the surging behaviour of debris cous effects were neglected in the simulations and
flows (Hungr 2000) and the influence of entrainment the required frictional input parameters were based on
on landslide mobility (Hungr & Evans 2004). independently measured static values. Significant dif-
A 3D extension of DAN, based on a Lagrangian ferences between the measured and simulated travel
finite element method, was proposed by Chen and distance and travel time were observed, which were
Lee (2000). Internal stiffness and compatibility of the attributed to multiple surges that developed in the
multi-dimensional internal stresses were neglected. experiment but could not be modelled.
Instead, instantaneous stress response was imple- Iverson and Denlinger (2001) and Denlinger and
mented in orthogonal, but otherwise arbitrarily- Iverson (2001) presented a 3D extension of the theory,
oriented, horizontal directions (depth-averaging was which they implemented in a conventional Eulerian
performed in the vertical direction). The frictional framework. The theory was demonstrated using ana-
basal rheology was implemented to back-analyze two lytical solutions of simplified forms of the governing
real landslides. Bingham and Voellmy rheologies were equations and the numerical model was tested using
implemented in subsequent back-analyses (Chen & both small-scale and large-scale flume experiments.
Lee 2002, Chen & Lee 2003, Crosta et al. 2004). A kinematic criterion was used to identify relatively
Entrainment capabilities were incorporated by Chen high resistance zones near frictional flow margins,
et al. (2006) and applied to the back-analysis of a debris allowing the spatial distribution of pore pressure to
avalanche. evolve automatically throughout each simulation. The
An alternative 3D extension of DAN was pro- internal stress distribution computed in the model was
posed by McDougall and Hungr (2004). The model dependent on the orientation of the local reference
(DAN3D) was based on Smoothed Particle Hydro- frame, which was itself arbitrarily-oriented.
dynamics (SPH), a Lagrangian numerical method Denlinger and Iverson (2004) presented a revised
capable of handling large deformations. As in DAN, 3D model for dry flows based on a unique Eulerian-
the internal stresses were coupled to the basal shear Lagrangian hybrid numerical method. The governing
stress and incremented in proportion to the magni- equations were solved in an Eulerian framework, sim-
tude of the prevailing strain. Stress dominance in ilar to the previous model, while a Lagrangian finite
the direction of motion was assumed in order to element method was simultaneously used to track def-
maintain stress compatibility. A material entrainment ormations and redistribute the internal stresses accord-
feature, based on the concept of natural exponential ingly. The model was demonstrated using hypothetical
volume growth, was presented by McDougall and experiments and tested using analytical solutions of the
Hungr (2005). The user-specified growth rate and dam-break problem. Iverson et al. (2004) presented
maximum erosion depths along the path could be con- additional experimental tests involving the motion of
strained by field observations and a change in rheology granular materials across an irregular surface.
could be implemented at the onset of entrainment. Mangeney-Castelnau et al. (2003) presented a 3D
The method was demonstrated through back-analysis model for simulating dry granular flows using a kinetic
of a rock slide-debris avalanche involving substantial scheme. The model (SHWCIN) was based largely on
entrainment. DAN3D has subsequently been used to the Savage-Hutter theory but used the multi-parameter
back-analyze a large number of case studies and good velocity and depth-dependent basal resistance rela-
correspondence with DAN results using the same set tionship proposed by Pouliquen (1999). A hydro-
of input parameters has been repeatedly demonstrated static, isotropic internal stress state was assumed.
(McDougall 2006). Using the growing database of cal- Pirulli (2005) proposed modifications to SHWCIN
ibrated parameters, the model is currently being used to combat observed mesh-dependency problems, per-
in practice for runout prediction. mit simulation of motion across irregular 3D terrain,

149
incorporate the influence of internal strength and allow for improvement. In particular, the problem of how
the selection of more than one possible basal resis- to choose the input parameters for forward-analysis
tance relationship. The modified model (RASH3D) remains a major challenge.
employed an unstructured finite volume mesh and
Iverson and Denlinger’s (2001) frame-dependent
3.1 Input parameter measurement vs. calibration
method to account for non-hydrostatic internal stresses
(Pirulli et al. 2007). RASH3D has been used to A distinction is sometimes made between models that
back-analyze a large number of case studies. require the input of measured rheological parameters
Pitman et al. (2003a) presented a 3D model based and models that must be calibrated by back-analysis.
on the Savage-Hutter theory. The model (TITAN2D) The difference between the two classes is not always
was designed to provide high resolution using a clear because parameter measurement and parameter
parallel, adaptive mesh numerical method, which calibration are not necessarily characteristics of the
runs on distributed memory supercomputers. Iverson models themselves, but rather the modelling approach
and Denlinger’s (2001) frame-dependent method was adopted by the model developers and/or users.
employed to account for non-hydrostatic internal Parameter measurement is advocated by work-
stress states. Entrainment capabilities, based on an ers who maintain that rapid landslide dynamics can
empirical erosion rate, were incorporated by Pitman be described by constitutive relationships that are
et al. (2003b) and demonstrated with a back-analysis functions of intrinsic material properties, that these
of a historic rock avalanche. TITAN2D is currently relationships can be incorporated into dynamic mod-
available as an open source program. els and that the required material properties can be
Naef et al. (2006) presented a 2D model (DFEM) measured using independent methods. Measurement-
which allowed the selection of a number of resistance based models are typically tested using established
relationships. A 3D version of the model was pre- analytical solutions and back-analyses of controlled
sented by Rickenmann et al. (2006), in which DFEM, experiments before being applied to the analysis of
FLO-2D and the 2D model developed by Laigle and real cases. This could be considered the traditional
Coussot (1997) were applied to the back-analysis of scientific approach to dynamic analysis.
two different field cases. In contrast, in the calibration-based approach,
A true 2D (i.e., non-depth-averaged) finite element rheological parameters are constrained by system-
model based on the program TOCHNOG (Roddeman atic adjustment during trial-and-error back-analysis
2001) was presented by Crosta et al. (2006). Entrain- of full-scale prototype events. Simulation is typically
ment of weak path material was incorporated and a achieved by matching the simulated travel distance,
number of resistance relationships could be imple- velocities and extent and depth of the deposit to those
mented. The model was demonstrated using a series of the prototype. Calibration-based models may utilize
of hypothetical cases. the same physically-justifiable constitutive relation-
Alternative solution methods have also been ships as measurement-based models and should simi-
explored. Sampl (1993) used the Particle-in-Cell larly be tested using analytical solutions and controlled
method to solve the 3D Savage-Hutter equations and experiments before being calibrated to specific classes
model the dense flow component of snow avalanches. of field problems. The calibration-based approach is
Frenette et al. (1997) presented a 3D model based on rooted in the empirical methods of classical hydraulic
a pseudoconcentration function. Quecedo and Pastor engineering.
(2001) presented a two-phase 3D model based on the Critics of the calibration-based approach argue that
level set method. True 3D SPH has been applied to model calibration, especially without controlled test-
the simulation of pyroclastic flows (Nagasawa and ing, is equivalent to tuning or curve-fitting (i.e.,
Kuwahara 1993) and dam-breaks (Cleary and Prakash successful simulation can be achieved by arbitrary
2004). Bursik et al. (2003) coupled the numeri- adjustments of the right type and number of variables),
cal methods of Cellular Automata and SPH in a and that model adaptability can therefore be mistaken
depth-averaged form to model erosion and consequent for model accuracy (Iverson 2003). Parameter selec-
surface evolution by debris floods in 3D. tion for the purposes of prediction can also be difficult
because, while simulation of a single event can be
performed quite efficiently, back-analysis of a signif-
3 INPUT PARAMETER SELECTION icant number of similar prototype events is required
for the calibrated parameters to be physically and/or
Landslide dynamic models have advanced incremen- statistically justifiable. Accurate data for several suit-
tally in the past three decades to the point where, when able events can be hard to acquire and the calibration
used in combination with careful engineering and geo- process can be time-consuming.
science judgement, first-order runout prediction is On the other hand, given the extreme complex-
possible. However, there is still considerable room ity of landslide dynamics, the measurement-based

150
approach could be considered idealistic. Although it to reproduce the bulk external behaviour of a prototype
is scientifically appealing to be able to measure the event, including the travel distance and duration and
input parameters independently, no standard tests are the spatial distribution of velocities and flow/deposit
available to measure, for example, the properties of depths (wherever comparable estimates are available
coarse rock avalanche debris travelling at extremely from field observations).
rapid velocities. Such properties, even if measurable, The calibrated parameters are considered apparent,
may change significantly during the course of motion, rather than actual, material properties and cannot, in
along with the rheology itself, and may be scale- general, be measured in the laboratory. As mentioned
dependent. True measurement-based models must be previously, extensive back-analysis is required to build
able to account for any evolution in behaviour without a database of input parameters that can be used for
the need for user-imposed changes and without using prediction. However, this approach reduces reliance
constitutive laws that include ad hoc assumptions. on laboratory-derived material properties and consti-
These requirements are extremely challenging. tutive relationships that may not be valid at full-scale.
Because the equivalent fluid rheologies are simple and
need only a limited number of controlling parameters,
3.2 The ‘‘equivalent fluid’’ approach the models are generally easy to constrain. Both veloc-
In our own modelling, the authors have adopted ity and deposit distributions are strongly affected by
the semi-empirical ‘‘equivalent fluid’’ approach the chosen rheology.
(Figure 3), defined by Hungr (1995) but used for a The premise of the equivalent fluid approach is per-
long time by many other workers (e.g., Sousa and haps summarized best by Voight & Pariseau (1978),
Voight 1991, O’Brien et al. 1993, Rickenmann & Koch who wrote, ‘‘Any model that allows the slide mass
1997). In this framework, the heterogeneous and com- to move from its place of origin to its resting place
plex landslide material is modelled as a hypothetical in the time limits that bound the slide motion is
material governed by simple internal and basal rhe- likely to be consistent with the principal observable
ological relationships, which may be different from fact—that of the slide occurrence itself.’’ In the same
each other. The internal rheology is typically assumed sense, continuum dynamic models that can accurately
to be frictional, based on the method developed by simulate/predict the extent and duration of a land-
Savage & Hutter (1989). In contrast, a unique basal slide and the distribution of intensity (e.g., flow depth
rheology is not imposed. Instead, the basal rheolog- and velocity) within the impact area, regardless of
ical model and its associated parameters, generally the underlying micro-mechanics, should be consid-
only one or two, are selected based on an empiri- ered useful. For the practical purposes of landslide
cal calibration procedure, in which actual landslides risk assessment, this is the only information that is
of a given type are subjected to trial-and-error back- relevant.
analysis. The results are judged in terms of their ability As noted above, a common criticism of this
approach is that, with a sufficiently flexible model,
nearly any event can be simulated by choosing a cer-
tain set of parameters. What is necessary is to seek
patterns of rheological type and parameter ranges that
reproduce the behaviour of groups of events of simi-
lar description. Predictive capability will result only
once such consistent patterns have been identified.
This work is now in progress.

3.3 Calibrated parameters

The three models developed by the authors (DAN,
DAN3D and RASH3D) have been used to back-
analyze a number of landslides of various types and
scales, including several of the same cases. It should
be noted again that, although there has been exten-
sive cooperation between the authors, RASH3D was
developed largely independently as an extension of the
model SHWCIN, originally introduced by Mangeney-
Figure 3. Schematic illustration of the equivalent fluid Castelnau et al. (2003). In keeping with the equivalent
approach (after Hungr 1995). The complex landslide mate- fluid approach, several alternative rheological rela-
rial is modelled as a hypothetical material governed by simple tionships, including frictional, Voellmy and Bingham
internal and basal rheologies. models, have been used for each case study.

151
 Rock avalanches are an interesting subset of the Twenty-three rock avalanches in the data set pre-
combined case study database. Hungr & Evans (1996) sented above were also back-analyzed by Hungr and
and Pirulli (2005) used DAN to back-analyze 34 dif- Evans (1996) using DAN with the Voellmy model.
ferent rock avalanches using the frictional rheology Total runout distance and observed/estimated veloc-
with a constant bulk basal friction angle, φb (i.e., ity and duration were used as calibration criteria to
assuming a constant pore pressure ratio). Total runout constrain the two-parameter model. Calibrated fric-
distance was the single calibration criterion in each tion coefficients ranged between 0.03 and 0.24, while
case. The calibration results are summarized in the calibrated turbulence parameters ranged between 100
histogram in Figure 4. Note that nine duplicate events and 1000 m/s2 . Using f = 0.1 and ξ = 500 m/s2 ,
and two outlier events analyzed by Pirulli (2005) were Hungr & Evans (1996) found that the total runout dis-
excluded from the data set. The results approximately tance in 70% of the cases could be simulated within
obey a normal distribution with a mean φb of 16◦ and an error of approximately 10%.
a standard deviation of 4.3◦ . Similar ranges of calibrated frictional and Voellmy
The rock avalanche data set can be further sub- parameters have been found in more recent back-
divided based on event volume, path morphology, analyses of rock avalanches using DAN3D and
source and path composition and many other factors, RASH3D (McDougall & Hungr 2004, McDougall
providing further constraint to the calibration results. 2007, Pirulli 2005, Pirulli & Mangeney 2007).
For example, Pirulli (2005) suggested that the data set For example, the 1987 Val Pola rock avalanche in the
could be split into three groups, those with elongated, Italian Alps (cf., Govi et al. 2002) has been back-
tongue and T-shaped runout areas (after Nicoletti & analyzed using both of the 3D models, with similar
Sorriso-Valvo 1991), and noted three distinct ranges results. Using the frictional model, for both DAN3D
of calibrated friction angles. and RASH3D the best simulation of the extent of the
Overestimation of velocities and often unrealistic impact area was achieved using φb = 16◦ (McDougall
forward-tapering deposits are well-known characteris- 2006, Pirulli & Mangeney 2007). The same calibrated
tics of the frictional model. The Voellmy model, which friction angle was obtained by Hungr and Evans (1996)
includes frictional and velocity-dependent terms (cf., in their DAN back-analysis of the event (included in
Hungr 1995), typically produces better simulations the data set shown in Figure 4). The DAN3D and
of velocity and deposit distribution, but requires the RASH3D simulation results are compared in Figure 5.
input of two parameters: a friction coefficient, f , and While some differences are evident, the results are
a turbulence parameter, ξ. Since these two parameters qualitatively very similar.
are not necessarily independent, statistical analysis of Pirulli and Mangeney (2007) also presented a sys-
calibration results is not straightforward. tematic calibration of RASH3D for the Val Pola case
study using the Voellmy model. The best simulation
of the event, in terms of the extent of the impact
area and the distribution of deposits, was achieved
using f = 0.24 and ξ = 1000 m/s2 . Figure 6 shows
Pirulli and Mangeney’s (2007) results, along with the
results of a corresponding DAN3D simulation using
the same pair of input parameters. Relatively higher
runup against the opposite slope, particularly by the
northern lobe of the slide mass, was simulated by
DAN3D.
While some of the differences between the DAN3D
and RASH3D results shown in Figures 5 and 6 are
no doubt due to differences in the governing equa-
tions and numerical methods used in each model, it
is important to note that different assumptions about
the geometry of the source failure and the sliding sur-
face were also made in each case. Significantly, the
DAN3D simulations were based on a bulked source
volume of approximately 50 Mm3 (McDougall 2006),
while the RASH3D simulations were based on a
source volume of only 30 Mm3 (Pirulli & Mangeney
2007). This difference has an important influence on
Figure 4. Histogram of calibrated bulk basal friction angles the results when the Voellmy model is used, since
based on DAN back-analyses of 34 different rock avalanches the flow resistance relative to the driving forces is
by Hungr and Evans (1996) and Pirulli (2005). inversely related to the flow depth; the relatively

152
Figure 5. Comparison of DAN3D and RASH3D simula- Figure 6. Comparison of DAN3D and RASH3D simula-
tions of the Val Pola rock avalanche using the frictional model tions of the Val Pola rock avalanche using the Voellmy model
with φb = 16◦ . The sliding surface shown is based on the with f = 0.2 and ξ = 1000 m/s2 (note that the DAN3D
DAN3D analyses. Elevation contours are shown at 50 m case is not calibrated and uses a bulked volume of the source
intervals. mass). The sliding surface shown is based on the DAN3D
analyses. Elevation contours are shown at 50 m intervals.

higher volume and flow depths in the DAN3D case

contributed significantly to the higher runup against DAN3D and RASH3D with these input parameter
the opposite slope. With DAN3D, a better simula- values did not produce satisfactory results.
tion of the runup was achieved using f = 0.2 and
ξ = 500 m/s2 .
Similar to the frictional case, in both the DAN3D 4 DISCUSSION
and RASH3D analyses, the calibrated Voellmy para-
meters are within the range of values reported by As shown above, unique values of calibrated resis-
Hungr and Evans (1996). However, in this case they do tance parameters do not necessarily apply to each
not correspond exactly with Hungr and Evans’ (1996) event. The calibrated values depend on the govern-
calibrated values of f = 0.1 and ξ = 500 m/s2 for ing equations and numerical solution methods that
the Val Pola event. This discrepancy may simply be are used, which vary between different models and
due to the inherent limitations of the 2D model, which even between different versions of the same model.
requires the input of a path profile. In the 2D case, As a result, calibrated values are not necessarily trans-
the deflection of the southern part of the slide mass ferable, and it may be necessary to revisit previous
down the Adda River valley was not modelled. Since back-analyses after any significant modifications are
the mean channel slope in this area is greater than 0.1, made to a particular model. Perhaps more importantly,
use of f = 0.1 in the 3D case should allow the south- the calibrated values depend on the input geometrical
ern lobe to continue travelling downstream, beyond assumptions made in each case. For this reason, the
the observed deposition zone. Nevertheless, using the collection of more accurate field and remote obser-
3D model developed by Chen & Lee (2000), Crosta vations of case studies and the development of more
et al. (2004) obtained calibrated values of f = 0.1 objective methods of setting up and running the numer-
and ξ = 200 m/s2 in their back-analysis of the Val ical models are key factors in the future success of the
Pola rock avalanche. Comparative simulations using calibration-based approach. In particular, improved

153
methods of estimating flow velocities are needed to avalanche, and Dr. Franco Godone (CNR-IRPI) and
help constrain multi-parameter models. These goals Dr. Luca Mallen (ARPA Piemonte) for their contribu-
may be somewhat conflicting, however; as better tions to the development of the DEM for the RASH3D
observations become available there is a tendency analyses of the Val Pola rock avalanche.
towards more detailed, and sometimes correspond-
ingly more subjective, numerical analyses.
In spite of these challenges, using DAN, DAN3D REFERENCES
and RASH3D, consistent ranges of input parameter
values have emerged for certain classes of events. Ayotte, D. & Hungr, O. 2000. Calibration of a runout
Rock avalanches have been used as an example in this prediction model for debris-flows and avalanches. In
paper, but similarly well-defined (although typically G.F. Wieczorek & N.D. Naeser (eds.), Proc. of the 2nd
region-specific) calibration results have also been pre- Int. Conf. on Debris-Flow Hazards Mitigation, Taipei:
sented for debris avalanches and debris flows in Hong 505–514. Rotterdam: Balkema.
Kong (Hungr et al. 1998, Ayotte and Hungr 2000), Bagnold, R.A. 1954. Experiments on a gravity-free disper-
sion of large solid spheres in a Newtonian fluid under
Italy (Revellino et al. 2004) and California (Bertolo shear. Proc. of the Royal Society of London A 225: 49–63.
and Wieczorek 2005) and flow slides of coal mine Bertolo, P. & Wieczorek, G.F. 2005. Calibration of numerical
waste in British Columbia (Hungr et al. 2002). In each models for small debris flows in Yosemite Valley, Cal-
of these cases, constant or ‘‘bulk’’ rheologies with only ifornia. Natural Hazards and Earth System Sciences 5:
one or two adjustable input parameters were imple- 993–1001.
mented and objective calibration criteria were defined, Brufau, P., Garcia-Navarro, P., Ghilardi, P., Natale, L. &
so model tuning has been kept to a minimum. Savi, F. 2000. 1D mathematical modelling of debris flow.
These ranges of calibrated input values can be used Journal of Hydraulic Research 38(6): 435–446.
for first-order, parametric runout prediction. Reason- Bursik, M., Martínez-Hackert, B., Delgado, H. & Gonzalez-
Huesca, A. 2003. A smoothed-particle hydrodynamic
able confidence limits can often be applied, but the automaton of landform degradation by overland flow.
calibration database is still too small to be used for true Geomorphology 53: 25–44.
probabilistic forward-analysis. With continued expan- Chau, K.T. & Lo, K.H. 2004. Hazard assessment of debris
sion of the database, a probabilistic approach, which flows for Leung King Estate of Hong Kong by incorpo-
would fit well within the framework of quantitative rating GIS with numerical simulations. Natural Hazards
landslide risk assessment, may eventually be possible. and Earth System Sciences 4: 103–116.
In any case, landslide continuum dynamic models Chen, H. & Lee, C.F. 2000. Numerical simulation of debris
should never be used as ‘‘black boxes’’, and experi- flows. Canadian Geotechnical Journal 37: 146–160.
ence and judgement will always play important roles in Chen, H. & Lee, C.F. 2002. Runout analysis of slurry
flows with Bingham model. Journal of Geotechnical and
runout prediction. Back-analyses of real landslide case Geoenvironmental Engineering December: 1032–1042.
studies using dynamic models can serve to enhance Chen, H. & Lee, C.F. 2003. A dynamic model for rainfall-
this experience. induced landslides on natural slopes. Geomorphology 51:
269–288.
Chen, H., Crosta G.B. & Lee, C.F. 2006. Erosional effects on
runout of fast landslides, debris flows and avalanches: a
5 CONCLUSIONS
numerical investigation. Geotechnique 56: 305–322.
Chiou, M.C., Wang, Y. & Hutter, K. 2005. Influence of
As demonstrated in this paper, different calibrated obstacles on rapid granular flows. Acta Mechanica 175:
values may be obtained using different models to back- 105–122.
analyze the same event. These differences may be Cleary, P.W. & Prakash, M. 2004. Discrete-element mod-
amplified by the different input geometrical assump- elling and smoothed particle hydrodynamics: potential
tions made in each case. With the authors’ own models, in the environmental sciences. Phil. Trans. of the Royal
consistent ranges of calibrated values have been found Society of London A 362: 2003–2030.
for specific classes of events. These ranges are rea- Costa, J.E. 1997. Hydraulic modeling for lahar hazards
at Cascades Volcanoes. Environmental and Engineering
sonably well-defined and can be used for parametric
Geosciences 3(1): 21–30.
forward-analyses. Eventually, the calibration database Coussot, P. 1994. Steady, laminar flow of concentrated
may be large enough to permit probabilistic input mud suspensions in open channel. Journal of Hydraulic
parameter selection. Research 32(4): 535–559.
Crosta, G.B., Chen, H. & Lee, C.F. 2004. Replay of the 1987
Val Pola Landslide, Italian Alps. Geomorphology 60(11):
127–146.
ACKNOWLEDGEMENTS Crosta, G.B., Imposimato, S. & Roddeman, D.G. 2006.
Continuum numerical modelling of flow-like landslides.
The authors gratefully acknowledge Prof. Giovanni In S.G. Evans, G. Scarascia Mugnozza, A. Strom &
Crosta (University of Milan), who provided the R.L. Hermanns (eds.) Landslides from Massive Rock
DEM for the DAN3D analyses of the Val Pola rock Slope Failure: 211–232. Springer.

154
Denlinger, R.P. & Iverson, R.M. 2001. Flow of vari- Hungr, O., Sun, H.W. & Ho, K.K.S. 1998. Mobility of
ably fluidized granular masses across three-dimensional selected landslides in Hong Kong—Pilot back-analysis
terrain, 2. Numerical predictions and experimental tests. using a numerical model. Report of the Geotechnical
Journal of Geophysical Research 106(B1): 553–566. Engineering Office. Hong Kong SAR Government.
Denlinger, R.P. & Iverson, R.M. 2004. Granular avalanches Hungr, O. 2000. Analysis of debris flow surges using the
across irregular three-dimensional terrain: 1. Theory and theory of uniformly progressive flow. Earth Surface
computation. Journal of Geophysical Research 109: Processes and Landforms 25: 1–13.
F01014. Hungr, O., Dawson, R., Kent, A., Campbell, D. &
Dent, J.D. & Lang, T.E. 1980. Modeling of snow flow. Morgenstern, N.R. 2002. Rapid flow slides of coal-mine
Journal of Glaciology 26(94): 131–140. waste in British Columbia, Canada. In S.G. Evans &
Dent, J.D. & Lang, T.E. 1983. A biviscous modified Bingham J.V. DeGraff (eds.) Catastrophic Landslides: Effects,
model of snow avalanche motion. Annals of Glaciology Occurrence and Mechanisms 15: 191–208. Geological
4: 42–46. Society of America.
Egashira, S. & Ashida, K. 1997. Sediment transport in steep Hungr, O. & Evans, S.G. 2004. Entrainment of debris in
slope flumes. Proc. of the Roc Japan Joint Seminar on rock avalanches; an analysis of a long run-out mecha-
Water Resources. nism. Geological Society of America Bulletin 116(9/10):
Egashira, S., Honda, N. & Itoh, T. 2001. Experimental 1240–1252.
study on the entrainment of bed material into debris flow. Hürlimann, M., Corominas, J., Moya, J. & Copons, R. 2003.
Physics and Chemistry of the Earth (C) 26(9): 645–650. Debris-flow events in the eastern Pyrenees: preliminary
Evans, S.G., Hungr, O. & Clague, J.J. 2001. Dynamics of the study on initiation and propagation. In D. Rickenmann &
1984 rock avalanche and associated distal debris flow on C.L. Chen (eds.) Proc. of the 3rd Int. Conf. on Debris-
Mount Cayley, British Columbia, Canada; implications flow Hazards Mitigation: Mechanics, Prediction and
for landslide hazard assessment on dissected volcanoes. Assessment, Davos: 115–126. Rotterdam: Millpress.
Engineering Geology 61: 29–51. Hutchinson, J.N. 1986. A sliding-consolidation model
Fread, D.L. 1988. The National Weather Service DAMBRK for flow-slides. Canadian Geotechnical Journal 23:
Model: Theoretical Background and User Documenta- 115–126.
tion. Hydrologic Research Laboratory, National Weather Hutchinson, J.N. & Bhandari, R.K. 1971. Undrained loading,
Service, Silver Spring, MD, 320 pp. a fundamental mechanism of mudflow and other mass
Frenette, R., Eyheramendy, D. & Zimmermann, T. 1997. movements. Géotechnique 21: 353–358.
Numerical modelling of dam-break type problems for Hutter, K. & Savage, S.B. 1988. Avalanche dynamics: The
Navier-Stokes and granular flows. In C.L. Chen (ed.) motion of a finite mass of gravel down a mountain side.
Debris-flow hazards mitigation: mechanics, prediction In C. Bonnard (ed.) Proc. of the 5th Int. Symp. on
and assessment: 586–595. Landslides, Lausanne: 691–697. Rotterdam: Balkema.
Ghilardi, P., Natale, L. & Savi, F. 2001. Modeling of debris Hutter, K. & Koch, T. 1991. Motion of a granular avalanche
flow propagation and deposition. Physics and Chemistry in an exponentially curved chute: experiments and the-
of the Earth 26(9): 651–656. oretical predictions. Phil. Trans. of the Royal Society of
Govi, M., Gullà, G. & Nicoletti, P.G. 2002. Val Pola rock London A 334: 93–138.
avalanche of July 28, 1987, Valtellina (Central Italian Hutter, K., Siegel, M., Savage, S.B. & Nohguchi, Y. 1993.
Alps). In S.G. Evans & J.V. DeGraff (eds.) Catastrophic Two-dimensional spreading of a granular avalanche down
Landslides: Effects, Occurrence and Mechanisms 15: an inclined plane, Part 1: Theory. Acta Mechanica 100:
71–89. Geological Society of America. 37–68.
Gray, J.M.N.T., Wieland, M. & Hutter, K. 1999. Gravity- Iverson, R.M. 1997. The physics of debris flows. Reviews of
driven free surface flow of granular avalanches over Geophysics 35(3): 245–296.
complex basal topography. Proc. of the Royal Society of Iverson, R.M. & Denlinger, R.P. 2001. Flow of variably
London A 455: 1841–1874. fluidized granular masses across three-dimensional ter-
Gray, J.M.N.T., Tai, Y.C. & Noelle, S. 2003. Shock waves, rain, 1. Coulomb mixture theory. Journal of Geophysical
dead zones and particle-free regions in rapid granular free- Research 106(B1): 537–552.
surface flows. Journal of Fluid Mechanics 491: 161–181. Iverson, R.M., Logan, M. & Denlinger, R.P. 2004. Granular
Greve, R. & Hutter, K. 1993. Motion of a granular avalanche avalanches across irregular three-dimensional terrain: 2.
in a convex and concave curved chute: experiments and Experimental tests. Journal of Geophysical Research 109:
theoretical predictions. Phil. Trans. of the Royal Society F01015.
of London, Physical Sciences and Engineering 342(1666): Jakob, M., Anderson, D., Fuller, T., Hungr, O. & Ayotte, D.
573–600. 2000. An unusually large debris flow at Humming-
Greve, R., Koch, T. & Hutter, K. 1994. Unconfined flow of bird Creek, Mara Lake, British Columbia. Canadian
granular avalanches along a partly curved surface, Part 1: Geotechnical Journal 37: 1109–1125.
Theory. Proc. of the Royal Society of London A 445: Jaky, J. 1944. The coefficient of earth pressure at rest. Jour-
399–413. nal of the Society of Hungarian Architects and Engineers
Hungr, O. 1995. A model for the runout analysis of rapid October: 355–358.
flow slides, debris flows, and avalanches. Canadian Jeyapalan, J.K. 1981. Analyses of flow failures of mine tail-
Geotechnical Journal 32: 610–623. ings impoundments. Ph.D. thesis, University of California,
Hungr, O. & Evans, S.G. 1996. Rock avalanche runout pre- Berkeley.
diction using a dynamic model. In K. Senneset (ed.) Proc. Jeyapalan, J.K., Duncan, J.M. & Seed, H.B. 1983a. Anal-
of the 7th Int. Symp. on Landslides, Trondheim: 233–238. yses of flow failures of mine tailings dams. Journal of
Rotterdam: Balkema. Geotechnical Engineering 109(2): 150–171.

155
Jeyapalan, J.K., Duncan, J.M. & Seed, H.B. 1983b. Inves- O’Brien, J.S., Julien, P.Y. & Fullerton, W.T. 1993. Two-
tigation of flow failures of tailings dams. Journal of dimensional water flood and mudflow simulation. Journal
Geotechnical Engineering 109(2): 172–189. of Hydraulic Engineering 119(2): 244–261.
Johnson, A.M. 1970. Physical Processes in Geology. Free- Papa, M., Egashira, S. & Itoh, T. 2004. Critical conditions
man, Cooper and Co., San Francisco. 577 pp. of bed sediment entrainment due to debris flow. Natural
Koch, T., Greve, R. & Hutter, K. 1994. Unconfined flow of Hazards and Earth System Sciences 4: 469–474.
granular avalanches along a partly curved surface, Part Pastor, M., Quecedo, M., Fernandez Merodo, J.A., Herrores,
2: Experiments and numerical computations. Proc. of the M.I., Gonzalez, E., & Mira P. 2002. Modelling tailings
Royal Society of London A 445: 415–435. dams and mine waste dumps failures. Geotechnique 52(8):
Laigle, D. 1997. A two-dimensional model for the study of 579–591.
debris-flow spreading on a torrent debris fan. In C.L. Perla, R., Cheng, T.T. & McClung, D.M. 1980. A two-
Chen (ed.) Debris-flow hazards mitigation: mechanics, parameter model of snow-avalanche motion. Journal of
prediction and assessment: 123–132. Glaciology 26(94): 197–207.
Laigle, D. & Coussot, P. 1997. Numerical modeling of Pirulli, M., Preh, A., Roth, W., Scavia, C. & Poisel, R. 2003.
mudflows. Journal of Hydraulic Engineering 123(7): Rock avalance run out prediction: combined application of
617–623. two numerical methods. In Proceedings of the Int. Symp.
Lang, T.E., Dawson, K.L. & Martinelli Jr., M. 1979. on Rock Mechanics, Johannesburg: 903–908.
Application of numerical transient fluid dynamics to Pirulli, M. 2005. Numerical modelling of landslide runout:
snow avalanche flow, Part 1: Development of computer A continuum mechanics approach. Ph.D. thesis, Politech-
program AVALNCH. Journal of Glaciology 22(86): nico di Torino, Italy. 204 pp.
107–115. Pirulli, M. & Mangeney, A. 2007. Results of back-analysis
Lang, T.E. & Martinelli Jr., M. 1979. Application of numer- of the propagation of rock avalanches as a function of the
ical transient fluid dynamics to snow avalanche flow, Part assumed rheology. Rock Mechanics and Rock Engineering
2: Avalanche modeling and parameter error evaluation. (published online June 21, 2007).
Journal of Glaciology 22(86): 117–126. Pirulli, M., Bristeau, M.O., Mangeney, A. & Scavia C.
Lo, K.H. & Chau, K.T. 2003. Debris-flow simulations for 2007. The effect of the earth pressure coefficients on the
Tsing Shan in Hong Kong. In D. Rickenmann & C.L. runout of granular material. Environmental Modelling &
Chen (eds.) Proc. of the 3rd Int. Conf. on Debris-flow Haz- Software 22: 1437–1454.
ards Mitigation: Mechanics, Prediction and Assessment, Pitman, E.B, Nichita, C.C., Patra, A., Bauer, A.,
Davos: 577–588. Rotterdam: Millpress. Sheridan, M. & Bursik, M. 2003a. Computing granu-
MacArthur, R.C. & Schamber, D.R. 1986. Numerical lar avalanches and landslides. Physics of Fluids 15(12):
methods for simulating mudflows. In Proc. of the 3rd Int. 3638–3646.
Symp. on River Sedimentation, Mississippi: 1615– 1623. Pitman, E.B., Nichita, C.C., Patra, A.K., Bauer, A.C.,
Mangeney-Castelnau, A., Vilotte, J.P., Bristeau, O., Bursik, M. & Webb, A. 2003b. A model of granular
Perthame, B., Bouchut, F., Simeoni, C. & Yerneni, S. flows over an erodible surface. Discrete and Continuous
2003. Numerical modeling of avalanches based on Saint Dynamical Systems Series B 3(4): 589–599.
Venant equations using a kinetic scheme. Journal of Pouliquen, O. 1999. Scaling laws in granular flows down
Geophysical Research 108(B11): 2527. rough inclined planes. Physics of Fluids 11(3): 542–548.
McDougall, S. & Hungr, O. 2004. A model for the analysis of Pudasaini, S.P. & Hutter, K. 2003. Rapid shear flows of
rapid landslide motion across three-dimensional terrain. dry granular masses down curved and twisted channels.
Canadian Geotechnical Journal 41: 1084–1097. Journal of Fluid Mechanics 495: 193–208.
McDougall, S. & Hungr, O. 2005. Dynamic modelling of Pudasaini, S.P., Wang, Y. & Hutter, K. 2005. Rapid motions
entrainment in rapid landslides. Canadian Geotechnical of free-surface avalanches down curved and twisted chan-
Journal 42: 1437–1448. nels and their numerical simulation. Phil. Trans. of the
McDougall, S. 2006. A new continuum dynamic model for Royal Society of London A 363: 1551–1571.
the analysis of extremely rapid landslide motion across Pudasaini, S.P. & Hutter, K. 2006. Avalanche Dynamics:
complex 3D terrain. Ph.D. thesis, University of British Dynamics of rapid flows of dense granular avalanches.
Columbia, Canada. 253 pp. Springer. 626 pp.
Naaim, M., Faug, T. & Naaim-Bouvet, F. 2003. Dry granular Quecedo, M. & Pastor M. 2001. Application of the level set
flow modelling including erosion and deposition. Surveys method to the finite element solution of two-phase flows.
in Geophysics 24: 569–585. Int. Journal of Numerical Methods in Engineering 50:
Naef, D., Rickenmann, D., Rutschmann, P. & McArdell B.W. 645–663.
2006. Comparison of flow resistance relations for debris Rankine, W.J.M. 1857. On the stability of loose earth. Phil.
flows using a one-dimensional finite element simulation Trans. of the Royal Society of London 147.
model. Natural Hazards and Earth System Sciences 6: Revellino, P., Hungr, O., Guadagno, F.M. & Evans, S.G.
155–165. 2004. Velocity and runout simulation of destructive debris
Nagasawa, M. & Kuwahara, K. 1993. Smoothed particle sim- flows and debris avalanches in pyroclastic deposits, Cam-
ulations of the pyroclastic flow. Int. Journal of Modern pania region, Italy. Environmental Geology 45: 295–311.
Physics B 7(9/10): 1979–1995. Rickenmann, D., Laigle, D., McArdell, B.W. & Hübl, J.
Nicoletti, G. & Sorriso-Valvo, M. Geomorphic controls of 2006. Comparison of 2D debris-flow simulation mod-
the shape and mobility of rock avalanches. Geological els with field events. Computational Geosciences 10:
Society of America Bulletin 103: 1365–1373. 241–264.

156
Roddeman, D.G. 2001. TOCHNOG User’s Manual. FEAT, Sovilla, B. & Bartelt, P. 2002. Observations and modelling of
177 pp. snow avalanche entrainment. Natural Hazards and Earth
Sailer, R., Rammer, L. & Sampl, P. 2002. Recalculation of System Sciences 2: 169–179.
an artificially released avalanche with SAMOS and val- Tai, Y.C. & Gray, J.M.N.T. 1998. Limiting stress states in
idation with measurements from a pulsed Doppler radar. granular avalanches. Annals of Glaciology 26: 272–276.
Natural Hazards and Earth System Sciences 2: 211–216. Takahashi, T. 1991. Debris flow. International Association
Sampl, P. 1993. Current status of the AVL avalanche for Hydraulic Research monograph. Rotterdam: Balkema.
simulation model—Numerical simulation of dry snow Terzaghi, K. 1920. Old earth pressure theories and new test
avalanches. In L. Buisson & G. Brugnot (eds.) Proc. of results. Engineering News Record 85: 632.
the Pierre Beghin Int. Workshop on Rapid Gravitational Trunk, F.J., Dent, J.D. & Lang, T.E. 1986. Computer
Mass Movements, Grenoble: 269–276. modeling of large rock slides. Journal of Geotechnical
Sassa, K. 1985. The mechanism of debris flows. In Pro- Engineering 112: 348–360.
ceedings of the 11th Int. Conf. on Soil Mechanics and Turnbull, B. & Bartelt, P. 2003. Mass and momentum bal-
Foundation Engineering, San Francisco 1: 1173–1176. ance model of a mixed flowing/powder snow avalanche.
Sassa, K. 1988. Geotechnical model for the motion of land- Surveys in Geophysics 24: 465–477.
slides. In C. Bonnard (ed.) Proc. of the 5th Int. Symp. on Voight, B. & Sousa, J. 1994. Lessons from Ontake-san:
Landslides, Lausanne: 37–56. Rotterdam: Balkema. a comparative analysis of debris avalanche dynamics.
Savage S.B. & Hutter, K. 1989. The motion of a finite mass of Engineering Geology 38: 261–297.
granular material down a rough incline. Journal of Fluid Wang, Y., Hutter, K. & Pudasaini, S.P. 2004. The Savage-
Mechanics 199: 177–215. Hutter theory: A system of partial differential equa-
Savage S.B. & Hutter, K. 1991. The dynamics of avalanches tions for avalanche flows of snow, debris, and mud.
of granular materials from initiation to runout, Part 1: Zeitschrift fur Angewandte Mathematik und Mechanik
Analysis. Acta Mechanica 86: 201–223. 84(8): 507–527.
Schamber, D.R. & MacArthur, R.C. 1985. One-dimensional Wieland, M., Gray, J.M.N.T. & Hutter, K. 1999. Channelized
model for mud flows. In Proceedings of the ASCE Spe- free-surface flow of cohesionless granular avalanches in
cialty Conf. on Hydraulics and Hydrology in the Small a chute with shallow lateral curvature. Journal of Fluid
Computer Age 2: 1334–1339. American Society of Civil Mechanics 392: 73–100.
Engineers, New York.
Sousa, J. & Voight, B. 1991. Continuum simulation of flow
failures. Géotechnique 41(4): 515–538.

157
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Deformation and failure mechanisms of loose and dense fill slopes

with and without soil nails

C.W.W. Ng
Department of Civil Engineering, The Hong Kong University of Science and Technology, HKSAR

ABSTRACT: Although fill slopes have been constructed worldwide to meet the needs and development of
various human activities for years, the effects of fill density, material type, including gap-graded or well-graded
soils, various stabilisation methods, including soil nails, and destabilising agents, including rainfall infiltration,
rising ground water and earthquakes, are still not well understood. In this paper, various studies of deformation
and failure mechanisms of unreinforced and nailed fill slopes using a geotechnical centrifuge are described and
key findings are reported and explained.

1 INTRODUCTION The effects of using soil nails for stabilising both

loose and dense fill slopes are also investigated in this
Fill slopes have been constructed across the world for paper. Failure and deformation mechanisms of slopes,
highway and railway embankments, road widening which cannot be obtained easily and economically in
projects, earth dams, landfills and levees for years. The the field or reliably from numerical analysis are pro-
construction of fill slopes generally involves borrow- vided and highlighted in this paper. All test results are
ing fill materials from local sources or from foreign reported in prototype scale unless stated otherwise.
sites and compaction. Catastrophic failures of uncom-
pacted and improperly compacted fill slopes resulting
in loss of lives and severe damage to property have 2 ROLES OF FIELD WORK AND NUMERICAL
been documented (Hong Kong Government 1977, Sun AND CENTRIFUGE MODELLING
1999). In addition, excessive settlement and surface
erosion failure to densely compacted new slopes due to Geotechnical centrifuge modelling has become an
rainfall infiltration have been reported. It is clear that alternative modelling tool to investigate failure mech-
the deformation and failure mechanisms of fill slopes anisms and deformations of slopes and soil structures
subjected to rainfall infiltration and earthquakes are (Kimura 1998, Ng et al. 2003, 2006a). The basic
not fully understood. In this paper, the performance principle of geotechnical centrifuge modelling is to
of both loose and dense fill slopes is investigated recreate the stress conditions that would exist in a
wih the aid of centrifuge modelling technology. Both full-scale construction (prototype), using a model on
gap-graded Leighton Buzzard (LB) Fraction E fine a greatly reduced scale. This is done by subjecting
sand and well-graded completely decomposed granite the model’s components to an enhanced body force,
(CDG), i.e. silty sand, are used as model fill materials. which is provided by a centripetal acceleration of mag-
The model fill slopes are destabilised by rainfall infil- nitude, ng, where g is the acceleration due to the earth’s
tration, rising ground water and simulated earthquakes gravity (i.e. 9.81 m/s2 ). Stress replication in an nth
using a hydraulic shaker installed on the geotechnical scale model is achieved when the imposed ‘‘gravita-
centrifuge at the Hong Kong University of Science tional’’ acceleration is equal to ng. Thus, a centrifuge
and Technology (HKUST; Ng et al. 2001, 2004b). is suitable for modelling stress-dependent problems.
The 400 g-ton centrifuge is 8.4 m in diameter and Moreover, a significant reduction of testing duration
is equipped with a hydraulic bi-axial shaker that can such as consolidation and seepage can be achieved by
simulate earthquakes in two horizontal directions (Ng using a reduced size model (Ng 1999).
et al. 2004b) with a maximum input acceleration of Figure 1 shows the inter-relationships between field
50 g (model scale). In addition, a four-axis robotic investigation, numerical modelling and physical
manipulator is installed in the centrifuge for simulat- (including centrifuge) modelling. These three approa-
ing some construction activities in-flight such as soil ches are definitely not mutually exclusive. On
nailing (Ng et al. 2002a). the contrary, they are complementary as no

159
 Deviator stress
Field (a) B Strain hardening
investigation Dilation
Investigation &
Evaluation verification C
Evaluation Strain softening Strain hardening
& Limited
&
calibration liquefaction
Failure calibration
mechanisms,
new theories A
Investigation & Investigation & Liquefaction Strain softening
& design
verification verification Axial strain
guidelines
(b) B Dilation

Deviator stress
Centrifuge Numerical Undrained
modelling Evaluation & calibration modelling strength increase
Phase C due to dilative
transformatio
tendency
n point
Figure 1. The inter-relationships between field investiga- Limited
liquefaction
tion, numerical modelling and centrifuge modelling. contractive tendency
A
Liquefaction Undrained strength reduction
due to contractive tendency
approach is perfect in solving every geotechnical or Mean effective stress
geo-environmental problem in terms of the quality of Contractive tendency Liquefaction
(c) A
Excess pore pressure
results, time and cost. Ideally, two or three approaches
should be utilized to calibrate their results against each
other and to verify any conclusion drawn. Small-scale
laboratory model tests without correctly considering Axial strain
stress effects or using an incorrect constant stress Limited C
distribution with depth can be very misleading as liquefaction
contractive tendency
soil behaviour is stress dependent. These tests should B
therefore be treated with caution. Dilation Dilative tendency

Figure 2. Liquefaction, limited liquefaction, and dilation in

monotonic loading tests (modified from Castro 1969, Kramer
3 FAILURE MECHANISMS OF LOOSE SAND 1996).
FILL SLOPES WITH AND WITHOUT SOIL
NAILS

Liquefaction failure due to cyclic and earthquake large strains at a low effective confining pressure and
loading has received much attention and significant a low shear strength at large strain. This behaviour
progress has been made to improve our understand- is often loosely called ‘‘liquefaction’’ or ‘‘flow liq-
ing of the failure mechanisms and viable stabilisation uefaction’’. No matter whether it is called ‘‘flow
measures. On the contrary, the mechanisms of static liquefaction’’ or ‘‘liquefaction’’, the terminology to
liquefaction failure of loose fill slopes are not yet fully describe the material behaviour observed in the labo-
understood. It is clear that an under-design of sta- ratory is rather confusing and, strictly speaking, incor-
bilisation measures for existing loose fill slopes may rect. Would it be clearer and more precise to describe
lead to catastrophic results. In contrast, an over-design the material behaviour of loose specimen, A, and
of upgrading methodologies could result in exces- dense specimen, B, as ‘‘strain-softening’’ and ‘‘strain-
sive costs and adverse environmental impacts, such hardening’’, respectively, in the deviator stress-axial
as unnecessarily removing trees and shrubs that have strain space (see Fig. 2a)? In the mean effective stress-
been growing on existing fill slopes for years. It is deviator stress space (see Fig. 2b), would it be more
clear that more research work is needed. precise to use the terms ‘‘undrained strength reduc-
tion’’ (or so-called collapse (Sladen et al. 1985)) and
‘‘undrained strength increase’’ to describe the strength
3.1 Clarification of the terminology relating
changes of specimen A and specimen B, respectively?
to static liquefaction
Of course, it is well-recognised that a reduction and an
Figure 2 shows some typical results from monotonic increase in the undrained shear strength are caused by
triaxial tests on saturated, anisotropically consolidated the respective tendency of sample contraction and dila-
sand specimens. The figure shows that the very loose tion, leading to a respective increase and a reduction
sand specimen, A, exhibits a peak undrained shear in the pore water pressure (u) for specimens A and
strength at a small shear strain and then ‘‘collapses’’ to B during undrained shearing (see relationship between

160
u and axial strain in Fig. 2). It must be pointed out 800

that these are simply material element behaviour that (a)

700
do not capture the global behaviour of an entire fill
slope or an earth structure. 600

500
e =0.973
Liquefaction potential (LP)
3.2 Investigation of the failure mechanism

q (kPa)
400
of liquefied flow in the centrifuge e =0.970
300
3.2.1 Model material Quasi-steady state
200
Centrifuge model tests were previously carried out
to investigate the failure mechanisms of static lique- 100 e =0.983
faction flow of loose fill slopes subjected to rainfall e =0.992
0
and a rising ground water table at HKUST (Zhang 0 10 20 30 40
2006, Zhang et al. 2006, Ng et al. 2007). Leighton
Buzzard (LB) Fraction E fine sand was selected as
the fill material for the model tests. Figure 3 shows (b)
700
the gap-graded particle size distribution of LB sand.
D10 and D50 of the sand were 125 μm and 150 μm,
respectively. Following BS1377 (1990), the maximum 500
and minimum void ratios of the LB sand were found
to be 1.008 and 0.667, respectively (Cai 2001). The
estimated saturated coefficient of permeability was 300
1.6×10−4 m/s. LB sand was chosen because of its pro-
nounced strain-softening characteristics with its high
liquefaction potential, LP, i.e. it experiences substan- 100

tial strength reduction (see Fig. 4a). The results from

four loose specimens with different initial void ratios
(eo ) shown in the figure are obtained from isotropically
consolidated undrained compression triaxial tests. The
loose sand clearly shows pronounced strain-softening Figure 4. Contractive behaviour of loose LB sand under
behaviour and substantial strength reduction in the consolidated undrained tests (a) in the εq − q and (b) in p − q
planes (modified from Zhang 2006; data from Cai 2001).
deviator stress and shear strain (q − εq ) space and
contractive responses in the mean effective stress (p )
and deviator stress (q) space, i.e. p decreases con-
tinuously as q increases until a peak state is attained sand is 30◦ (Cai 2001). Following the approach pro-
(see Fig. 4b), where p and qare equal to (σ1 + 2σ3 )/3 posed by Lade (1992), the angle of instability (φins 
)
and (σ1 − σ3 ), respectively. After the peak state, q ◦
determined for the sand is 18.6 . It is well-known that

drops (the soil collapses) with a large deformation φins is dependent on void ratio and stress level (Chu &
develops until the quasi-steady state (a shear strain Leong 2002). For engineering assessment and design
of about 15%) or the critical state (shear strain = 30%) of remedial work for loose fill slopes, it may be rea-
is reached. The critical state friction angle (φc ) of the sonable to assume this angle is a constant as the first
approximation.
100
LB-Sand
80
SKW-CDG 3.2.2 Model package and test procedures
Percentage finer (%)

CKL-CDG Figure 5 shows an instrumented loose sand fill slope

BH-CDG
60 WTS-CDG
model together with the locations of the pore water
pressure transducers (PPTs) (Zhang & Ng 2003, Ng
40 et al. 2008). The model slope was prepared by moist
tamping. The initial slope angle and relative com-
20 paction were 29.4◦ and 68%, respectively.
The body of the sand slope was instrumented
0
with seven PPTs and arrays of surface markers were
0.001 0.01 0.1 1 10
Particle size (mm) installed for image analysis of soil movements. Lin-
ear variable differential transformers (LVDTs) and a
Figure 3. Particle size distributions of LB sand and CDG laser sensor were mounted at the crest of the slope to
samples. monitor its settlement.

161
 LVDT
LVDT & Laser sensor

Drainage
board Model
container

Temporary Inlet hole

reservoir

305
Reflectory
Outlet hole
29.4

x
Sand
1130,7

Figure 5. Centrifuge model of a loose sand fill slope sub-

jected to rising ground water table at 60 g (Zhang & Ng 2003).
The dimensions are given in model scale.

The ratio of the model height to the average particle

size was on the order of 1000. According to Goodings
& Gillete (1996), particle size effects would be small
if the slope height were higher than 158 times the grain
diameter of the model material.

3.2.3 Observed static liquefaction mechanism

Although the initial angle of the loose slope at 1 g was
29.4◦ , the slope was densified to 80% of the maximum
relative compaction due to self-weight compaction at
60 g. The slope angle was therefore flattened to 24◦ Figure 6. Slope profile in a loose sand fill test (a) before
(see Fig. 6a), which is steeper than the angle of instabil- rising ground water table; (b) after static liquefaction (Zhang
ity of 18.6◦ . This implies that the slope was vulnerable & Ng 2003).
to instability, which could lead to liquefaction (see
Fig. 4). At 60 g, the 18 m-height (prototype) slope was 1.0
de-stabilised by rising ground water from the bottom 0.8
Slope failure
of the model (Zhang 2006, Ng et al. 2008). The loose
Excess pore pressure ratio ( u / )

0.6 PPT7
sand slope liquefied statically and flowed rapidly (see
0.4 PPT4
Fig. 6b), i.e. it followed a process in which the loose PPT5
slope was sheared under the undrained condition, lost 0.2
PPT2
its undrained shear strength as a result of the induced 0.0 PPT1
high pore water pressure and then flowed like a liq- -0.2 PPT6
uid, which we call ‘‘liquefied flow’’. The liquefaction -0.4
of the loose sand slope was believed to be initially -0.6 PPT3
triggered by seepage force (Ng et al. 2008).
-0.8
Figure 7 shows the measured rapid increases in the
excessive pore water pressure ratio (u/σv ) within -1.0
37.8 38.2 38.6 39.0 39.4 39.8 40.2 40.6 41.0 41.4 41.8 42.2 42.6 43.0
about 25 seconds (prototype) at a number of locations Duration (min)
in the slope during the test. The maximum measured
u/σv was about 0.6, which would be much higher if Figure 7. Measured sudden and substantial increases in
a properly scaled (i.e. more viscous) pore fluid were pore water pressure at seven locations inside the slope (Zhang
used to reduce the rate of dissipation of the excess pore & Ng 2003).
pressure correctly in the centrifuge test. This means
that the slope would liquefy much more easily. As
shown in Figure 6b, the completely liquefied slope water pressures during the test clearly demonstrated
inclines at about 4◦ to 7◦ to the horizontal after the the static liquefaction of the loose sand fill slope. It
test. The observed fluidization from in-flight video should be noted that measurements of the sudden and
cameras and the significant rise in excessive pore significant rise of excessive pore water pressures are

162
 3.3 The role of soil nails in liquefiable flow
(Zhang et al. 2006)
The soil nailing technique has been used worldwide to
stabilise slopes because of its ease of construction and
robustness. In Hong Kong, soil nailing has been used
for improving soil slope stability for many years and is
currently the predominant method used for upgrading
the stability of existing cut slopes.
Figure 8. Postulated failure zones during the liquefaction
of slope SG30 (Ng et al. 2008).
3.3.1 Model preparation, instrumentation
and test procedures
Three dynamic centrifuge model tests using LB sand
essential to ‘‘prove’’ or verify the occurrence of static were carried out to investigate the effect of soil nails on
liquefaction of loose fill slopes if no video recording the failure mechanism of a liquefiable loose sand fill
is available. slope (Zhang et al. 2006). The tests were subjected to
Figure 8 shows five postulated zones, Z1 −Z5 , severe groundwater conditions and then large dynamic
corresponding to the sequence of the failure and lique- loading in the centrifuge. To create reliable liquefac-
faction process of the slope (Ng et al. 2008). Z1 is the tion, the in-flight hydraulic shaker was used in the
failure region de-stabilised by the loss of its toe due centrifuge tests (Ng et al. 2001). Figure 9 shows the
to the seepage force in the gully. The soil mass at the initial geometry and instrumentation of a nailed model
toe of Z1 slid with the soil at the gully head to trigger slope (Zhang et al. 2006). Due to page constraints here,
the failure of Z2 . The soil mass in Z2 collapsed rapidly only two tests are summarised in Table 1. LB sand was
which was then followed by the collapse of Z3 without used to construct the model slopes. In the two tests, i.e.
inducing obvious deformation in the lower part. The DS25 and DSN25 (D stands for dynamic, S for sand,
collapses of Z2 and Z3 were due to the strain-softening N for nail and 25 is the initial slope angle), the slopes
associated with the significant strength reduction (i.e. were initially subjected to severe seepage controlled
the high liquefaction potential) of the loose LB sand as by the boundary conditions before shaking.
illustrated in Figure 4. The rapid undrained collapses All the slope models had an initial slope height
of Z2 and Z3 were evident from the measured large of 250 mm and an initial angle of 25o , as shown in
excess positive pore pressures at PPT7 (see Fig. 7). Figure 9. Sand with water content of about 5% was
The collapse of the soil mass induced by the collapse compacted into 50 mm thick layers at a void ratio of
of the other parts of the slope inclined at an angle of about 1.5. To minimize side friction, the container
10◦ was also reported by Moriwaki et al. (2004). They wall was covered with a thin layer of smooth plas-
observed a large-scale landslide in loose sandy soil tic membrane. The upstream drainage board (Fig. 9)
induced by artificial rainfall at 1 g. was covered by a rubber plate to decrease the wave
Subsequently, Z4 collapsed as a result of the strain- reflection effects. A thin layer of sand-glue mixture
softening associated with the significant strength was used on the inner bottom surface of the container
reduction (high liquefaction potential) of the loose LB to increase the roughness.
sand (see Fig. 4). In this zone, relatively smaller excess As shown in Figure 9, soil nails were installed in
positive pore pressure ratios were recorded at PPT4 slope DSN25. Each modelled nail was equivalent to a
and PPT5 (see Fig. 7) as compared to that at PPT7. steel bar 35 mm in diameter with a grouting layer of
This likely indicates the smaller extent of shearing in 210 mm on the outer diameter of the prototype. The
Z4. The fluidised soil was carried by the water wave nails were arranged in four rows and five columns
and finally deposited at an angle of 4o to the horizon- with a horizontal distance of 67 mm (2.1 m in the
tal. The dotted line in this figure represents the upper prototype) and a vertical distance of 42.3 mm (1.3 m
boundary of the stable region (Z5 ), monitored by mark- in the prototype). The inclined angle of each nail was
ers and the small excess pore pressures at PPT1 and 40◦ to the horizontal. Depending on the locations of
PPT2 (see Fig. 7) during the liquefaction process. the nails, the lengths of the nails (i.e. A, B, C and D)
Based on the centrifuge results described above varied. The soil nails located in the centre column were
and other centrifuge model tests on unreinforced and instrumented with five sets of semiconductor strain
reinforced loose sand fill slopes subjected to rainfall gauges.
infiltration and shaking (Ng et al. 2006b, Zhang et al. The instrumentation layout for all tests included
2006), it is fairly clear that the strain-softening charac- five miniature accelerometers (ACCs) installed in the
teristic of a material is a necessary but not a sufficient slope to measure soil acceleration in the X-direction,
condition to induce liquefied flows. A trigger such as eight PPTs installed at the base of the slope or near
the seepage force in the gully is required. the accelerometers to record excessive pore pressures

163
 (a)
1m
Post shaking ground
surface

6m

(b)
1m
Post shaking ground
Figure 9. Initial geometry and instrumentation of model surface
slope for tests DS25 (without nails) and DSN25 (nailed slope)
(Zhang et al. 2006).
4m

Table 1. Slope geometries, soil properties and input motion

(data from Zhang et al. 2006).
Figure 10. Measured displacement vectors of slopes during
Model identity DS25 DSN25 shaking by image analysis (a) for DS25, (b) for DSN25 with
the slope profiles before the earthquake (Zhang et al. 2006).
Total height at 30 g before
shaking (m) 8.13 8.16
Initial density (kg/m3 ) 1373 1314 3.3.2 The role of soil nails during slope liquefaction
Initial water content (%) 4.8 5.6
(comparisons of tests DS25 and DSN25)
Void ratio: initial 1.022 1.125
at 30 g before shaking 0.858 0.976 Due to the page length constraints, only some results
Duration of shaking (s) 15.2 66.2 for DS25 and DSN25 are reported in this paper. Other
Peak amplitude (g) 0.02 0.16 test details and results are described by Zhang et al.
(2006).
Note: The initial slope angle and the initial total height was Figure 10 shows localised seepage failure induced
25o and 9.75 m, respectively. by the high hydraulic gradient created between the
upstream and the downstream prior to shaking. The
pre-shaking ground surface profiles are also shown.
No static liquefaction was observed at this stage. After
during the shaking, and two LVDTs installed to shaking, displacement vectors obtained from image
measure the crest settlement. analysis of PIV data of DS25 (see Fig. 10a) indicated
The deformation of the model slopes was captured a deep-seated rotational failure of the liquefied soil,
using a texture-driven image analysis system, GeoPIV, reducing the once sloping ground to an almost hor-
which combines the techniques of digital photography, izontal surface. Figure 11a shows the slope profile
multi-threshold centroiding, particle image velocime- captured by a digital camera after the shaking. The
try (PIV), and close-range photogrammetry (White sand mass was shaken down in the left of the slope
et al. 2003, Zhang et al. 2006). The accuracy of the (or the part close to the slope crest), whereas the sand
measurements in the centrifuge tests was estimated to mass at the right (or the sand close to the slope toe)
be about 0.1 mm. moved horizontally. The estimated maximum depth of
When the acceleration in the centrifuge reached a the liquefied zone was up to about 6.0 m.
nominal value of 30 g, the slopes were subjected to In slope DSN25, the soil nails restricted the devel-
transient seepage controlled by an upstream water sup- opment of this deep-seated failure mechanism, with
ply. A high hydraulic gradient was created between the soil deformations being smaller in magnitude and
upstream and the downstream. The outlet located at the nearer to the surface (Fig. 10b). The sand slid downs-
downstream was kept open throughout the tests. Dur- lope along the failure surface (see Fig. 11b). The
ing an increase in the hydraulic head at the upstream, estimated maximum thickness of the failure zone was
the slope deformed and some seepage failure occurred up to about 4.0 m. Based on the comparison of slope
(Zhang et al. 2006). Subsequently, a sinusoidal earth- profiles between the two tests, the final lateral dis-
quake was triggered by applying acceleration along placement of the sand mass at the slope toe was
the base of the slope. Some details of each shaking are estimated to be 0.71 m, which was 41% less than that
given in Table 1. of DS25. The estimated maximum settlement at the

164
 )/ Crest settlment (m)
100 0.8
7th cycle End of the shaking
80 8th cycle PPT7
C3 0.6
9th cycle
60
Soil profile before shaking 0.4
40
0.2
20

Nail force (KN)

PPT1
0.0

/
0

Excess pore pressure ratio (

-20 -0.2
D3
-40 D3 -0.4
-60 C3
C3 -0.6
-80 D3
LVDT3
-100 -0.8

-120 -1.0
0 10 20 30 40 50 60 70 80 90
Duration (s)
(a)
Figure 12. Typical mobilized nail forces with excess pore
pressure ratios and crest settlement in DSN25 (Zhang et al.
2006).

Slope profile before shaking

Slope profile after shaking Table 2. Maximum excess pore pressure ratios at PPTs 1 to
8 for the two tests (data from Zhang et al. 2006).

Test ID PPT1 PPT2 PPT3 PPT4*

DS25 0.06 0.51 0.39 0.22
DSN25 0.30 0.12 0.36 0.22
Postulated sliding surface
PPT5 PPT6 PPT7 PPT8*
DS25 0.41 0.20 0.80 0.96
DSN25 0.28 0.02 0.66 1.35
(b)
* Estimated locations of PPT4 and PPT8 may be ques-
Figure 11. Slope profiles after the shaking for (a) DS25 and tionable.
(b) DSN25 (Zhang et al. 2006).

slope crest was 0.76 m, which was 62% less than that of pore pressure ratio at PPT7 (close to the crest, see
DS25, even though the magnitude and duration of the Fig. 9) continued to rise and reached its peak of 0.66
shaking were substantially larger and longer, respec- after about 50 cycles of shaking. The sudden reduction
tively, for DSN25 than that for DS25, (see Table 1). of nail force might be attributed to the destruction of
Hence, the presence of soil nails greatly reduced the bond strength between the soil and the nail, resulting
deformation of the sand mass during the shaking. from the significant reduction of shear strength after
Figure 12 shows the typical responses of nail forces the peak as shown in Figure 4.
recorded by strain gauges C3 and D3, which were Table 2 summarises the measured excess pore pres-
located at the middle of nail C and nail D, respec- sure ratios at PPT1-8 in the two tests. For both slopes,
tively (see Fig. 9). Tension (+ve) and compression peak ratios were measured at PPT7 and the values were
forces (−ve) in the nails increased when large defor- 0.80 and 0.66 for slopes DS25 and DSN25, respec-
mation was initiated and excess pore pressure started tively. The measured ratios would be larger if a scaled
to build up at the eighth cycle of the shaking. All forces viscous liquid were used in the tests. With the excep-
reached their peaks at the end of the eighth cycle (or tion of PPT1, the estimated peak ratios for DS25 were
at the beginning of the ninth cycle) when the excess generally larger than those at the corresponding loca-
pore pressure ratio, uw /σv , was about 0.27 and 0.23 tions (PPT2∼PPT7) of DSN25 by up to 100%, despite
at PPT1 and PPT7, respectively, where σv is the esti- the facts that a denser soil, a much smaller peak ampli-
mated effective overburden stress just before failure tude and a shorter duration of shaking were adopted for
and uw is measured excess pore water pressure. At DS25 (see Table 1). These differences demonstrate the
this time, LVDT3, located at the crest of the slope, role of soil nails in reducing the volume contraction of
also reached the end of its travel. As the shaking con- the sand mass during earthquake-induced liquefaction
tinued, the two nail forces decreased and eventually by resisting the relative displacement between the soil
reached their respective ultimate values but the excess and the nails.

165
4 EXCESSIVE SETTLEMENTS OF THICK 300
(a)
CU050 CU100 CU200 CU300 CU400
LOOSE CDG FILL SLOPES 250
M=1.54 ( '=37.8 ) M =1.12 ( =28.2 )

Shear stress, q (kpa)

200
Critical state line
4.1 Monotonic and cyclic behaviour of CDG
150
from Beacon Hill (Ng et al. 2004b) Instability line
100
Prior to the centrifuge model tests, a series of undr- e=0.82
e=0.78

e=0.85
ained monotonic and cyclic triaxial tests on normally 50
e=0.94
consolidated CDG specimens 70 mm in diameter and 0 e=1.05
0 50 100 150 200 250 300 350 400 450
140 mm in height were performed to assist in the Effective mean nor Pa)
interpretation of the centrifuge test results. Figure 3 300
(b)
CU050 CU100 CU200 CU300 CU400
shows the particle size distribution of the well-graded 250
CDG samples obtained from Cha Kwo Ling (CKL),

Shear stress, q (kpa)

e=0.78, =400kPa
200
Kowloon. In the figure, the well-graded CDG taken
from Beacon Hill (BH) is also included for com- 150
e=0.82, =300kPa

parison. The mean particle size, D50 , of the CDG e=0.85, =200kPa
100
from CKL is 1.18 mm and the sample contains about e=0.94, =100kPa
15% fines content. According to the British Standard, 50
e=1.05, =50kPa
BS1377 (1990), CDG can be classified as well-graded 0
5 10 15 20 25 30
silty sand. The maximum dry density and the optimum 0
Axial strain, (%)
water content are 1.82 g/cm3 and 14.4%, respectively.
The triaxial specimens were prepared by moist (c)
tamping at the optimum moisture content. The ini-
tial relative compaction of the specimens was 70%
before saturation. Enlarged lubricated end platens
were used in the tests to reduce the end constraints on
the soil specimens. In the undrained monotonic triaxial
compression tests, the soil specimens were consoli-
dated isotropically to different initial mean effective
stresses before shearing. Figure 13a shows the effec-
tive stress paths of five isotropically consolidated
(d)
undrained compression tests with the initial p rang-
ing from 50 kPa to 400 kPa (corresponding to void
ratios varying from 1.05 to 0.78). The effective stress
path of each loosely compacted specimen is character-
ized by its initial increasing q with decreasing p , due
to an increase in pore water pressure during undrained
shearing resulting from the contractive tendency of the
soil. After a peak is reached, q decreases with a fur-
ther reduction in p until the critical state (M = 1.54,
φ  = 37.8o ) is reached, illustrating the unstable nature Figure 13. Triaxial tests on loose CDG: (a) Stress paths of
of the specimen. By joining the stress origin and the static triaxial tests; (b) Stress-strain relationships of static tri-
peak of each stress path, an instability line (Lade 1992) axial tests; (c) Cyclic triaxial test with CSR = 0.1; (d) Cyclic
can be identified with its slope equal to 1.12, cor- triaxial test with CSR = 0.15.

responding to φins =28.2o . Strain-softening behaviour
with very small liquefaction potential but without any
phase transformation phenomenon was noted in these pressure build-up decreased as the number of cycles
tests (see Fig. 13b). increased, due to the relatively low CSR. Eventually,
During the cyclic tests, a cyclic deviator stress of the pore water pressure ceased to develop further as the
equal magnitude in compression and extension was contractive and dilative tendency of the soil specimen
applied to the specimens. Figure 13c shows a typical balanced out. The total deviator strain developed was
result of CDG (e = 0.821) from a cyclic triaxial test less than 0.2% at the end of the test. On the other
with a cyclic stress ratio (CSR) of 0.1, where CSR hand, for a cyclic test on CDG with CSR = 0.15 (e =
is defined as the single amplitude cyclic shear stress 0.821) as shown in Figure 13d, the pore water pressure
(σd ) divided by twice the initial effective confining accumulated continuously and resulted in a continuous
pressure (σ3 ), i.e. CSR = σd / (2σ3 ). In the test, p decrease in p , illustrating a typical cyclic mobility
decreased monotonically but the rate of the pore water phenomenon (Castro 1969).

166
4.2 Response to rainfall infiltration
(Take et al. 2004)
With the support of the Geotechnical Engineering
Office (GEO) of the Civil Engineering and Devel-
opment Department of HKSAR, collaborative and
complementary centrifuge model tests were carried
out at the University of Cambridge and HKUST. Bulk
samples of CDG taken from BH were delivered to the
two universities for centrifuge model tests.
Figure 14 shows an initially 45o loose fill model
slope. The model was constructed by moist-tamping
with only a minimal compaction effort. To reduce par-
ticle size effects, the fill material was first sieved to
remove all particles in excess of 5 mm in diameter.
To simulate effects of rainfall infiltration by control-
ling water contents (or moisture), the fill slope was
installed in an atmospheric chamber, which was sealed
from the external environment (Take et al. 2004). It is
well known that suction is related to the water contents
in soil pores (Ng & Menzies 2007). In the test, posi-
tive and negative pore water pressures were measured
using a network of miniature PPTs and pore pressure
tension transducers (PPTT), respectively, at each of Figure 15. Observed behaviour of a loose CDG model slope
locations indicated by open circles in the figure. The (Take et al. 2004).
deformation of the model fill slopes was captured by
PIV (White et al. 2003).
During the centrifuge test at Cambridge University,
the model was slowly brought to the testing accelera-
rapidly decreased its void ratio. The biggest pores in
tion of 60 g in 10 g increments. The response of the
the soil shed their pore water vertically downwards
loose fill slope during the self-weight consolidation
into the fill (consolidation), creating an initial suc-
and rainfall (see Fig. 15a) at PPTT1 and displacements
tion distribution that increased with elevation from a
of the crest region (a 32 × 32 pixel patch PIV1 ) are
value of approximately zero at the toe to −25 kPa at
shown in Figures 15b and c, respectively.
the crest (PPTT1 in Fig. 15b). Figure 16 illustrates
As the loose fill material became incrementally
the changes of the initially moist-tamped structure of
heavier, a cumulatively larger percentage of the loose
the model fill at the crest during the test. At 1 g, the
fill no longer supported the increase in total stress and
very loose soil had an initially very open structure
(see Fig. 16a), which consisted of large voids sup-
ported by capillary suction. One such void is circled
in the figure. At 60 g, many of these macro-voids
were observed to collapse (Fig. 16b). However, not
all the voids collapsed. In particular, the voids at low
stress levels (i.e. shallow depths) such as the high-
lighted void in Figure 16a simply settled along with
the fill. The observations of the collapse and the mech-
anisms shown in these two figures cannot be easily
obtained from the field or numerical analyses even
with large-strain formulations.
After the initial self-weight consolidation, the fill
slope was subjected to the equivalent of six weekly
periods of rainfall infiltration (Fig. 15a). As shown in
Figure 15b, the arrival of rainfall on the slope surface
at time A destroyed a significant portion of the soil
suction very rapidly at the shallow location of PPTT1 .
The loose model fill responded immediately to this
Figure 14. Model geometry of a CDG fill slope (Take et al. loss of surface tension by collapsing the macro-voids
2004). that had survived self-weight consolidation (Fig. 16c).

167
Figure 16. The soil structure of the crest region (Take et al. 2004).

The resulting settlement of the crest was purely

vertically downwards.
As the rainfall infiltration continued, the rate of suc-
tion loss decreased (see Fig. 15b), as did the rate of
settlement (Take et al. 2004). Further rainfall infiltra-
tion resulted in the development of a water table at
the toe of the slope. By the time the model had been
subjected to rainfall for an equivalent duration of 1
week (Time B in Fig. 15b), the rate of the pore water
pressure increase diminished to almost zero, becom-
ing asymptotic to a small negative value. At this point,
the mist nozzles were turned off and the fill slope was
allowed to consolidate for an equivalent period of one
week.
Despite being subjected to an additional five infil-
tration events of identical severity, the model slope
was observed neither to achieve positive pore water
pressures nor to experience any significant additional
deformation. Figure 17 shows the spatial distribution
of the settlement vectors, revealing the region of insta-
bility recorded during the first rainfall event. Although
the fill experienced instability, a liquefied flow was not
triggered. The finding is consistent with that reported Figure 17. Observed wetting collapse of the loose model
by Ng et al. (2002b). This was probably because of the fill slope (Take et al. 2004).
small liquefaction potential of the CDG (see Fig. 13b).

horizontal and the initial relative compaction of the

4.3 Response to rising ground water
fill was less than 80%. At 60 g, a 300 mm high model
(Ng et al. 2002b)
slope was equivalent to an 18 m high prototype slope.
To complement the rainfall infiltration tests carried Figure 18 shows the measured displacement vectors of
out at Cambridge, a series of centrifuge model tests on an unreinforced loose CDG fill slope destabilised by
loose CDG fill slopes with and without soil nails was the rise of the ground water. Excessive settlement was
subjected to rising ground water at HKUST (Ng et al. measured but no sign of liquefied flow or slide of the
2002b, Zhang 2006). The CDG fill material used for slope was observed during and after the test. This was
the tests in Hong Kong was also from BH. A model probably because of the small liquefaction potential of
slope was initially prepared to incline at 45o to the the CDG (Ng et al. 2004a). Similar findings are also

168
 18.900 140
Prototype Scale
LVDT-v3 LVDT-v2 LVDT-v1

LVDT-h1 LS-h1
24.240

X
PPT4 ACC4-Y
ACC4-X
ACC3-Y

150
PPT3 ACC5-Y ACC3-X ACC2-Y
PPT2
ACC2-X
0.600 ACC-T-X,Y,Z ACC5-X ACC1-X

30
6.240
ACC1-Y

Unit in metre Scale PPT1

660 Z

Figure 18. Displacement vectors in unreinforced slope

(CG45) (Ng 2007). Figure 19. Configuration of the model slope and instru-
mentation (Ng et al. 2004b).

reported by Take et al. (2004) as described in Section

4.2 of this paper. clay (putty) was sandwiched between a silicon rubber
Tang & Lee (2003) report on a large-scale field trial sheet and the sidewall of the model container. This
on a loose CDG fill slope compacted with an initial set-up was used to reduce reflected stress waves at
angle of 33o to the horizontal. Again, the bulk fill the boundaries in the Y-direction. Five pairs of minia-
material was taken from BH. The height and width ture accelerometers were installed in the embankment.
were 4.75 m and 9 m, respectively. It was constructed Each pair was arranged to measure soil accelera-
by the end-tipping method and resulted in a loose state tions in two horizontal directions (i.e. the X- and
with an initial dry density ranging from 70% to 75% Y-directions). Four miniature pore pressure transduc-
of the maximum dry density. It is considered that the ers were installed in the soil near the accelerometers
stress state of this slope would represent reasonably to record pore water pressures during shaking. On
well that of many existing fill slopes in Hong Kong. top of the embankment, three LVDTs were mounted
The fill slope was subjected to a rise in the ground to measure the crest settlement, and one LVDT and
water and rainfall infiltration. Only excessive settle- one laser sensor (LS) were used to measure horizontal
ment of the slope was recorded and no sign of liquefied movement of the crest.
flow or non-liquefied slide was observed (Tang & Lee To simulate the correct dissipation rate of excessive
2003). Consistent results are obtained between the pore pressures in the centrifuge tests, sodium car-
independent centrifuge model tests at Cambridge and boxy methylcellulose (CMC) powder was mixed with
at HKUST and the field trial. distilled deionized water to form the properly scaled
viscous pore fluid and to saturate the loose CDG model
slopes.
4.4 Response to earthquake loading After model preparation, the speed of the centrifuge
was increased to 38 g. Once steady state pore pressure
4.4.1 Centrifuge model and test procedures condition was reached at all transducers, a windowed
(Ng et al. 2004b, Ng 2007) 50 Hz (1.3 Hz prototype), 0.5 s (19 s prototype) dura-
To further investigate the possibility of flow liquefac- tion sinusoidal waveform was then applied (Ng et al.
tion of loose CDG fill slopes, uni-axial and bi-axial 2004b). After triggering each earthquake, the cen-
dynamic centrifuge tests were carried out using soil trifuge acceleration was maintained long enough to
samples taken from BH (Ng et al. 2004b). Model allow the dissipation of any excess pore pressure. Due
embankments were subjected to shaking ranging from to page limits, only some results from one biaxial shak-
0.08 g to 0.28 g (prototype) in the centrifuge at ing test are discussed here. Other details of all the tests
HKUST. All the models were essentially the same in are presented in Ng et al. (2004b).
geometrical layout and made of loose CDG with the
same initial dry density. Figure 19 shows a typical 4.4.2 Measured responses of the loose CDG
model slope (6 m in prototype) initially inclined at fill slope subjected to bi-axial shaking
30o to the horizontal with its instrumentation. A rigid (M2D-0.3) (Ng et al. 2004b)
rectangular model box was used to contain the CDG Figure 20 shows some measured horizontal accel-
samples compacted to an initial dry density of about eration time histories in the X- and Y-directions
1.4 g/cm3 (or 77% of relative compaction). The insides together with their normalized amplitudes in the
of the container were lined with a lubricated plas- Fourier domain. In the biaxial shaking test, the base
tic membrane to minimize friction of the soil with input accelerations (recorded by ACC-T-X & ACC-
the container walls. For two-directional (X-Y) hori- T-Y as shown in the figure) were 11.26 g (0.28
zontal shaking tests, a layer of 5 mm soft modelling g prototype) and 7.77 g (0.19 g prototype) in the

169
 Ground water level

LVDT Laser sensor

Laser sensor

Figure 20. Seismic acceleration history and Fourier ampli- Figure 22. A typical profile of a loose fill slope after
tude spectrum (M2D-0.3) (extracted from Ng et al. 2004b). shaking (Ng et al. 2004b, Ng 2007).

1.0
1.0
PPT1
PPT1 and PPT2 recorded about the same maximum
PPT2 (Z=100mm) PPT2
0.5
pore pressure ratio of 0.87, whereas PPT4 registered
the smallest of 0.75. These measured values were
PPT1 (Z=145mm) 0.0
0.5 PPT4
less than the theoretical value of 1.0 for liquefaction,
-0.5 even though the pore fluid was correctly scaled in the
0.0 0.2 0.4 0.6 0.8 1.0
Time (s) test. The excess pore pressures dissipated to zero at
about 12 s (6.8 minutes in prototype) after the start of
0.0 shaking.
Figure 22 is a photograph of the model taken after
PPT4 (Z=10mm) the completion of a shaking test. The deformation pro-
file for the slope was similar in both the uni-axial
-0.5 and bi-axial shaking tests. The observed profile of
0 5 10 15
the deformed slope clearly illustrates that no lique-
fied flow and non-liquefied slide took place during
the shaking. The significant difference between the
Figure 21. Measured excess pore-water pressure ratios in observed physical test results from the loose LB sand
bi-axial shaking test M2D-0.3 (Ng et al. 2004b).
and CDG fill slopes may be attributed to the signif-
icant difference in liquefaction potential of the two
materials (see Figs 4a & 13b).
X-direction and Y-direction, respectively. The win-
dowed sinusoid waveform applied in the Y-direction
lagged the X-direction input signal by 90◦ . Recorded
by the accelerometer near the crest, the peak acceler- 5 NON-LIQUEFIED SLIDE MECHANISMS
ation in the X-direction increased 45% at ACC4-X, OF SHALLOW CDG FILL SLOPES
higher than that measured in a corresponding uni-
axial shaking test (Ng et al. 2004b). A similar trend
5.1 Destabilisation of loose shallow CDG fill
of variations in the acceleration was also found in
slopes near the crest (Ng et al. 2007)
the Y-direction. The normalized spectral amplitudes
of acceleration at the predominant frequency of 50Hz The Housing Department of HKSAR has been actively
decreased about 9% in the X-direction but increased looking for innovative methods to preserve the envi-
about 4% in the Y-direction in the upper portion of the ronment by minimizing the need for felling trees when
embankment. improving the stability of existing shallow loose CDG
Figure 21 shows the time history of the excess pore fill slopes. Centrifuge model tests were commissioned
pressure ratios along the height of the model embank- to investigate possible failure mechanisms of loose fill
ment during shaking. Peak acceleration occurred at slopes. Figure 23 shows an instrumented centrifuge
about 0.25 s after the start of shaking. The maxi- model created to study the potential static liquefaction
mum pore pressure ratio occurred at about 0.33 s at of a loose shallow CDG fill slope subjected to a ris-
each of the three transducers (PPT1, PPT2 & PPT4). ing ground water table. The particle size distribution

170
 PPT
Unit: mm

Model box
Upstream drainage
board Coarse soil

PPT1PPT2
Downstream drainage
PPT3 board
LS3
Loose CDG (WTS)
PPT4 Downstream temporary
Upstream temporary LS2 reservior
reservior PPT5
Wood block LS1 Outlet hole
Coarse soil
Inlet hole PPT6

PPT7 PPT8

PPT B PPT9 PPT C

Figure 23. Model package of an instrumented shallow fill

slope (Ng et al. 2007).
Figure 25. Variations in the measured pore water pressure
at the crest (PPT2) and at the toe (PPT7) of the slope over
time (Ng et al. 2007).

could occur but static liquefaction was very unlikely

to happen in the slopes.

5.2 Destabilisation of loose shallow CDG

fill slope at the toe (Take et al. 2004)
Take et al. (2004) also carried out centrifuge model
tests to investigate the possible slide-flow failure
mechanism of a loose thin CDG fill layer. The CDG
used was taken from Beacon Hill. Figure 26 shows
the geometry adopted. The slope angle was 33o . At
30 g, the model corresponded to a fill slope of 9 m in
Figure 24. Top view of the model showing a non-liquefied height, with a vertical depth of fill of 3 m. The chosen
slide (Ng et al. 2007). soil profile for the model fill also represents an ideal-
ized case of layering in which the CDG fill material
has been sieved and separated into its coarse and fine
fractions and placed one on top of the other to form
of the CDG used is denoted as WTS in Figure 3. The a layered backfill. The layer ends blindly at the toe of
initial fill density was 66%. This model was used to the slope to generate elevated transient pore pressures
simulate a 1.5 m thick, 24 m high layered fill slope (Take et al. 2004). This ensures that the rate of arrival
when tested at 60 g. In addition to laser sensors (LSs) of the seepage water at the toe greatly exceeds that of
installed for monitoring soil surface movements, PPTs the leakage, thereby ensuring a more rapid local tran-
were installed to measure excess pore water pressures sient build up of pore water pressures in this region
during the tests. Effects of layering were considered by than would have existed in the absence of layering.
titling the model container during model preparation. In this experiment, the impermeable bedrock layer was
The slope was destabilised by downward seepage cre- modelled by a solid wooden block, the top surface of
ated by a hydraulic gradient, which was controlled by which was coated with varnished coarse decomposed
the water level inside the upstream temporary reser- granite to ensure a high interface friction angle.
voir and the conditions of the outlet hole located The density of the fill material in the first layered
downstream (see Fig. 23). slope model was very loose, with an approximate rela-
Figures 24 and 25 show the occurrence of a non- tive compaction of 77%. After preparation, the model
liquefied slide and the measured excessive pore water fill slope was installed on the centrifuge and slowly
pressure during two failures, respectively. The slide brought to the testing acceleration of 30 g.
was initiated near the crest. Based on the observed Figure 27 shows the arrival of the transient pore
failure mechanisms and the small excessive pore water at the toe of the slope. Once the line source of
water pressures measured, it was concluded that non- seepage water was activated, the high transmissivity
liquefied slide of loose shallow CDG fills slopes of the coarse layer quickly delivered water to the toe

171
 Figure 28. Displacement field prior to final acceleration of
Figure 26. A slide-to-flow landslide triggering mechanism loose fill model (Take et al. 2004).
model (Take et al. 2004).

destabilised by downward seepage in the test for

the Housing Department in which the slope was not
blinded hydraulically at the toe (see Fig. 23). The ini-
tiation of the non-liquefied slides differed in these two
slopes.

5.3 Destabilisation of a dense shallow CDG fill

slope at the toe (Take et al. 2004)
Unlike the static liquefaction mechanism of loose sand
fill slopes, the non-liquefied slide triggering mecha-
nism is argued to be independent of soil density (Take
et al. 2004, Ng 2007). In order to verify this hypothe-
sis, the experiment was therefore repeated with a fill
compacted to 95% maximum Proctor density while all
Figure 27. Observed behaviour of slide-to-flow models other factors remained constant (Take et al. 2004).
(Take et al. 2004). As before, seepage water was introduced to the crest
of the model fill slope and it was quickly transmitted
to the toe of the slope, building up localized tran-
of the fill slope. As intended, the rate of water trans- sient pore water pressures at an identical rate as in
fer into the toe region exceeded the seepage velocity the loose fill model (Fig. 29a). Since the slope mate-
through the model fill, causing a transient increase in rial was dry, the position of the wetting front could be
the pore water pressure at the toe. The local pore water observed. This dense slope exhibited a much stiffer
pressure was observed to increase at a nearly constant response to the build up of pore water pressures, with
rate reaching a maximum value of 16 kPa at point B in less than one half of the pre-failure displacements sig-
Figure 27a. As this seepage front progressed towards naling the onset of failure (see Fig. 27b). Just before
the toe, the slope was slowly creeping (Fig. 27b). reaching the failure pore water pressure, the brittle fill
After time B, the slope mass is observed to accel- material cracked and water rapidly entered the fill. As
erate (points B–C on Fig. 27b). Analysis of this pair high-pressure water entered the crack, the acceleration
of images indicates that, at the onset of more rapid of the slide increased. The extent to which this crack
failure, the toe accelerated horizontally with an aver- injected water into the fill material at time B is shown
age velocity of approximately 6 mm/s (Fig. 28). The in Figure 27a. After time B, the slope mass acceler-
observed displacement field over this time interval ated, although at a slower slide velocity than observed
indicates that the surface of the model fill moved in the loose fill slope (points B–C in Fig. 27b). The
down-slope at a slower velocity. When the fill mate- subsequent behaviour of the model fill slope is laid out
rial finally came to rest, it formed a low-angle run-out. pictorially in the remainder of Figure 29. As the toe
This failure mechanism differs from that of the slope continued to accelerate horizontally, the surface of the

172
 (a) (b)

LVDT

Nailing Model
Drainage system box
board
Mark Nail Actuator

(c) Water
(d) CDG Leading plate Water
Pore water pressure transducers

Figure 30. A typical centrifuge model slope package

equipped with an in-flight soil nailing system (Ng et al.
2002b).

Figure 29. Failure mechanism in the dense fill model

(modified from Take et al. 2004).

model fill accelerated towards the toe (Fig. 29b), with

the velocity increasing to such a point that it exceeded
the shutter speed of the camera (Fig. 29c). Eventually,
the slope came to rest (Fig. 29d). Similarly to in the
shallow loose fill slope, the landslide event triggered
from localized transient pore water pressures formed
a low-angle run-out. The densification of the fill slope
slightly increased the pore water pressure required to
initiate failure (see Fig. 27a), but it made the failure
more brittle (Take et al. 2004).
Figure 31. Comparisons of measured soil displacements
without (CG45) and with soil nails (CGN45) in two centrifuge
6 THE ROLE OF SOIL NAILS IN CDG tests using CDG loose fill at 60 g (dimensions in metres at
FILL SLOPES prototype scale) (Ng et al. 2002b).

6.1 Performance of soil nailed loose fill slopes

Centrifuge model tests on CDG fill slopes subjected fill at Cambridge University and by Tang & Lee (2003)
to rising ground water and rainfall were conducted by from large-scale field tests conducted at Hong Kong
Ng et al. (2002b) and reported by Ng et al. (2003). The University.
main objective of these tests was to investigate static
liquefaction in loose CDG fill slopes and the ability
of soil nails to stabilise these loose slopes. Figure 30
6.2 Effects of soil nails on steep dense CDG fill
shows a model package of a loose CDG fill slope
slopes (Zhou et al. 2006)
equipped with an in-flight soil nailing system. The rel-
ative compaction of the fill model slope was less than 6.2.1 Centrifuge model and test procedures
80%. At 60 g, the 300 mm high, 45◦ model slope was Figure 32 illustrates an instrumented model of a nailed
equivalent to an 18 m high prototype slope. Figure 31 slope centrifuge test (CGN65_30) on model scale.
compares the displacement vectors of the loose CDG A similar model was used for an unreinforced slope
fill slopes obtained from two model tests, one without (CG65_30) except that no soil nails were installed.
and one with soil nails. The soil nails were installed The soil used in the experiment was sieved CDG from
in-flight at 60 g and it can be seen that the soil nails BH with particles larger than 2 mm removed. D10 and
substantially reduced soil movements by at least a fac- D50 were found to be 8 μm and 600 μm, respectively.
tor of 5. No sign of static liquefaction of the slopes was The specific gravity of the CDG was 2.62. The max-
observed during and after the tests. Similar findings imum dry density was 1845 kg/m3 and the optimum
are also reported by Take et al. (2004) from indepen- water content was 14.2% as established by the standard
dent centrifuge model tests using the same loose CDG Proctor compaction test.

173
 PPT
Unit:mm
LVDT1 LVDT2
Model box
Upstream drainage Nail F
board

PPT13 PPT14
Nail F Downstream
PPT12
drainage board
PPT10 Nail E PPT11
Coarse CDG Nail D Downstream Nail E
Upstream temporary temporary reservior
block PPT9
reservior
Outlet hole
Inlet hole PPT7 Nail C PPT8
PPT6
PPT5 Nail B Nail D
PPT1 PPT2 PPT4 PPT3

Nail C
Figure 32. A nailed slope model on the model scale
(CGN65_30) (Zhou et al. 2006).
Nail B

A horizontal bed of CDG was first formed by layer- Figure 33. Positions of instrumented soil nails (front view)
by-layer moist tamping. The mass of soil required in (Zhou et al. 2006).
each layer was calculated based on a dry density of
1753 kg/m3 (relative compaction 95%) at a water con-
tent of about 16% (wet of the optimum). A rectangular 6.2.2 Measured settlement responses and crack
block of coarse CDG was used to facilitate the seepage formations due to rising of the ground water
of water into the slope mass. Silicon grease was used Due to space limitations, only selected results are pre-
on both sides of the container to reduce side friction sented here. Figure 34 shows the relationships between
and preferential water flow at the interfaces. After the measured settlement at LVDT2 and pore pressure at
model was formed, the horizontal bed of soil was cut PPT4 for both the unreinforced and nailed slopes.
to form a slope that was 500 mm high with a slope In the unreinforced slope (CG65_30), initially, set-
angle of 65◦ . The dimensions of a 1/30th-scale steep tlement at the crest was not sensitive to the increase
slope model were chosen to represent the prototype in pore water pressure below the slope toe (O-A). At
behaviour of a 15 m high steep slope when tested point A, the positive pore water pressure indicates a
at 30 g. complete loss of suction. Settlement increased sud-
Holes were drilled for inserting nails into the slope denly at point A, corresponding to the occurrence
mass at 1 g. The inclination of each nail was 20◦ below of the first crack near the slope toe (see Fig. 35).
the horizontal. The nails were arranged in nine rows Then, the settlement gradually increased with the
and in five columns with horizontal spacing of 67 mm pore water pressure. When the pore water pressure
(2.0 m in prototype) and vertical spacing of 50 mm increased to about 14 kPa (point B), the settlement
(1.5 m in prototype) (Fig. 33). Five nails (Nail B to significantly increased again. Point B exactly corre-
Nail F) of the middle column were instrumented with sponds to the observed second crack at the crest. From
strain gauges (Fig. 33). The length of the nails was point B to point C, a slight increase in pore water pres-
400 mm (12 m in the prototype). The modelled nail sure induced a significant vertical deformation at the
was equivalent to a steel bar 32 mm in diameter with crest. This observed deformation-suction relationship
a grouting layer of 205 mm in the prototype. Seven is consistent with results from field stress path triaxial
sets of full-bridge semiconductor strain gauges were tests reported by Ng & Chiu (2003). At point C, the
mounted on each nail to monitor the axial nail forces. third crack occurred below the crest. Then, instabil-
Two LVDTs were used to monitor the settlement ity from the slope toe quickly developed upwards with
of the modelled slope (Fig. 32). In addition, the test increased pore water pressure (PPT4). At point D, the
process and slope deformations were monitored in the third crack spread to the slope surface as indicated by
control room by four digital cameras and four video the formation of a continuous slip plane. After point D,
cameras. the settlement at LVDT2 could not represent the defor-
During a test, the g-level was slowly increased to mation of the slipped mass because of the detachment
the target value of 30 g. Then, the rising water table of the soil mass below the crest (see Fig. 35).
was simulated by allowing water to flow into the stand- In the reinforced slope (CGN65_30), the initial
ing pipe outside the model container on the upstream decrease at PPT4 indicated that consolidation was
side. On the downstream side, the water level was kept possibly still in process before the water flowed near
within 3 m (in the prototype) of the model base. The the toe. Once the water from the upstream reached the
supply of groundwater was stopped when no further slope toe, the pore water pressure started to increase at
failure was observed after over four months of seepage point A . Localized failure between nails near the toe
in prototype time. occurred (see Fig. 36). Then, the crest settlement

174
 0.4

CG65_30
A continuous slip
0.3
Settlement at LVDT2, m

plane was formed

D
0.2 3 rd crack occurred
below the crest CGN65_30

2nd crack occurred

Localized failure at the crest C
0.1
occurred
B'
A'
B Crack occurred at the
1st crack occurred near the toe
O A O' crest
0
-10 -5 0 5 10 15 20 25 30
Figure 36. Side view of the nailed slope (CG65_30) (Zhou
Pore water pressure, kPa
et al. 2006).
Figure 34. Settlements at LVDT2 versus pore water pres-
sure (PPT4) in two models (CG65_30 & CGN65_30) (Zhou
et al. 2006). 250
6 Nail B
3
Nail C
200 2 4 5 Nail D
Nail E
150 Nail F
Axial force, kN

1
100

50 7
Second crack

Third crack A continuous slip plane 0

-50
0 2 4 6 8 10 12 14
Distance from the rear end, m

First crack
Figure 37. Distribution of axial forces at each instrumented
nail when the groundwater stopped rising (CGN65_30) (Zhou
et al. 2006).

Figure 35. Side view of the unreinforced model (CG65_30)

(Zhou et al. 2006). was 0.16 m in the prototype, which was only about
half of that in the unreinforced model. This implies
that the scale of failure is reduced with the use of soil
nails.
increased with the pore water pressure at PPT4 slowly
to about 20 kPa (point B ). After point B , there was
a significant increase of crest settlement, which cor- 6.2.3 Distribution of nail forces
responded to the appearance of a crack at the crest. Figure 37 shows the measured axial nail forces when
In comparison to the measured pore water pressure the groundwater stopped rising. The maximum axial
of 14 kPa in the unreinforced slope, the critical pore tensile force (216 kN) was mobilised at Nail C, which
water pressure at PPT4, corresponding to the ini- was 31% larger than that when the g-level stopped ris-
tiation of the slope instability, increased by about ing. The axial force mobilised at Nail B was about 20%
6 kPa in the nailed slope. This corresponded to a smaller than that at Nail C. The largest mobilized max-
longer duration of seepage prior to the formation of imum tensile force did not occur at Nail B, which was
the crack at the crest. After that, the crest settle- located in the lowest row. This is in good agreement
ment increased significantly with a minor change at with numerical predictions reported by Shiu & Chang
PPT4. However, unlike in the unreinforced model, a (2005). Nails D, E and F mobilised similar but the
continuous slip plane in this test was not observed. smallest axial forces. The measured maximum tensile
When the seepage flow was stopped in the nailed force at Nail F increased by about 220% as compared
slope, the crest settlement in the reinforced model with that recorded when the g-level stopped rising.

175
7 CONCLUSIONS event triggered by high localized transient pore water
pressures at the toe results in a low-angle run-out
By making use of geotechnical centrifuge modelling in both shallow loose and dense CDG fill slopes. A
technology, the deformation and failure mechanisms denser CDG fill slope shows stiffer and more brittle
of sand and CDG fill slopes with and without soil nails responses than does a looser fill slope.
were investigated. Both static and dynamic centrifuge The use of soil nails in steep dense CDG fill slopes
tests were carried out and in-flight rainfall infiltration can prevent the slope from forming a continuous slip
and rising ground water were simulated. Based on the plane and reduce settlements at the crest by about
static centrifuge tests, it can be concluded that static 50%. In addition, the presence of soil nails delays and
liquefaction/fluidization of a loose sand fill slope due minimizes crack formation at the crest and localized
to a rising ground water table was successfully created failures between nails near the toe.
in the centrifuge. The occurrence of liquefaction in
sand was observed by in-flight video cameras and veri-
fied by the significant and sudden buildup of excessive ACKNOWLEDGEMENTS
positive pore water pressures measured at various loca-
tions in the slope. It is found that strain softening of the The work presented here was supported by resea-
material is a necessary but not a sufficient condition rch grants HKUST6053/97E, HKUST6046/98E and
to cause flow liquefaction. A trigger such as seepage CA99/00.EG01 provided by the Research Grants
force or additional loading is needed. Council of the Hong Kong Special Administra-
Dynamic centrifuge model tests on loose sand fill tive Region (HKSAR) and DAG00/001.EG36 from
slopes with and without soil nails reveal that soil nails HKUST. The author is grateful for research contracts
increase the stability of the slope against dynamic liq- provided by the Geotechnical Engineering Office of
uefaction failure (a larger dynamic loading is needed the Civil Engineering and Development Department
to trigger failure). This is because soil nails can resist and the Housing Department of the HKSAR. More-
soil mass movements and reduce the amount of con- over, the author thanks Mr Chen Rui for assisting in
traction of the soil. Measured settlements and lateral checking and formatting the paper.
movements of the sand mass were reduced at least by
62% and 41%, respectively, due to the presence of
soil nails. The maximum excess pore pressure ratio at
the corresponding locations could be reduced by up to REFERENCES
100% at certain locations.
No liquefied flow or slide was observed in thick British Standards Institution (BSI). 1990. BS1377: Methods
loose CDG fill slopes when they were subjected to of tests for soils for civil engineering purposes. BSI,
rising ground water tables and/or heavy rainfall infil- London.
Cai, Z.Y. 2001. A comprehensive study of state-dependent
tration. Only excessive settlements were measured. dilatancy and its application in shear band formation
The centrifuge test results are consistent with the field analysis. PhD thesis, HKUST.
observations from a large-scale field trial of loose Castro, G. 1969. Liquefaction of sands. Harvard Soil
CDG fill slope. With the use of the two-dimensional Mechanics Series 87, Harvard University, Cambridge,
shaking table at HKUST, both one-dimensional and Massachusetts.
two-dimensional earthquakes with up to 0.3 g peak Chu, J. & Leong, W.K. 2002. Géotechnique 52 (10): 751–755.
ground acceleration (prototype) were imposed on Goodings, D.J. & Gillette, D.R. 1996. Model size effects in
loose CDG fill slopes saturated with correctly scaled centrifuge models of granular slope instability. Geotech.
pore fluid. The earthquakes did not liquefy the loose Testing J. 19(3): 277–285.
Hong Kong Government. 1977. Report on the Slope Failures
slopes and they also did not induce any slide but at Sau Mau Ping 25th August 1976. Vol. 1–3, Hong Kong
resulted in excessive positive pore pressures and Government Printer, Hong Kong.
settlements. The significant difference between the Kimura, T. 1998. Development of geotechnical centrifuge
observed physical test results on the sand and CDG in Japan. Proc. Centrifuge 98, Tokyo, Pre-print volume:
models may be attributed to the difference in the fine 23–32.
contents, gradation and liquefaction potential of the Kramer, S.L. 1996. Geotechnical earthquake engineering,
two materials. Prentice-Hall, Inc. New Jersey, USA.
Although static and dynamic liquefaction is very Lade, P.V. 1992. Static instability and liquefaction of loose
unlikely to occur in loose CDG fill slopes, because fine sandy slopes. J. Geotech. Eng. 118(1): 51–71.
Moriwaki, H., Inokuchi, T., Hattanji, T., Sassa, K., Ochiai,
of CDG’s small liquefaction potential, non-liquefied H. & Wang, G. 2004. Failure processes in a full-scale
shallow slides were observed in both loose and dense landslide experiment using a rainfall simulator. Landslide
shallow fill slopes. Depending on the boundary con- 1: 277–288.
ditions, different initiations of non-liquefied shallow Ng, C.W.W. 1999. Applications of Geotechnical Centrifuge
slides were captured in the centrifuge. The landslide Modelling Techniques for Engineering Designs. Proc.

176
 Construction Challenges into the Next Century. Dec., Ng, C.W.W, Zhang, L.M. & Wang, Y.H. 2006a. Proceedings
1999, Hong Kong Institution of Engineers: 241–252. of 6th Int. Conf. on Physical Modelling in Geotechnics.
Ng, C.W.W. 2007. Keynote: Liquefied flow and non- Volumes 1 and 2. Publisher: Taylor & Francis. ISBN:
liquefied slide of loose fill slopes.Proc. 13thAsian 978–0-415–41587–3 and 978-0-415-41588-0.
Regional Conference on Soil Mechanics and Geotechnical Ng, C.W.W., Zhang, E.M. & Zhou, R.Z.B. 2006b. Centrifuge
Engineering, 10–14, Kolkata. Vol. 2. In press. modelling of use of soil nails in loose and dense fill slopes.
Ng, C.W.W & Chiu, C.F. 2003. Laboratory study of loose GEO Report, CEDD of HKSAR.
saturated and unsaturated decomposed granitic soil. Jour- Shiu, Y.K. & Chang, G.W.K. 2005. Nail head review.
nal of Geotechnical and Geoenvironmental Engineering, Geotechnical Engineering Office, HKSAR.
ASCE 129(6): 550–559. Sladen, J.A., D’Hollander, R.D. & Krahn, J. 1985. The lique-
Ng, C.W.W., Fung, W.T., Cheuk, C.Y. & Zhang, L.M. 2004a. faction of sands, a collapse surface approach. Canadian
Influence of stress ratio and stress path on behaviour of Geotechnical Journal 22: 564–578.
loose decomposed granite. J. Geotech. and Geoenviron. Sun, H.W. 1999. Review of fill slope failures in Hong Kong.
Eng., ASCE 130(1): 36–44. GEO Report No. 96. Geotechnical Engineering Office,
Ng, C.W.W., Kusakabe, O. & Leung, C.F. 2003. Theme lec- Hong Kong.
ture: Applications of centrifuge modelling technology in Take, W.A., Bolton, M.D., Wong, P.C.P. & Yeung, F.J. 2004.
geotechnical engineering practice. Proc. of 12th Asian Evaluation of landslide triggering mechanisms in model
Regional Conference on Soil Mechanics and Geotechnical fill slopes. Landslides, Japan, Vol. 1: 173–184.
Engineering, August, Singapore, Vol. 2: 1277–1285. Tang, W.H. & Lee, C.F. 2003. Potential use of soil nails in
Ng, C.W.W., Li, X.S., Van Laak, P.A. & Hou, Y.J. 2004b. loose fill slope: an overview. Proceedings of the Inter-
Centrifuge modelling of loose fill embankment subjected national Conference on Slope Engineering, Hong Kong,
to uni-axial and bi-axial earthquake. Soil dynamics and China: 974–997.
earthquake Engineering 24(4): 305–318. White, D.J., Take, W.A. & Bolton, M. D. 2003. Soil deforma-
Ng, C.W.W. & Menzies, B. 2007. Advanced Unsaturated Soil tion measurement using particle image velocimetry (PIC)
Mechanics and Engineering. Publisher: Taylor & Francis. and photogrammetry. Géotechnique 53(7): 619–631.
ISBN: 978-0-415-43679-3 (Hardcopy). 687p. Zhang, M. 2006. Centrifuge modelling of potentially liq-
Ng, C.W.W., Pun, W.K., Kwok, S.S.K., Cheuk, C.Y. & Lee, uefiable loose fill slopes with and without soil nails.
D.M. 2007. Centrifuge modelling in engineering prac- PhD Thesis, the Hong Kong University of Science and
tice in Hong Kong. Proc. of the Geotechnical Division’s Technology.
Annual Seminar, Hong Kong Institution of Engineers Zhang, M. & Ng, C.W.W. 2003. Interim Factual Test-
(HKIE), 55–68. ing Report I—SG30 & SR30. Hong Kong University of
Ng, C.W.W., Van Laak, P., Tang, W.H., Li, X.S. & Zhang, Science & Technology.
L.M. 2001. The Hong Kong Geotechnical Centrifuge. Zhang, M., Ng, C.W.W., Take, W.A., Pun, W.K., Shiu, Y.K. &
Proc. 3rd Int. Conf. Soft Soil Engineering: 225–230. Chang, G.W.K. 2006. The role and mechanism of soil nails
Ng, C.W.W., Van Laak, P.A., Zhang, L.M., Tang, W.H., in liquefied loose sand fill slopes. Proc. of 6th Int. Conf.
Zong, G.H., Wang, Z.L., Xu, G.M. & Liu, S.H. 2002a. Physical Modelling in Geotechnics (TC2), Hong Kong.
Development of a four-axis robotic manipulator for cen- Vol. 1: 391–396.
trifuge modeling at HKUST. Proc. Int. Conf. on Physical Zhou, R.Z.B., Ng, C.W.W., Zhang, E.M., Pun, W.K., Shiu,
Modelling in Geotechnics, St. John’s, Canada: 71–76. Y.K. & Chang, G.W.K. 2006. The effects of soil nails in a
Ng, C.W.W., Zhang, M., Pun, W.K., Shiu, Y.K. & dense steep slope subjected to rising groundwater. Proc. of
Chang, G. W.K. 2008. Investigation of static liquefac- 6th Int. Conf. Physical Modelling in Geotechnics (TC2),
tion mechanisms in loose sand fill slopes. Re-submitted Hong Kong. Vol. 1: 397–402.
to Géotechnique.
Ng, C.W.W., Zhang, M. & Shi, X.G. 2002b. Keynote (In Chi-
nese): An investigation into the use of soil nails in loose
fill slopes. Proc. of the 1 st Chinese Symposium on Geoen-
vironment and Geosynthetics, 17–19 Nov., Hangzhou,
Zhejiang, China: 61–80.

177
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Capturing landslide dynamics and hydrologic triggers

using near-real-time monitoring

M.E. Reid
U.S. Geological Survey, Menlo Park, California, USA

R.L. Baum
U.S. Geological Survey, Denver, Colorado, USA

R.G. LaHusen
U.S. Geological Survey, Vancouver, Washington, USA

W.L. Ellis
U.S. Geological Survey, Denver, Colorado, USA

ABSTRACT: Near-real-time monitoring of active landslides or landslide-prone hillslopes can provide imme-
diate notification of landslide activity, as well as high-quality data sets for understanding the initiation and
movement of landslides. Typical components of ground-based, near-real-time landslide monitoring systems
include field sensors, data acquisition systems, remote telemetry, and software for base-station data process-
ing and dissemination. For the last several decades, we have used these monitoring tools to investigate different
landslide processes. Some of our field applications have determined the groundwater conditions controlling slow-
moving landslides, detected 3-D displacements of large rock masses, and characterized the transient near-surface
hydrology triggering shallow landsliding.

1 INTRODUCTION high-quality data sets about landslide behavior by

helping to maintain continuity of monitoring during
Most landslide investigations are autopsies of inactive critical periods. If the systems or field sensors mal-
slides or inventories of past slope failures. Such stud- function, they can be quickly repaired to minimize
ies, however, reveal little about the dynamics of active data interruption. In addition, such systems tend to
landslides. Reliable landslide warning systems require promote the evolution of better landslide monitor-
accurate short-term forecasts of landslide activity, ing by identifying the need for additional or different
which in turn demand a detailed understanding of sensors to better detect changing field conditions. The
current field conditions and a quantitative framework resulting data sets are valuable for improved geo-
for interpreting those conditions. This knowledge is technical designs or emergency actions aimed at mit-
difficult to extract solely from landslide postmortem igating landslide hazards. They are also crucial for
studies. advancing scientific understanding of active landslide
Real-time or near-real-time landslide monitoring behavior.
can provide insight into the dynamics of landslide The term real-time monitoring has become com-
initiation and movement. In addition, this monitor- mon in many settings, from finance to computer
ing can provide immediate notification of landslide performance to environmental conditions. However,
activity that may be critical to protecting lives and remote monitoring systems are not truly real time;
property. Displaying current landslide conditions on there is always some delay between sampling condi-
the Internet can be extremely valuable to a wide vari- tions and displaying those conditions to users. Here
ety of end users, including emergency responders, we use the term near-real-time monitoring to designate
land managers, geotechnical engineers, researchers, observations that are delayed slightly (typically min-
teachers, and the general public. These groups may utes to hours) but still close enough in time to represent
have very different uses for near-real-time monitor- the current status of field conditions. The degree to
ing data. Near-real-time systems also tend to ensure which a remote system approaches real time depends

179
on the frequency of 1) data sampling in the field, 2) Automated warning systems may need redundant field
data transmission, and 3) data updates available to sensors, power supplies, and data serving computers
users. to help ensure continuous operation during landslide
Near-real-time monitoring systems have been used triggering events. The desired frequency of data updat-
throughout the world to detect or forecast landslide ing, i.e. how close to real time the system is, can be
activity. In Hong Kong, the USA, and Brazil, regional greatly influenced by the end purpose of the system.
warning systems have been operated to forecast con- 2) The type of landslide to be monitored. Instrumen-
ditions for rainfall-induced shallow landslides, using tation techniques and sampling frequencies to detect
near-real-time rainfall observations (Finlay et al. 1997, rapid debris flows differ greatly from those used to
Ortigao & Justi 2004, Wilson 2005, Chleborad et al. monitor slow-moving landslides. Moreover, designs
2006). Frameworks for similar systems have been to monitor displacement of a currently active slide
developed for mountainous regions of Italy, New can differ from those monitoring the hydrologic condi-
Zealand, and Taiwan (Aleotti 2004, Chien-Yuan et al. tions the might trigger future sliding. 3) The physical
2005, Schmidt et al. 2007). Site-specific, near-real- setting of the field site. Landslides in urban settings
time systems have been applied in many countries to may have access to AC power and readily available
monitor critical structures, such as dams, or hazardous telecommunications, whereas very remote sites may
landslides (e.g. Angeli et al. 1994, Berti et al. 2000, need multiple radio repeaters or satellite links to relay
Husaini & Ratnasamy 2001, Froese & Moreno 2007). data to a secure base-station computer.
Since 1985, researchers with the U.S. Geological Although designs and configurations can vary con-
Survey (USGS) have used near-real-time monitoring siderably, most ground-based, near-real-time landslide
systems for regional warning systems (Keefer et al. monitoring systems have certain components in com-
1987, NOAA-USGS Debris Flow Task Force 2005) mon. These include: 1) sensors on or within the
and for recording the dynamics of hazardous active landslide mass or landslide-prone area, 2) data acqui-
landslides or landslide-prone hillslopes (e.g. Reid & sition systems to sample and control the sensors, 3)
LaHusen 1998, e.g. Baum et al. 2005). a communication system to relay data from the field
In this paper, we discuss some of the design consid- to base-station computers or the Internet directly, and
erations and components typical of ground-based, site- 4) software for data analysis and visualization. An
specific, near-real-time landslide monitoring systems. example system configuration is shown in Figure 1.
We then discuss some USGS applications of such mon- Below, we provide brief overviews of each compo-
itoring systems. Finally, we present three brief case nent with an emphasis on techniques we have found
studies that illustrate monitoring system configura- successful in USGS landslide monitoring.
tions and that document landslide dynamics or hydro-
logic triggering in very different geologic settings.
These studies include: 1) identifying the groundwater
pressures controlling a slow-moving, coastal land-
slide, 2) detecting 3-D displacement of a large rock
block using inexpensive GPS receivers, and 3) cap-
turing the transient, rainstorm-induced, soil-moisture
conditions triggering a shallow landslide.

2 COMPONENTS OF NEAR-REAL-TIME
LANDSLIDE MONITORING SYSTEMS

2.1 System design considerations

The technologies used in ground-based, near-real-time
landslide monitoring systems have evolved rapidly in
recent years; the development of new sensors, low-cost
methods of telemetry, and new software for data dis-
semination have made this type of monitoring readily
available and affordable. However, there is no stan-
dard setup that will work for all landslide monitoring.
The design and implementation of near-real-time sys- Figure 1. Example of a ground-based, near-real-time land-
tems depends on considerations that vary from site to slide monitoring system. Typical components include field
site, including: 1) The end purpose of the monitor- sensors, field data acquisition system, remote communica-
ing. Systems for public safety often differ from those tions (here via radio telemetry), data processing (on a base
intended to record data for scientific research studies. computer) and data dissemination (via the Internet).

180
2.2 Field sensors or installed in a cased borehole, whereas tensiometers
and moisture sensors typically need direct soil contact.
For ground-based monitoring, detecting and measur-
ing landslide activity typically requires an array of
sensors placed on the ground surface and/or in the 2.3 Data acquisition systems
subsurface at the slide. For USGS monitoring, we
have used both commercially available and USGS Remote sites usually require on-site data acquisition
custom-built sensors. A wide variety of commercial- systems to power and control sensor sampling, log
off-the-shelf electronic and/or mechanical geotech- sensor data, and interact with the remote communica-
nical and hydrologic sensors exist (Dunnicliff 1993, tion system. Site conditions and system configurations
Mikkelsen 1996). For remote sites, sensors need to often dictate the number of data acquisition systems
be rugged, weather resistant, portable, and have low required; most systems can handle multiple sensors.
power consumption; power is often the most criti- Seismic and GPS instruments can require relatively
cal issue. Sensors also need to easily interface with high frequency sampling (0.1 to 100 Hz), whereas
data acquisition systems and have adequate sensitiv- most geotechnical and hydrologic sensors are sampled
ity and resolution. Very-high precision instruments, at relatively low frequency (minutes to hours). USGS
common in scientific laboratories, are seldom needed landslide monitoring has used both commercially
for field monitoring. Instead, relatively inexpensive, available data loggers and USGS systems custom-
yet precise, sensors are preferred because they may be developed for monitoring active volcanoes (Hadley &
destroyed by landslide activity. In addition, low-cost LaHusen 1995). A reliable power supply is an abso-
sensors can allow more units to be deployed. Often, lute necessity for each acquisition system, as well as
sensor arrays with more complete spatial coverage, for the remote communication system. We commonly
both on the ground and in the subsurface, are more use batteries and solar panels to supply power at remote
useful than a few high-precision sensors. sites; AC power is sometimes available at urban sites.
Monitoring sensors typically provide two types of For sites without AC power or adequate solar exposure,
information: 1) actual displacement of a landslide we often use air-alkaline batteries.
or debris flow or 2) environmental conditions that
affect slide activity. Slow-moving landslide displace-
2.4 Remote communications
ment can be measured using surface or subsurface
extensometers, tiltmeters, ultrasonic or laser distance To provide near-real-time data updates from remote
meters, radar (Tarchi et al. 2005), digital cameras, monitoring systems, some form of dependable com-
videos cameras, or downhole inclinometers. Although munication system is needed. There are many options
surface cable extensometers are relatively inexpensive, using either dedicated telemetry or commercially
they are particularly subject to disturbance by weather, available services. Typically, the crucial link is
animals, and local ground failure. Subsurface defor- between the remote monitoring stations and a secure
mation can also be detected using grouted-in TDR base station with Internet access or a dedicated tele-
(Time Domain Reflectometry) cables (Kane & Beck phone line. This remote link can be provided by
1996). Measurements from Global Positioning System radio transceivers, satellite uplinks, or telephone ser-
(GPS) receivers located on a slide can be processed rel- vices. The choice of a communication component
ative to a known reference station to provide sub-cm depends, in part, on site remoteness, power avail-
positions and 3-D displacements (Kramer & Rutledg ability, the frequency of data transmission desired,
2000, LaHusen & Reid 2000). Rapidly moving debris reliability, data throughput, and recurring expense.
flows can be detected with tripwires, geophones, or Low-power radio transceivers can use either a fixed
flow-height sensors (LaHusen 2005). frequency or license-free spread-spectrum technol-
Because most landslide and debris flow activity is ogy, but line-of-site transmission may require repeater
triggered by hydrologic conditions, it is often impor- stations. Meteor burst radio communications can be
tant to detect changes in subsurface pore-water pres- used over long distances for low data-rate applica-
sures. Furthermore, it may be desirable to understand tions. Satellite uplinks may use a dedicated service
the hydrologic processes modifying those pore pres- such as GOES (Geostationary Operational Environ-
sures, including the infiltration of rainfall or snowmelt. mental Satellite). Commercial vendors can provide
Commonly used hydrologic sensors include precipita- satellite phone modems or a fixed Internet address
tion gauges, tensiometers and dielectic soil-moisture accessible through a low-cost VSAT (Very Small
probes for measuring unsaturated soil conditions, and Aperture Terminal) satellite ground station. Although
piezometers for recording positive pore pressures. For satellite uplinks may reach many remote areas, their
automated recording, tensiometers and piezometers transceivers can require more power than line-of-sight
can be fitted with transducers for measuring fluid pres- radios and their use can incur service charges. Tele-
sure. Piezometers can be directly buried in a sand phone services, either land line, cellular or satellite,
pack, grouted in place (Mikkelsen & Green 2003), can be reliable options for low data-rate transmission.

181
However, they entail recurring service charges and pressures, and soil suction were transmitted using
may not be available at remote sites. For monitoring ALERT system radio telemetry. Starting in 1997, the
where close to real-time response and/or high data- Cleveland Corral landslide, threatening U.S. Highway
rate transmissions are required, we often use dedicated 50 in California, was our first monitoring site with
radio telemetry. Importantly, battery-powered radio automated data dissemination via publicly accessible
links are usually very reliable during stormy weather web pages on the Internet (Reid & LaHusen 1998, Reid
when landslides may be active. AC power or telephone et al. 2003). Since then, the USGS has operated many
communications may fail during these stormy periods. other near-real-time monitoring sites. With the excep-
tion of La Honda, all of the systems listed in Table 1
use or used USGS web servers to disseminate data
2.5 Data processing and dissemination
over the Internet, typically with updates at a frequency
After remote monitoring data are collected and similar to that of the listed data transmission. Our
relayed to a protected base-station computer, addi- sampling, transmission, and update frequencies were
tional actions are needed to provide information to end selected to capture changes in the physical processes
users. Base-station computer actions typically include occurring in the field (e.g. movement, rain infil-
receiving the data, processing the data if needed, cre- tration) while minimizing field station power usage.
ating graphs and/or tables, and archiving the data. Sites with geophones, such as the Cleveland Corral
Remote data transmission may be controlled by the landslide, scan data every second and transmit imme-
field acquisition system or by the base-station com- diately if ground vibrations exceed a chosen threshold;
puter. There are a variety of software options for thus these sites are closer to true real-time monitor-
performing these tasks, including commercially avail- ing. USGS researchers have also played key roles in
able software packages, some using Open Process designing and installing near-real-time landslide mon-
Control (OPC) protocols (http://opcfoundation.org), itoring systems to monitor alpine debris-flow activity
that can handle real-time data flow and processing. in Italy (Berti et al. 2000) and volcanic debris flows at
For many USGS systems, we use custom-written Ruapehu Volcano in New Zealand.
base-station software controlled by automated batch Below, we briefly present three USGS case studies
processing. Graphs can be generated, using commer- using near-real-time monitoring that illustrate some
cial or license-free software, at specified intervals or of the advantages and complexities involved. Each
in response to user requests. Once graphs and tables study examines a different type of slide, uses differ-
are created, they can be disseminated via a user’s ent monitoring instrumentation and communication
local computer network, or commonly, on web pages telemetry, and addresses different scientific questions.
with public or password protected access. We typi- In particular, we focus on how near-real-time moni-
cally use USGS web servers to disseminate monitoring toring provides crucial insight into different landslide
information. triggering and behavior. Each of the three cases is
located near the Pacific Coast of the USA (Fig. 2)
where rainfall-induced landslide activity occurs pri-
3 USGS APPLICATIONS OF NEAR-REAL- marily during the winter/spring-wet season. Brand
TIME LANDSLIDE MONITORING names for sensors and data acquisition systems are pro-
vided for descriptive purposes only and do not imply
Over the last several decades, researchers with the endorsement by the USGS; other vendors can provide
USGS have used monitoring systems to understand similar equipment.
both the dynamic behavior of individual slides and
the hydrologic conditions triggering widespread land-
sliding. Many of these efforts involved remote data
3.1 Case study 1: Identifying groundwater controls
acquisition and some invoked periodic transfer of data
on the motion of a slow-moving landslide,
via cellular telephone service. (Ellis et al. 2002).
Newport, Oregon
USGS automated, near-real-time landslide monitoring
sites are listed in Table 1 with a brief summary of their Most landslide movement is activated or reactivated
field sensors, data acquisition systems, and remote by increased pore-water pressures acting on a slide’s
communication set-ups. Publicly accessible monitor- slip surface (Terzaghi 1950, Sidle & Ochiai 2006).
ing data from USGS systems currently in operation can These pressure increases can result from many pro-
be viewed at http://landslides.usgs.gov/monitoring. cesses. (e.g. Reid & Iverson 1992, e.g. Iverson 2000);
Early USGS monitoring efforts at La Honda, understanding the timing and pathways of subsurface
California contributed to a San Francisco Bay regional water flow leading to landslide movement is crucial to
landslide warning system that operated between 1985 forecasting future slide behavior, developing warning
and 1995 (Keefer et al. 1987, Wilson 2005). Here, strategies, and designing effective mitigation mea-
near-real-time observations of rainfall, shallow pore sures. Our first brief case study illustrates the use

182
Table 1. USGS near-real-time landslide monitoring sites.

Communication
system and
Location and Data acquisition transmission
period of operation Type of slide Field sensors system∗∗∗ frequency

La Honda, California Shallow earth Rain gauges, Sierra Misco Radio network
(1985–1995) slide∗∗ piezometers, ALERT with repeater
tensiometers, system (15 minutes)
extensometers
Cleveland Corral Translational Rain gauges, geophones, USGS custom Radio network
landslide, earth slide piezometers, system with repeater
U.S. Highway 50, extensometers (15 minutes)
California
(1997-present)
Woodway, Washington Rotational debris Rain gauge, piezometers, Campbell CR10X Telephone
(1997–2006) slide extensometers data logger (15 minutes)
Rio Nido, California Earth slide Rain gauge, geophones, USGS custom Radio network
(1998–2001) piezometers, system with repeater
extensometers (10 minutes)
Headscarp of Mission Rock block slide L1-GPS receivers, Environmental Cellular telephone
Peak landslide, extensometers, air Cellular initially, initially, then
Fremont, California temperature sensor then USGS spread-spectrum
(1998-present)∗ custom system radio network
(30 minutes
or hourly)
Edmonds, Washington Shallow Rain gauges, soil tempera- Campbell CR10X Radio network
(2001–2006)∗ translational ure probe, soil-moisture data logger (hourly and
earth slide∗∗ profilers, tensiometers, 15 minutes)
piezometers
Everett, Washington Shallow earth Rain gauge, water- Campbell CR10X Radio network
(2001-2006) slide∗∗ content reflectometers, data logger (hourly and
piezometers 15 minutes)
State Route 20, Rock block slide Geophones, tiltmeters, USGS custom Radio network
Newhalem, extensometers system repeater
Washington with
(2004–2005) (15 minutes)
Johnson Creek Translational slide Rain gauge, downhole Campbell CR10X Cellular telephone
landslide, extensometers, piezome- data logger (daily)
Newport, Oregon ters, soil-moisture sensors,
(2004-present)∗ air and ground temperature
sensors
Florida River landslide, Ancient translational Rain gauge, extensometers, Campbell CR1000 Cellular telephone
Durango, Colorado slide and recent tiltmeters, and CR200 data (hourly)
(2005-present) debris slides in piezometers, air loggers with radio
wildfire burn area temperature sensor network
Ferguson rockslide, Rock block slide L1-GPS receivers, USGS custom Spread-spectrum
near Yosemite Natl. geophones system radio network
Park, California with repeater
(2006-present) (hourly)
Portland, Oregon Shallow earth Rain gauges, tensiometers, Campbell CR1000 Cellular telephone
(2006-present) slide∗∗ piezometers, soil-moisture data logger (15 minutes)
sensors
∗ Monitoring at this site is discussed further in a case study.
∗∗ Instruments monitor(ed) hydrologic conditions in landslide-prone hillslope. Slide occurred at end of monitoring at
Edmonds site.
∗∗∗ Brand names are provided for descriptive purposes only and do not imply endorsement by the USGS.

183
 Figure 3. Map of the Johnson Creek landslide showing
major structural features, location of rain gauge, and the three
sites of grouped instrumentation boreholes. Slide motion is
west towards the Pacific Ocean. Modified from Landslide
Technology (2004).

Industries and the Oregon Department of Transporta-

tion. Our efforts were focused on understanding the
Figure 2. Map of western USA showing the locations of our subsurface pore pressures controlling movement as
three case studies (Newport, Fremont, Edmonds) illustrating well as the timing of and processes creating elevated
near-real-time landslide monitoring. pore pressures on the basal slip surface. Earlier inves-
tigators installed a rain gauge and pairs of borings at
three sites along a longitudinal section of the slide
of near-real-time monitoring at a slow-moving land- (Fig. 3). At each site, a vibrating-wire piezometer
slide to identify the relations between rainfall, pore (Slope Indicator) had been installed just above the
pressure, and slide movement. slide plane in one boring and inclinometer casing
in the companion boring. Wire-rope extensometers,
anchored beneath the landslide’s basal slip surface
3.1.1 Background and setting had also been installed in the inclinometer borings.
Many large, episodically active landslides disrupt U.S. (Landslide Technology 2004). We added electronic
Highway 101, the major north-south transportation cable extensometers (Celesco) to the down-hole, wire-
corridor that links towns along the Pacific Ocean coast rope extensometer cables for automated recording. In
of Oregon, USA. The Johnson Creek landslide, near November 2006, we installed two vertical arrays of
Newport, Oregon, has a history of repeated move- six vibrating-wire piezometers each (Slope Indicator)
ments during winter rainy seasons, and frequently between depths of 3 m and 26 m at sites 1 and 2 within
impacts the highway. This translational slide, about the central and upper parts of the landslide (Fig. 3) so
200 m long, 360 m wide, and 26 m thick, occurs in sea- that both the lateral and vertical distribution of pore
ward dipping (15–20◦ ) siltstone, sandstone, mudstone pressures could be monitored. These were installed in
and tuffaceous claystone of the Astoria Formation and grout and are capable of measuring unsaturated soil
is located on a nearly flat Pleistocene marine terrace. suctions as well as positive pore pressures. We also
Total landslide displacement is about 28 m horizon- installed two sets of dielectric soil-moisture content
tal and 6 m vertical. (Priest et al. 2006). The largest sensors (Decagon Devices) at shallow depths of 1.5 m
recent movement episode occurred between January and 3 m at sites 1 and 3 to assess the contribution of
2002 and February 2003, when the central part of the vertical rainfall infiltration to pore pressure changes at
slide moved about 25 cm horizontally and dropped depth within the slide (Schulz & Ellis 2007). Two data
several cm vertically. (Landslide Technology 2004). loggers (Campbell Scientific), powered by batteries
and solar panels, record data in 15-minute intervals.
3.1.2 Near-real-time monitoring These data are transmitted automatically every 24
Beginning in November 2004, the USGS installed hours using cellular-telephone telemetry, graphed on
near-real-time monitoring at this site, in cooperation a USGS base-station computer, and placed on a USGS
with the Oregon Department of Geology and Mineral website for viewing.

184
3.1.3 Results
Our monitoring between 2004 and 2007 (Ellis et al.
2007), spanning both dry and wet years, showed that
basal-shear pore pressures begin to increase within
just a few hours following rainfall events. Monitor-
ing also indicated that landslide movement initiates
when pore pressures exceed a threshold, with some
minor variability (Fig. 4). Our monitoring also shows
that rainfall-induced pore-pressure increases travel
from near the headscarp (site 3) westward toward
the toe of the slide (site 1), and that the travel time
of the pore-pressure pulses decreases significantly
with increased antecedent pore-pressure conditions
(Fig. 4). Following rainstorms, there are almost simul-
taneous increases in pore pressures at all depths within
Figure 5. Pore-pressure head at various depths in the John-
each vertical array of piezometers located beneath
son Creek landslide at site 1 between December 2006 and
the water table while shallower unsaturated zone May 2007. Grey lines represent negative pore pressures from
responses lag (Fig. 5). This suggests that rapid pore- three different depths in the unsaturated zone; black lines rep-
pressure increases at depth within the slide do not resent pore pressures at four different depths in the saturated
result directly from vertical infiltration of rainfall, zone. Vertical dashed line shows timing of similar response
but are likely due to lateral pore-pressure response in the saturated zone; response is delayed in the unsaturated
from the headscarp graben area. Our observations zone.
demonstrate that enhanced forecasting of slow-moving
landslide activity requires detailed knowledge of the
links between pore-pressure response and movement, from slow creep to rapid movement. For some fail-
and that inferences based on other landslide studies ures, the timing of rapid movement can be predicted
may be inadequate. Near-real-time monitoring can by projecting the trend of 1/velocity (Fukuzono 1990,
provide the required information. Petley et al. 2002, Petley et al. 2005). Thus, detecting
slide displacement in near real time is crucial to most
forecasting efforts. The following brief case study
examines the use of inexpensive, single-frequency
3.2 Case study 2: Detecting 3-D movement using (L1) GPS receivers to detect the 3-D displacement of
inexpensive GPS receivers, Fremont, California a large rock block having the potential to fail rapidly.
Predicting the timing of rapid, catastrophic failure of
landslides and rockslides is a long sought after goal 3.2.1 Background and setting
of landslide science (e.g. Saito 1965, Varnes 1983, More than 75, 000 ancient and dormant landslides are
Voight 1989). Most forecasting approaches rely on scattered throughout the hills of the San Francisco Bay
detecting the acceleration associated with a transition region, California, USA (Pike 1997). During the wet,
El Niño influenced winter and spring of 1998, the large
(35 hectare) Mission Peak landslide reactivated, and
moved more than 5 m (Geolith Consultants 2000). The
slip surface of this 1.2 km long, 30–55 m thick, histor-
ically dormant earthflow is primarily in the clay-rich
Orinda Formation (Geolith Consultants 2000). The
reactivated slide is a small part of a much larger ancient
landslide complex located above the City of Fremont
in a tectonically active region, with on-going vertical
uplift. Near the head of the slide, the seismically active
Mission Fault crosses the slope. Upslope of this fault,
the steep (45◦ ) headscarp is composed of relatively
competent Briones Sandstone (Graymer et al. 1995)
that dips backward into the slope and contains several
persistent joint sets. The headscarp area shows geo-
Figure 4. Basal shear zone pore-pressure head, rainfall, and morphic evidence of prior large massive rock block
landslide movement (from downhole extensometers) at the failures (Fig. 6), as well as small sackung features,
Johnson Creek landslide between November 2004 and May suggesting that both slow and rapid movements are
2007. Site locations are shown in Figure 3. possible in this setting.

185
 2000). In February 2000, we installed a working pro-
totype of this system in the headscarp area, using
NovAtel GPS receivers, Micropulse GPS antennas,
and a USSG-designed data acquisition and controller
system. To obtain sub-cm measurements, we uti-
lize very short baseline, static differential processing
of GPS observations from two antenna/receiver sta-
tions, one located on the moving rock block and
another located off the block, about 67 m away. Power
for these two remote stations is supplied by solar
panels and batteries. The GPS antenna on the mov-
ing rock mass is located near the outer edge of the
block (Fig. 6), just upslope of the headscarp, to mea-
sure strain across the entire block (relative to the stable
GPS receiver) and to increase the likelihood of detect-
Figure 6. Photograph of active rock mass in headscarp of ing a rapid failure. Instead of continuously operating
the Mission Peak landslide, showing large tension crack and the GPS receivers, which use significant power, we
adjacent older failure scar. Locations of two extensometers employ a novel scheme of powering the receivers on
and two GPS stations also shown. Photo: Phil Stoffer, USGS. and off with a variable duty cycle controlled by base-
station computer software. Typically, we collect 30
minutes of GPS observations at 10-second intervals,
In March 1998, the main earthflow was active and transmit these data using 900 MHz spread-spectrum
threatened homes at the toe of slide. During this time, radio transceivers, and then power down the system
we observed disturbance of a pre-existing, prominent for the next 30 minutes. Independent, high-precision,
tension crack (previously open about 1.5 m) in the static GPS solutions, with fixed ambiguity resolutions,
headscarp area (Fig. 6). Our quadrilateral measure- for each 30-minute observation period are automat-
ments of 3-D displacement across the crack showed ically computed on the base-station computer using
continued movement and prompted concerns about GPS processing software (Waypoint). Results are then
potential catastrophic failure of a large rock block, automatically graphed and placed on a USGS website
partially bordered by this tension crack. The esti- for viewing.
mated volume of this block ranges between 50,000 and
170,000 m3 , depending on inferred thickness (Geolith
Consultants 2000). Rapid failure of the remaining 3.2.3 Results
entire rock mass might result in a rockfall avalanche. Although our monitoring between 1998 and 2007 did
Subsequent analyses of such an avalanche show poten- not record rapid, catastrophic failure, it did demon-
tial maximum runouts of about 500 m along slope strate the ability of our L1-only GPS system to detect
(Jurasius 2002). 3-D, sub-cm movement and accelerations of the rock
mass during wet seasons. More than 40 cm of downs-
3.2.2 Near-real-time monitoring lope motion of the block, measured by our lower
USGS monitoring at this site focuses on measuring extensometer crossing the large tension crack prior
surface displacement of rock in the headscarp region to installation of the GPS system, occurred during
where acceleration might be a possible precursor to the wet 1998 season. Subsequently, between 2000
rapid failure. Over time, our monitoring tools have and 2007 our differential GPS system showed long-
evolved to better identify the 3-D strain across the term northward, westward, and downward creep of
entire rock block as well as the time history of move- the block, resolved into 3-D components in Figure 7.
ment. Initially in March 1998, we used manually Creep might be expected because the GPS antenna
surveyed quadrilateral monuments located across the is located near the headscarp free face. During the
large tension crack. Over the next several months, we relatively wet springs of 2000 and 2006, the block
installed two surface cable extensometers (UniMea- accelerated slightly but then slowed during the fol-
sure) to record downslope displacement across tension lowing summers (Fig. 7). It did not exhibit creep to
cracks (Fig. 6), first using a data logger (Campbell rapid failure. Nevertheless, our observations indicate
Scientific) and later using cell-phone communications that significant movement of the block is related to wet
(Environmental Cellular). years or sequential wet years.
We then developed a low-cost, single-frequency When the block was active during the spring of
(L1) GPS receiver system designed for automated data 2000, we measured movement with both extensome-
acquisition, rapid deployment, and prolonged oper- ters and the differential GPS system (Fig. 8). The
ation in remote hazardous areas (LaHusen & Reid upper extensometer recorded about 5 cm of downslope

186
 movement, the lower extensometer about 2 cm, and the
GPS about 6 cm of cumulative slope displacement.
Detrended GPS solutions have a standard deviation of
about 2 mm in the horizontal and 4 mm in the verti-
cal. Our monitoring illustrates the advantage of using
high-precision GPS position solutions to detect 3-D
strain across larger areas than can be readily measured
using extensometers.

3.3 Case study 3: Capturing shallow landsliding

triggered by rainstorms, Edmonds, Washington
Shallow landslides and accompanying debris flows
pose a frequent and often devastating hazard world-
wide (Iverson et al. 1997, Jakob & Hungr 2005,
Sidle & Ochiai 2006). These slides typically occur in
thin colluvium and are often induced by intense rain-
storms or rapidly melting snow. Although such occur-
rences are exceedingly common, well-documented
Figure 7. 3-D components (northing, easting, and verti- field examples of the subsurface hydrologic condi-
cal) of the high-precision, differential GPS solutions for
the Mission Peak active rock block between February 2000
tions controlling slide initiation are rare. Some studies
and September 2007. Points shown are 5-point medians of have demonstrated that transient positive pore pres-
the independent static solutions for each 30-minute satel- sures in shallow saturated zones trigger failure (e.g.
lite observation period. Cumulative displacement is since Sidle & Swanson 1982, Reid et al. 1988), whereas oth-
installation in 2000; overall the block is moving northward, ers have inferred that suction changes in unsaturated
westward, and downward. Cumulative yearly rain begins materials might instigate failure (e.g. Wolle & Hachich
October 1 of each water year. 1989, Collins & Znidarcic 2004). Understanding the
near-surface transient hydrologic conditions control-
ling shallow failure is crucial to developing reliable
forecasting or warning systems. In our final brief case
study, we illustrate how near-real-time monitoring can
identify the transient, shallow subsurface hydrologic
conditions triggering a shallow landslide.

3.3.1 Background and setting

Shallow, rapidly moving landslides occur almost every
winter along steeper sections (45–60◦ ) of the coastal
bluffs of Puget Sound between Seattle and Everett,
Washington, USA (Baum et al. 2000). Although most
of these slides are less than 1000 m3 (Baum et al.
2000), they pose a continuing threat to public safety in
this area, including disruption of a railway at the base
of the bluffs and destruction of homes, other structures,
and utilities on the bluffs. Rainstorm events producing
one or more landslides have an average recurrence of
six times per year (Chleborad et al. 2006). In cooper-
ation with the Burlington Northern Santa Fe (BNSF)
Railway, we selected a coastal bluff near Edmonds,
Washington (20 km north of Seattle) for detailed mon-
itoring of subsurface hydrologic conditions (Fig. 9).
The purpose of our monitoring in this area was not to
Figure 8. Daily rainfall, extensometer and GPS displace-
ment measurements at Mission Peak during the 1999-2000
provide warning of individual landslides, but rather to
wet season. Extensometers have cables oriented downslope; determine when subsurface conditions are wet enough
therefore smoothed GPS displacement is computed along to make the slopes highly susceptible to landslides.
slope using the 3-D components. GPS data start in February During heavy rains in 1996–1997, several shallow
2000. landslides occurred near this site in weathered glacial

187
deposits and colluvium (Baum et al. 2000). The 50-m- tensiometers (Soil Moisture Equipment Corp.), rang-
high bluff that we selected for monitoring is underlain ing in depth from 20 cm to 150 cm, to measure soil
by subhorizontally bedded glacial and interglacial sed- suction (Baum et al. 2005). The soil-water instruments
iments. A 3-m-thick layer of glacial till caps the bluff; were installed in dense glacial outwash sand and col-
beneath the till is a layer of glacial advance out- luvium about 25–35 m above sea level (Fig. 9). Data
wash that overlies dense glaciolacustrine silt (Minard were relayed every hour using line-of-sight radio
1983). Mechanical weathering of the dense, uni- telemetry to a server at Meteor Communications, then
form, medium outwash sand produces a loose sandy received, reduced, and graphed on a USGS base-
colluvium mantle that covers much of the lower bluff. station computer, and finally placed on a USGS
website for viewing.
3.3.2 Near-real-time monitoring
USGS monitoring at this site, which operated on AC 3.3.3 Results
power, focused on identifying the transient subsurface On January 14, 2006, a shallow landslide occurred
hydrologic conditions triggering shallow failure. We at the Edmonds site, destroying much of the instru-
experimented with various kinds of sensors here in an mentation. However, we measured the near-surface
effort to find a combination that provided hydrologic hydrologic conditions through the previous three wet
monitoring data of sufficient quality, reliability, and seasons and just prior to failure. Our near-real-time
relevance to be suitable for forecasting landslide activ- monitoring revealed several relations between rain-
ity (Baum et al. 2005). Between September 2003 and fall, soil moisture, pore pressure, and the occurrence
January 2006, our remote station at the Edmonds site of shallow landslides in the Seattle area, including:
was equipped with sensors to monitor both unsaturated 1) The timing and magnitude of soil moisture/pore
and saturated volumetric soil-moisture contents and pressure response from rainfall is highly dependent on
pore-water pressure or suction (Fig. 9). We installed antecedent soil moisture. For example, during October
two adjacent tipping bucket rain gauges, two water- 2003, the soil was dry and wetting fronts moved
content profilers (Sentek EnviroSMART) equipped slowly in response to rainfall (Fig. 10). As soil wetness
with eight (soil capacitance) sensors each at depths increased throughout the winter season, pore pressure
ranging from 20 cm to 200 cm, and two nests of six and soil wetness at depth responded much more rapidly
to heavy rainfall. For example, heavy rainfall in mid-
October produced an increase in soil moisture at 2 m
depth after 6 days, whereas in mid-November, heavy
rainfall resulted in a similar increase after only 1 day
(Fig. 10). 2) The pattern of soil-moisture response
was consistent with vertical downward infiltration,

Figure 10. Soil-moisture response to rainfall at the

Edmonds site between October 2003 and January 2004. Sen-
Figure 9. Map of Edmonds monitoring site near Seat- sor depths (cm) are indicated next to response curves. Num-
tle, Washington showing instrument locations and extent of bers of landslides that occurred on each date are indicated in
January 2006 shallow landslide. parenthesis beside dashed vertical lines.

188
 being monitored and the end purpose of the monitor-
ing system, however most ground-based systems are
composed of field sensors, field data acquisition sys-
tems, remote communications, and base-station data
processing and dissemination, often over the Internet.
We have used near-real-time monitoring to under-
stand the dynamic behavior and hydrologic conditions
triggering different types of landslides. Our investiga-
tions include identifying the groundwater conditions
controlling slow-moving landslides, detecting 3-D dis-
placements of large rock masses, and documenting the
transient near-surface hydrology triggering shallow
landsliding. Knowledge of both current field condi-
tions and likely future behavior are crucial to develop-
ing better landslide forecasting and warning systems.
Near-real-time monitoring systems can provide the
current field conditions, but more work is needed on
Figure 11. Soil-moisture response to rainfall at the techniques to rapidly forecast future landslide behavior
Edmonds site between October 2005 and January 2006. A based on these near-real-time observations.
shallow landslide destroyed the instrumentation on January
14, 2006. Numbers of landslides that occurred on each date
are indicated in parenthesis beside dashed vertical lines.
ACKNOWLEDGEMENTS

rather than lateral flow. 3) Landslide occurrence was Many people and agencies contributed to the landslide
strongly correlated to wet antecedent soil conditions. monitoring described in this paper. We particularly
Intense rainfall that occurred in October 2003, Novem- thank Dianne L. Brien, Jonathan W. Godt, Jonathan
ber 2003, and November 2005 (Fig. 11) caused very P. McKenna, and William H. Schulz of the USGS for
few landslides because soil was relatively dry prior their multi-faceted assistance.
to these storms. However, storms of moderate inten-
sity during January 2006, when soil was relatively
moist after many successive days of rainfall, caused REFERENCES
moderate numbers of landslides, even though the
rainfall did not exceed our empirical rainfall thresh- Aleotti, P. 2004. A warning system for rainfall-induced shal-
old. (Baum et al. 2005, Godt et al. 2006). After low failures. Engineering Geology 73 (3–4): 247–265.
several weeks of rain, the landslide at our monitor- Angeli, M.G., Gasparetto, P., Menotti, R.M., Pasuto, A. &
Silvano, S. 1994. A system of monitoring and warning in
ing site occurred during light rainfall on January 14, a complex landslide in northeastern Italy. Landslide News
2006 (Fig. 11), when overall soil moisture was at 8: 12–15.
the highest since the beginning of the 2005–06 rainy Baum, R.L., Godt, J.P., Harp, E.L., McKenna, J.W. &
season. Overall, our monitoring illustrates that soil- McMullen, S.R. 2005. Early warning of landslides for rail
moisture conditions exert a strong control over the traffic between Seattle and Everett, Washington, USA. In
timing of shallow landsliding. The ability to measure O. Hungr, R. Fell, R. Couture & E. Bernhard (eds), Land-
these conditions in near real time makes it possible slide Risk Management, Proc. of the 2005 International
to determine when heavy rainfall is likely to cause Conference on Landslide Risk Management: 731–740.
landslides. New York: A.A. Balkema.
Baum, R.L., Harp, E.L. & Hultman, W.A. 2000. Map
showing recent and historic landslide activity on coastal
bluffs of Puget Sound between Shilshole Bay and Everett,
4 CONCLUSIONS Washington. U.S. Geological Survey Miscellaneous Field
Studies Map MF 2346.
Near-real-time systems for monitoring active land- Berti, M., Genevois, R., LaHusen, R.L., Simoni, A. & Tecca,
slides or landslide-prone hillslopes have advanced P.R. 2000. Debris flow monitoring in the Acquabona
rapidly in recent years and offer many advantages for watershed on the Dolomites (Italian Alps). Physics and
Chemistry of the Earth, Part B: Hydrology, Oceans and
understanding landslide processes. They can provide Atmosphere 26 (9): 707–715.
current information about remote landslide conditions, Chien-Yuan, C., Tien-Chien, C., Fan-Chieh, Y., Wen-Hui, Y.
help ensure high-quality data sets, and capture tran- & Chun-Chieh, T. 2005. Rainfall duration and debris-flow
sient and dynamic processes. Their configurations initiated studies for real-time monitoring. Environmental
vary widely depending on the style of landsliding Geology 47: 715–724.

189
Chleborad, A.F., Baum, R.L. & Godt, J.P. 2006. Rain- Keefer, D.K., Wilson, R.C., Mark, R.K., Brabb, E.E., Brown,
fall thresholds for forecasting landslides in the Seattle, W.M., Ellen, S.D., Harp, E.L., Wieczorek, G. F., Alger,
Washington, Area—exceedance and probability. U.S. C.S. & Zatkin, R.S. 1987. Real-time landslide warning
Geological Survey Open-File Report 2006–1064. during heavy rainfall. Science 238 (4829): 921–925.
Collins, B.D. & Znidarcic, D. 2004. Stability analyses of Kramer, J. & Rutledg, D. 2000. Advances in real-time
rainfall induced landslides. Journal of Geotechnical and GPS deformation monitoring to landslides, volcanoes,
Geoenvironmental Engineering 130 (4): 362–372. and structures. Annual Meeting—Assoc. of Engineering
Dunnicliff, J. 1993. Geotechnical Instrumentation for Mon- Geologists 43 (4): 97.
itoring Field Performance. New York: John Wiley & LaHusen, R.L. 2005. Debris-flow instrumentation. In M.
Sons. Jakob & O. Hungr (eds), Debris-flow Hazards and Related
Ellis, W.L., Kibler, J.D., McKenna, J.W. & Stokes, R.L. 2002. Phenomena: 291–304. Berlin: Praxis, Springer.
Near real-time landslide monitoring using cellular tele- LaHusen, R.L. & Reid, M.E. 2000. A versatile GPS system
phone telemetry—some recent experiences. Geological for monitoring deformation of active landslides and vol-
Society of America Abstracts with Programs 34 (6): 49. canoes. Eos, Transactions of the American Geophysical
Ellis, W.L., Priest, G.R. & Schulz, W.H. 2007. Precipita- Union 81 (48): F320.
tion, pore pressure, and landslide movement—detailed Landslide Technology. 2004. Geotechnical investigation
observations at the Johnson Creek landslide, coastal Ore- Johnson Creek Landslide, Lincoln County, Oregon. Ore-
gon. In V.R. Schaefer, R.L. Schuster & A.K. Turner gon Department of Geology and Mineral Industries
(eds), Conference Presentations: 1st North American Open-File Report O-04-05.
Landslide Conference, AEG Special Publication 23: Mikkelsen, P.E. 1996. Field instrumentation. In A.K. Turner
921–934. Vail, Colorado: Assoc. of Engineering Geol- & R.L. Schuster (eds), Landslides: Investigation and Mit-
ogists. igation, Special Report 247: 278–316. Washington, DC:
Finlay, P.J., Fell, R. & Maguire, P.K. 1997. The relation- National Academy Press.
ship between the probability of landslide occurrence and Mikkelsen, P.E. & Green, G.E. 2003. Piezometers in fully
rainfall. Canadian Geotechnical Journal 34 (6): 811–824. grouted boreholes. In F. Myrvoll (ed.), Field Measure-
Froese, C.R. & Moreno, F. 2007. Turtle Mountain Field Lab- ments in Geomechanics, Proc. of the 6th International
oratory (TMFL): Part 1—overview and activities. In V.R. Symposium, Oslo: 545–554. Lisse: Swets & Zeitlinger.
Schaefer, R.L. Schuster & A.K. Turner (eds), Conference Minard, J.P. 1983. Geologic map of the Edmonds East and
Presentations: 1st North American Landslide Conference, part of the Edmonds West Quadrangles, Washington.
AEG Special Publication 23: 971–980. Vail, Colorado: U.S. Geological Survey Miscellaneous Field Studies Map
Assoc. of Engineering Geologists. MF-1541.
Fukuzono, T. 1990. Recent studies on time prediction of slope NOAA-USGS Debris Flow Task Force. 2005. NOAA-
failure. Landslide News 4: 9–12. USGS Debris-Flow Warning System—Final Report. U.S.
Geolith Consultants. 2000. Executive Summary, Mission Geological Survey Circular 1283.
Peak Landslide, Fremont, California. Ortigao, B. & Justi, M.G. 2004. Rio-Watch: the Rio de
Godt, J.G., Baum, R.L. & Chleborad, A.F. 2006. Rainfall Janeiro landslide alarm system. Geotechnical News 22
characteristics for shallow landsliding in Seattle, Wash- (3): 28–31.
ington, USA. Earth Surface Processes and Landforms 31: Petley, D.N., Bulmer, M.H. & Murphy, W. 2002. Patterns
97–110. of movement in rotational and translational landslides.
Graymer, R.W., Jones, D.L. & Brabb, E.E. 1995. Geologic Geology 30 (8): 719–722.
map of the Hayward fault zone, Contra Costa, Alameda, Petley, D.N., Higuchi, T., Petley, D.J., Bulmer, M.H. & Carey,
and Santa Clara Counties, California: A digital database. J. 2005. Development of progressive landslide failure in
U.S. Geological Survey Open-File Report 95–587. cohesive materials. Geology 33 (3): 201–204.
Hadley, K.C. & LaHusen, R.L. 1995. Technical manual for Pike, R.J. 1997. Index to detailed maps of landslides in the
an experimental acoustic flow monitor. U.S. Geological San Francisco Bay region, California. U.S. Geological
Survey Open-File Report 95–114. Survey Open-File Report 97-745D.
Husaini, O. & Ratnasamy, M. 2001. An early warning system Priest, G.R., Allen, J., Niem, A., Christie, S.R. &
for active landslides. Quarterly Journal of Engineering Dickenson, S.E. 2006. Interim Report: Johnson Creek
Geology and Hydrogeology 34: 299–305. Landslide Project, Lincoln County, Oregon. Oregon
Iverson, R.M. 2000. Landslide triggering by rain infiltration. Department of Geology and Mineral Industries Open-File
Water Resources Res. 207: 59–82. Report O-06-02.
Iverson, R.M., Reid, M.E. & LaHusen, R.G. 1997. Debris- Reid, M.E., Brien, D.L., LaHusen, R.L., Roering, J.J., de
flow mobilization from landslides. Annual Review of la Fuente, J. & Ellen, S.D. 2003. Debris-flow initiation
Earth and Planetary Sciences 25: 85–138. from large slow-moving landslides. In D. Rickenmann &
Jakob, M. & Hungr, O. 2005. Introduction. In M. Jakob C. Chen (eds), Mechanics, Prediction and Assessment:
& O. Hungr (eds), Debris-flow Hazards and Related Proc. of the Third International DFHM Conference,
Phenomena: 1–7. Berlin: Praxis, Springer. Davos, Switzerland: 155–166. Rotterdam: Millpress
Jurasius, M. 2002. Rock slope kinematics of the Mission Science Publishers.
Peak Landslide, Fremont, CA (M.S. thesis), San Jose State Reid, M.E. & Iverson, R.M. 1992. Gravity-driven ground-
University. water flow and slope failure potential, 2, Effects of
Kane, W.F. & Beck, T.J. 1996. Rapid slope monitoring. Civil slope morphology, material properties, and hydraulic
Engineering-ASCE 66 (6): 56–58. heterogeneity. Water Resources Res. 28 (3): 939–950.

190
Reid, M.E. & LaHusen, R.L. 1998. Real-time monitoring of of ground-based SAR interferometry for slope failure
active landsides along Highway 50, El Dorado County. early warning—the Cortenova rockslide (Italy). In K.
California Geology 51 (3): 17–20. Sassa, H. Fukuoka, F. Wang & G. Wang (eds), Land-
Reid, M.E., Nielsen, H.P. & Dreiss, S.J. 1988. Hydro- slides—Risk Analysis and Sustainable Disaster Manage-
logic factors triggering a shallow hillslope failure. Assoc. ment: 337–342. Berlin: Springer.
Engineering Geologists Bulletin 25 (3): 349–361. Terzaghi, K. 1950. Mechanism of landslides. In S. Paige
Saito, M. 1965. Forecasting the time of occurrence of slope (ed.), Application of Geology to Engineering Practice
failure. Proc. of the Sixth ICSMFE, Montreal 2: 537–541. (Berkey Volume): 83–123. New York: Geological Society
Schmidt, J., Turek, G., Clark, M. & Uddstrom, M. 2007. of America.
Real-time forecasting of shallow, rainfall-triggered land- Varnes, D.J. 1983. Time-deformation relations in creep to
slides in New Zealand. Geophysical Research Abstracts failure of earth materials. Proc. of the 7th Southeast Asian
9: 5778. Geotechnical Conference 2: 107–130.
Schulz, W.H. & Ellis, W.L. 2007. Preliminary results of Voight, B. 1989. A relation to describe rate-dependent
subsurface exploration and monitoring at the Johnson material failure. Science 243: 200–203.
Creek landslide, Lincoln County, Oregon. U.S. Geological Wilson, R.C. 2005. The rise and fall of debris-flow warning
Survey Open-File Report 2007-1127. system for the San Francisco Bay region, California. In
Sidle, R.C. & Ochiai, H. 2006. Landslides: Processes, Pre- T. Glade, M. Anderson & M.J. Crozier (eds), Land-
diction, and Land Use. Water Resources Monograph: slide Hazard and Risk: 493–516. West Sussex, United
American Geophysical Union. Kingdom: John Wiley & Sons.
Sidle, R.C. & Swanson, D. A. 1982. Analysis of a small debris Wolle, C.M. & Hachich, W. 1989. Rain-induced landslides
slide in coastal Alaska. Canadian Geotechnical Journal in southeastern Brazil. In Proc. of the 12th Interna-
19: 167–174. tional Conference on Soil Mechanics and Foundation
Tarchi, D., Antonello, G., Giuseppe, C., Farina, P., Fortuny- Engineering: 1639–1642. Rio de Janeiro: A.A. Balkema.
Guasch, J., Guerri, L. & Leva, D. 2005. On the use

191
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

The effects of earthquake on landslides – A case study of Chi-Chi

earthquake, 1999

M.L. Lin, K.L. Wang, & T.C. Kao

Department of Civil Engineering, National Taiwan University, Taiwan, China

ABSTRACT: The Chi-Chi earthquake struck central region of Taiwan on September 21, 1999, with a local
magnitude of 7.3, which induced extensive landslides covering a total area of more than 8000 ha. In this paper,
the effects of the ground motion of the Chi-Chi earthquake on the landslides are examined and discussed. The
relationship between the critical acceleration and threshold displacement to the landslides and ground motion is
analyzed based on the sliding block analysis for evaluation of the seismic stability of slope. Finally, the effects
of the earthquake to triggering of subsequent landslides are discussed.

1 INTRODUCTION was as shown in Figure 2 (Shin, 1999). Noted that

three triggered events with magnitude larger than 6
Taiwan is situated at the juncture zone of the Euro- were also recorded which caused higher seismic inten-
asia Continental Plate and the Philippine Sea Plate sity for some areas far away from the Chelungpu
with a northwest tectonic movement of the Philip- fault. The final ground displacements caused by the
pine Sea Plate. As the consequences of the tectonic thrusting of the fault were surveyed using global posi-
actions, the Central Mountain Range as well as the tioning system (GPS) (Huang et al., 1999), and the
Coastal Mountain Ranges is produced, and the pre- distribution of the displacements was as shown in
vailing geological formations and structural patterns Figure 3. The maximum horizontal displacement was
align approximately parallel to the longitudinal axis about 9 m, and the maximum vertical displacement
of the island. The topography and geological condi- was about 6 m near the north-end of the fault where it
tions are highly related to the tectonic activity, where turned eastward. With such strong ground motion and
the mountain area composes about 77% of the whole large displacements, various types of ground failure
area with highly fractured geological conditions. As occurred causing tremendous amount of damages and
the results of the tectonic activity, Taiwan is among the loss. Among all types of failure, landslides covered
most active seismic districts in the world. The Chi-Chi
earthquake struck central region of Taiwan on Septem-
ber 21, 1999, with a local magnitude of 7.3, had caused
severe ground failures and loss of lives and properties.
The earthquake was triggered by the faulting action
of the Chelungpu fault, with a fault rupture length
of 105 km. The Chelungpu fault is a shallow thrust
east-dipping fault which moved westward. The focal
depth of the earthquake was only 8 km, which meant
a tremendous amount of energy was released near the
ground surface. The maximum peak horizontal ground
acceleration recorded was about 1 g, and the maximum
peak vertical ground acceleration recorded was 0.7 g.
The PGA contour of motion caused by the main shock
is shown in Figure 1. During the Chi-Chi earthquake,
it was found that the ground motion is strong for area
next to the fault, and decreases rapidly with distance
as illustrated in Figure 1. The distribution of the mea- Figure 1. The distribution of horizontal peak ground accel-
sured ground motion along with the seismic intensity eration of the Chi-Chi earthquake.

193
 the most significant extents and caused tremendous
loss and profound effects (NCREE, 2000). The effects
of the earthquake on the landslides and the subsequent
effects of such incident will be discussed in this paper.

2 LANDSLIDES CAUSED BY THE CHI-CHI

EARTHQUAKE

The ground-based field reconnaissance called by the

NCREE for slope failure has reported a total num-
ber of 436 items of slope failure, and the distribution
of slope failure along with the surface rupture of
Chelungpu fault and strong motion stations are as
shown in Figure 4. A follow-up investigation was con-
ducted based on the aero-photo and SPOT satellite
images taken before and after the earthquake. More
than 21,900 items of pixel variation were identified
with a total area of more than 8,600 hectare as shown
in Figure 5. The slope failures distributed from Miao-
Li to Jia-Yi Counties as shown in the figures, and
almost all items located to the right or hanging wall
of the thrust fault, where the mountain terrain located.
Figure 2. The isoseismal map with the three triggered The slope failures in Tai-Chung County, and Nan-
events of the Chi-Chi earthquake (Shin, 1999). Tou County were most widely spread. In Tai-Chung
County, the most severe slope failure is along the
trans-island highway. In Nan-Tou County, a large scale
slope failure of Juo-Juo peaks covered an area as large
as 950 hectare, which was major shallow sliding and
spalling of gravelly material. In Juo-Feng-Err moun-
tain area, a dip slope sliding occurred with an area
of 200 hectare. In Yun-Lin County, a massive dip
slope failure occurred in the Tsao-Ling area which
covered an area of 400 hectare and with an amount
of 120 million cubic meter of deposit material. The
large amount of material deposited in the valley of
Ching-Sui river and formed a dammed-up lake.
Based on the data from the reconnaissance and field
observations, most of the landslides induced by the
earthquake were with small to medium scales, and

Figure 3. The distribution of final ground displacements

caused by Chi-Chi earthquake. The arrows indicate displace-
ment quantity and its direction whereas values appeared in Figure 4. The distribution of the slope failure cases and the
parenthesis. (Huang et al., 1999). strong motion stations (Lin & Tung, 2004).

194
about 70% of the landslides had area smaller than the debris slide and toppling/rock fall would occur on
4000 m2 . The distribution of different types of failure steep slope. The field observation also confirm that
of all items reported is as shown in Figure 6. In Figure the most common type of slides were shallow slides
6 the debris slide is the most encountered failure type on the steep slope occurring near the crest. The weather
which accounts for 63% of the failure, while toppling condition before the Chi-Chi earthquake was fairly dry
and rock fall is the second with 22%. Distribution of without much precipitation as shown in Figure 8. Thus
the slope angle of the failure slopes is as shown in it is unlikely to have significant cases of deep-seated
Figure 7. The landslides with slope angle in the range slides for lacking of ground water pressure effects.
of larger than 45◦ accounts for 90% of all cases, reflect- The horizontal peak ground acceleration (PGA-H)
ing the fact that the slope with high slope angle would and vertical peak ground acceleration (PGA-V) of
have high potential of failure. The results of the slope each slope failure were determined by interpolation
angle compared relatively well with the distribution of the strong motion station records. The results were
of the types of failure of the slope, because typically as shown in Figures 9 and 10, respectively, and it was
found that number of the slope failures increased sig-
nificantly when the vertical component of the peak
ground acceleration reached 200 gal. There were 74%
of the slope failures with the vertical component of
PGA-V larger than 200 gal. As for the horizontal peak
ground acceleration, about more than 81% of slope
failure were with mean PGA-H within the range from
150 to 450 gal. A cross-examination of the distribution
of the rock formations and topography suggested that
the effects of the geological formation and topography
did not appear to be significant. It was suggested that

Figure 5. Distribution of identified landslide caused by Figure 7. The distribution of slope angle of landslides.
Chi-Chi earthquake using SPOT images (Lin et al., 2002).

Figure 8. Rainfall record of the Sun Moon Lake station in

Figure 6. The distribution of failure types of landslides. September, 1999.

195
Figure 9. The distribution of mean horizontal peak ground
acceleration.

Figure 10. The distribution of vertical peak ground accel-

eration. Figure 11. The distribution of selected landslide cases.

the ground motion might be the most important factor

for earthquake induced slope failures (Lin et al. 2000).

3 THE EFFECTS OF GROUND MOTION

ON THE LANDSLIDES

In order to understand the effects of the ground motion

on the landslides, the pixel variation data by COA
(2000) combined with SPOT images and aerial photos
covering more areas were used. The data were care-
fully checked to ensure accuracy and representative of
landslides. Each identified ground surface variation
was screened in accordance to the variations of ele-
vation and slope angle (Lin & Tung, 2004). After the Figure 12. The distribution of horizontal peak ground
screening process, the number of data points in this acceleration versus slope angle.
study is 31702, and the study area with distribution
of landslide events are as shown in Figure 11. The
epicenter of Chi-Chi earthquake is also shown in slope angle, and there appeared to be threshold accel-
Figure 11 with the marked time 01:47:16 (local time, erations for both the vertical and horizontal ground
GMT +8) as the main shock and the other three are motions. Observing Figures 12 and 13, the threshold
triggered events with magnitude larger than 6. vertical peak acceleration is approximately 70 gal for
The distribution of the mean horizontal peak ground slope angle larger than 20◦ , and gradually increases
acceleration and vertical peak ground acceleration of as the slope angle decreases. The threshold horizontal
each event were determined and plotted with respect peak acceleration is approximately 100 gal for slope
to the slope angle in Figures 12 and 13, respectively. angle larger than 20◦ , and also gradually increases as
It was found that the effects of the ground motion the slope angle decrease. In both cases, the thresh-
were significant and relatively independent of the old peak ground accelerations are constants and being

196
Figure 13. The distribution of vertical peak ground accel- Figure 15. The distribution of vertical versus horizontal
eration versus slope angle. peak ground acceleration of the free-field strong motion
stations and landslide events.

800

First triggered
700 earthquake

large vertical motion could have significant effects on

600
the landslides. The main group of events distributed
500 in a bi-linear form with events located within about
PGAV (gal)

North fracture of
Chelun gpu fault

400
10 km to the epicenter and to the surface fracture of
the fault all appeared to have about constant vertical
300
Hypocentral
distance less
and higher horizontal peak ground accelerations with
than 10km
200 events close to the epicenter displaying larger ground
Distance to

100
fault less than
10km
motion. As moving further away from the epicenter
and fault rupture, the vertical and horizontal ground
0
0 100 200 300 400 500 600 700 800 motion of the events decreased more rapidly. Com-
PGAH (gal)
paring the ground motion of landslide events to the
ground motion recorded by the strong motion station
Figure 14. The distribution of horizontal peak ground in Figure 15, the distribution of the main group of
acceleration versus vertical peak ground acceleration. events fell in the similar trend as ground motion
recorded by the strong motion stations. However, it
was found that the landslide events appeared to locate
independent of the slope angle for the range of slope in the ranges with higher ground motion, and typically
angle larger than 20◦ . To examine the effects of the with high vertical acceleration.
two components of ground motion, the vertical peak The attenuation of the peak ground acceleration
acceleration versus horizontal peak acceleration of with respect to the hypocentral distance for vertical
each landslide was plotted in Figure 14. The distri- motion and horizontal motion were plotted in Figures
bution of the ground motions of the landslide cases 16 and 17, respectively. Generally, the attenuation
can be separated into few groups. The group of events of ground motion in both directions of the landslide
approximately parallel and to the left of the main events followed the same trend of attenuation recorded
group was landslides located close to the north end by the free field strong motion stations, and again
of the fault where the maximum ground displacement located in the range with larger ground motion and
occurred. This may be caused by the energy concen- closer to the epicenter. The attenuation curve proposed
tration where the fault rupture turned eastward, which by Wen et al. (2004) was also plotted in Figure 16. The
also induced large vertical and horizontal displace- curve provided satisfactory results but tended to under-
ments. While the second group of events with the estimate the ground motion of the landslide case within
higher vertical peak ground acceleration distributed 10 km of the hypocentral distance. The two small peaks
in the right upper portion of the main group was actu- observed in the figures were the groups affected by
ally affected by the first triggered earthquake. For the the large ground displacement in the north part and the
first triggered earthquake, large vertical and horizontal first triggered event as described previously. Again the
ground accelerations were also recorded. Both groups effects appeared to be more significant on the verti-
appeared to have higher vertical acceleration compar- cal ground motion than the horizontal ground motion.
ing to the main group of events, and it appeared that the Based on the previous discussions, the ground motion

197
 accordingly. Therefore, the determination of the crit-
ical acceleration and threshold displacement is vital
for assessment of slope stability. In order to under-
stand the effects of ground motion with respect to
the critical acceleration and threshold displacement,
cases of sliding failure caused by Chi-Chi earthquake
are analyzed, and the displacements of the slopes
are calculated based on the sliding block method.
The conditions of simple slope and homogeneous soil
properties are assumed and the log-spiral failure sur-
face method developed by Huang & Lin (2003) is
used. The study cases are selected and screened based
on the assumptions and requirements of data for the
analysis.
Figure 16. The distribution of horizontal peak ground It was suggested that the slope angle is the most
acceleration versus hypocentral distance. important factor affecting the critical acceleration of
the slope, and the method was appropriate for slope
with slope angle of 20 to 45 degrees of the range
most sliding occurred, while for slope with angle larger
than 45 degree rock fall and toppling were more likely
to occur (Huang & Lin, 2003). A cross comparison
of the sliding-block method and the numerical mod-
eling method was conducted, and it was suggested
that the displacement calculated in either method
was consistent for landslides with ratio of the criti-
cal acceleration to peak ground acceleration smaller
than 0.5. It implied that the peak ground accelera-
tion was quite large and more than two times of the
critical acceleration when landslides occurred. Analy-
ses were performed on landslide cases satisfying this
requirement, and the resulting toe displacements of
63 such cases were plotted versus the ratios of critical
acceleration to peak ground acceleration in Figure 18
Figure 17. The distribution of vertical peak ground accel- (Lin & Kao, 2005). The calculated toe displacements
eration versus hypocentral distance. appeared to increase with decreasing ratio of the crit-
ical acceleration to the peak ground acceleration and
followed a consistent trend. Noted that in Figure 18
most of the ratios of the critical acceleration to the
appeared to be the most important factor causing land- peak ground acceleration are much smaller than 0.5,
slides during the Chi-Chi earthquake, and the effects and typically the toe displacements thus calculated
of the vertical peak ground acceleration were quite are much larger than 10 centimeters. Such condi-
significant. tions indicated that strong ground motion could induce
large displacements and failure of slopes. However, in
Figure 18 there are a few cases with displacements
smaller than 10 centimeters. For those cases the ratios
4 THE THRESHOLD DISPLACEMENTS of the critical acceleration to the peak ground acceler-
VERSUS CRITICAL ACCELERATION ation are close to 0.5, implying that the ground motion
is not much larger than the critical acceleration. Still
Among the various seismic slope stability analysis for these cases, the calculated toe displacements are
methods, the sliding block method proposed by New- in the range of 5 to 10 cm, which are consistent
mark (1965) was often used for description of seismic with the threshold displacement proposed by Jibson
slope behavior. In Newmark’s method, the displace- (1993), Keefer & Wilson (1989), and Wieczorek et al.
ment of the slope is integrated from the acceleration (1985). Therefore, it is suggested that the peak ground
record when ground acceleration exceeds the criti- acceleration being two times larger than the criti-
cal acceleration determined from stability condition cal acceleration and a threshold displacement of 5 to
of the slope based on material strength and geome- 10 centimeters could be critical for initiation of land-
try of slope, and the stability of the slope is evaluated slide in the Chi-Chi earthquake, and the ratio of the

198
Figure 18. The toe displacement of landslide cases calcu-
lated using sliding block method versus the ratio of critical
acceleration to the peak ground acceleration.

critical acceleration to the peak ground acceleration Figure 19. Landslides and debris flows induced by typhoon
could provide information on the stability of the slope Toraji in 2001 versus landslides induced by the Chi-Chi
earthquake.
subjected to earthquake load.

5 EFFECTS ON SUBSEQUENT LANDSLIDES

Due to the severe slope failures caused by the Chi-Chi

earthquake, the slope material was loosened and with
fissures and cleavages. New landslides, rock falls, and
debris flow may be easily triggered by other earth-
quakes or rainfall. In 2001, typhoon Toraji caused
severe landslide and debris flow hazard, and it was
found that most of the landslides and debris flows
occurred at locations identified with previous land-
slides caused by the Chi-Chi earthquake as shown in
Figure 19. In 2004, typhoon Mindule again caused
severe landslide and debris flow hazard in the Cen-
tral Taiwan area. The reconnaissance (Lin, et al.,
2004) again indicated a close relationship between
the hazard induced by the typhoon Mindule and land- Figure 20. Landslides and debris flows induced by typhoon
slides induced by the Chi-Chi earthquake as shown Mindule in 2004 versus landslides induced by the Chi-Chi
earthquake (Lin, et al. 2004).
in Figure 20. For both events, it was found that the
magnitude and extend of failures of many reactivated
landslides and debris flows increased significantly.
Case histories of 14 debris flow torrents with known together with the reactivated landslide ratio by sub-
significant recurring hazards were documented for sequent typhoon were listed in Table 1. It was found
major typhoon events from 1985 to 2004 as shown that the landslides reactivated by subsequent typhoon
in Figure 21. Observing Figure 21, it was noted that events were significantly. In Ta-Chia river watershed
the number of debris flow occurrence increased signif- the reactivated landslides took up about 30% of land-
icantly after the Chi-Chi earthquake, and the recurring slides induced by both typhoons Toraji and Mindule.
period of severe events decreased. In Chen-You-Lan river watershed the reactivated land-
A close examination of the two most severely slides took up about 55% and 33% of landslides
impacted watersheds- the Ta-Chia river watershed, and induced by typhoons Toraji and Mindule, respectively.
the Chen-You-Lan river watershed was performed. Comparison of the subsequent landslides and land-
The landslide ratios caused by events of the Chi- slides induced by the Chi-Chi earthquake revealed that
Chi earthquake, typhoon Toraji, and typhoon Mindule the landslides caused by the earthquake occurred near

199
 Figure 22. Illustration of reactivation of landslides and trig-
gering of debris flows following the earthquake induced
Figure 21. Number of debris flow occurrences of 14 tor- landslides.
rents during major typhoon events (data from SWCB 2002).

6 CONCLUSIONS
Table 1. The landslide ratio of the Ta-Chia river watershed
and Chen-You-Lan river watershed triggered by three major In 1999 the Chi-Chi earthquake struck central
events. Taiwan and caused extensive landslide hazard. Based
on the analysis of ground motion data of the identi-
Triggering Ta-Chia river Chen-You-Lan river
fied landslide events, it is suggested that the ground
event watershed watershed
motion is the most important factor for causing land-
Chi-Chi EQ 0.48% 2.46% slides, and the vertical peak ground acceleration has
Toraji 1.63% 3.95% a significant effect. The threshold peak ground accel-
Toraji & Chi-Chi 0.48% 2.16% eration observed is approximately 70 gal in vertical
Accumulated direction, and 100 gal in horizontal direction. Further-
rainfall of typhoon more, the stability of slope can be evaluated using the
Toraji, mm 480 634 ratio of the critical acceleration to the peak ground
Mindule 3.19% 4.83% acceleration based on the sliding block method, and
Mindule & Chi-Chi 0.94% 1.57%
Accumulated
the threshold displacement of 5–10 cm. The fol-
rainfall of typhoon low up investigation of the landslides caused by the
Mindule, mm 1658.5 1418 typhoon events suggested that the landslides caused
by the earthquake could be easily reactivated in the
subsequent triggering events. Such effects could be
significant and prolonged, but it would be in a dimin-
ishing rate. However, efforts are still required in order
the crest of the slope due to the characteristic of seis- to reduce such effects of secondary hazard in the near
mic load and ground amplification. Subsequently, the future.
landslide scars and open cracks would easily lead to
landslide reactivation in the lower part of the slope
profile and triggering of debris flow caused by heavy REFERENCES
rainfall of typhoons as illustrated in Figure 22.
Thus, a large magnitude earthquake such as Chi- Huang, C.Y., Lin, C.W., Chen, W.S., Chen, Y.G., Yu, S.
Chi earthquake could lead to extensive ground failures B., Chia, I.P., Lu, M.D., Hou, C.S. & Wang, Y.S. 1999.
with landslides, cracks and fissures, which could Seismic Geology of the Chi-Chi Earthquake, Proceed-
lead to activation of the subsequent landslides and ings, International Workshop on the September 21, 1999
triggering of debris flows. It is expected that the Chi-Chi Earthquake, 25–42.
effects would be prolonged but in a diminishing rate, Huang, G.J. & Lin, M.L. 2003. Dynamic Slope Analysis
as suggested in the Table 1 that the accumulated Using Sliding Block Method, Journal of Chinese Institute
rainfall of typhoon Mindule was much higher than of Civil and Hydraulic Engineering, 15(4), 655–665.
Jibson, R.W., 1993. Predicting Earthquake-Induced Land-
that of typhoon Toraji for Chen-You-Lan watershed slide Displacements Using Newmark’s Sliding Block
without causing significant increase of reactivated Analysis, Transportation Research Record, 1411, 9–17.
landslides. However, efforts are still required in order Keefer, K.K. & Wilson, R.C. 1989. Predicting earthquake-
to reduce such effects of secondary hazard in the near induced landslides with emphasis on arid and semi-arid
future. environments, in Sadler, P. M., and Morton, DD. M., eds.,

200
 Landslides in a semi-arid environment. California, Iland NCREE, NAPHM, and Taiwan Geotechnical Society, 1999.
Geologicial Society, 2, 118–149. Reconnaissance report of the geotechnical hazard caused
Lin, M.L., Chen, T.C. & Chen, L.C., 2002. The Geotechnical by Chi-Chi earthquake, National Research Center on
Hazards and Related Mitigation after Chi-Chi Earthquake, Earthquake Engineering, Taiwan, 111p.
Proceedings, Disaster Resistant California Conference, Newmark, N.M. 1965. Effects of Earthquake on Dams and
City of Industry, California, USA. Embankments, Géotechnique, 15(2), 139–160.
Lin, M.L., Chen, C.H., Lin, M.L., Chen, H.Y., Lin, C.C. & Shin, T.C., 1999. Chi-Chi Earthquake-Seismology, Proceed-
Hsue, M.H. 2004. Reconnaissance of the Hazard Induced ings, International Workshop on the September 21, 1999
by Typhoon Mindule in Ta-Chia River Watershed, the Chi-Chi Earthquake, 1–14.
magazine of the Chinese Institute of Civil and Hydraulic Wen, K.L., Jan, W.Y., Chang, Y.W., Chen, K.T & Jian,
Engineering, 31(4), 19–25. J.S. 2004. Site effects on strong motion of Taiwan area,
Lin, M.L. & Kao, J.J. 2005. The Threshold Displacement Report, Central Weather Bureau, Report No. MOTC-
of Landslides Caused by Chi-Chi Earthquake, Interna- CWB-93-E-09.
tional Symposium on the Potential, Risk, and Prediction Wieczorek, G.F., Wilson, R.C. & Harp, E.L. 1985. Map
of Earthquake-induced Landslides, Taipai, Taiwan. showing slope stability during earthquakes of San Mateo
Lin, M.L. & Tung, C.C. 2004. A GIS-Based Potential Analy- County, California. U.S. Geological Survey Miscel-
sis of the Landslides Induced by the Chi_Chi Earthquake, laneous Geologic Investigations Map 1 1257E, scale
Engineering Geology, 71, 63–77. 1:62500.
Lin, M.L., Wang, K.L., & Chen, T.C., 2000. Characteris-
tics of the Slope Failure Caused by Chi-Chi Earthquake,
Proceedings of International Workshop on Annual Com-
memoration of Chi-Chi Earthquake, III-Geotechnical
Aspect, 199–209.

201
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

The role of suction and its changes on stability of steep slopes in unsaturated
granular soils

L. Olivares
Dept. of Civil Engineering, Second Univ. of Naples, Italy

P. Tommasi
Institute for Environmental Geology and Geo-Engineering, National Research Council, Rome, Italy

ABSTRACT: A significant part of Italian mountain areas are covered by pyroclastic deposits resting at slope
angles higher than 40–50◦ . The stability of these steep slopes in loose or poorly cemented pyroclastic materials
is essentially guaranteed by the positive effects of matrix suction on shear strength until an increase in saturation
(and hence a decrease in suction) is induced by seepage initiated by different processes. The Cervinara flowslide
(Campania, Italy) is a typical case where rainfall infiltration increased saturation and hence led to failure of
shallow pyroclastic deposits. At La Fossa Crater (Vulcano Island, north of Sicily, Italy) seepage induced by
condensation of vapor produced by the degassing activity of the active volcano seems instead to be correlated
to slope movements. In the paper the two case studies are examined by means of monitoring data, laboratory
investigations and numerical analyses.

1 INTRODUCTION In most rainfall-induced flowslides it is possible to

recognize some recursively similar elements:
Unsaturated cohesionless or slightly-bonded pyroclas-
tic materials can form relatively steep slopes whose – flowslides take place essentially in shallow cohe-
stability suffers from wetting, which reduces suction sionless air-fall deposits in primary deposition lying
and hence shear strength. on fractured limestone;
In this paper two cases of slope instability initi-
ated by wetting occurring through different seepage
processes are described and discussed.
The first is the Cervinara slide, which is represen-
tative of flows triggered by rainfall infiltrating in the
shallow pyroclastic deposits that overlie the limestone
massifs of the Campanian Apennines (southern Italy).
The second case is a landslide on the flank of the
pyroclastic cone of La Fossa volcanic edifice in Vul-
cano Island (North of Sicily). Its initiation is believed
to have been influenced by a sharp increase in con-
densed vapour produced by the degassing of the active
volcano.

1.1 Landslides induced by rainwater infiltration

In recent decades, a large number of landslides of flow
type (flowslides) have been triggered in pyroclastic
deposits in the southern Italian region of Campania by
rainwater infiltration, causing loss of life and prop-
erty (Olivares & Picarelli, 2003; Calcaterra & Santo,
2004). In Figure 1 a typical example of a landslide is Figure 1. Cervinara flowslide: plan-view and schematic
shown as it evolves into a flowslide. cross-section.

203
 Flowslidesin shallow
pyroclasticcovers
characteristics of each layer of the deposit and, above
layersparalleltobed-rock.
InitialconditionSr<1
all, by the state variables at the beginning of rain.
To remove this uncertainty at the Geotechnical Lab-
bed-rock: fracturedlimestone
oratory of the Second University of Naples a complex
experimental program was conducted. It included:
Frompre-failuretofailure Trigger:
Rain-water infiltration

Yes Slopeangle No – mechanical and hydraulic characterization of the

Gentleslope:
’ ashy soils in both saturated and unsaturated con-
Steepslope:
Failure⇒(Sr=1)
u>0 Failure⇒(Sr≤1) ditions (see section 2.1);
’ ’ – infiltration tests performed on both homogeneous
Failure⇒(Sr≈1) Failure⇒(Sr«1) and layered small-scale slopes in a well instru-
mented flume (FL tests) to investigate the mechan-
Post-failure
Drained
Soilresponse
Undrainedor partially
Soilresponse
Drained ics of rainwater infiltration leading to slope failure
Undrained
Liquefable ? (see section 2.2);
deposit
– calibration of a numerical model through back-
Yes No analysis of infiltration tests (see section 2.3.1);
Evolution: slide Evolution:flowslide Evolution: slide Evolution:slideor
– in situ monitoring of the slope involved in the 1999
(liquefactionor fluidization) debrisavalanches
flowslide, including suction measurements at dif-
ferent depths in the ashy layers and pluviometric
Figure 2. Conditions for rainfall-induced flowslide in
measurements (see section 2.3.2);
unsaturated soils (infinite slope) (Olivares and Damiano, – back-analysis of in situ suction measurements, by
2007). means of numerical simulations carried out with
boundary and initial conditions derived from in situ
monitoring (see section 2.3.2).

– deposits are constituted by alternating ashy and

pumiceous layers with an average thickness in the 1.2 Landslides induced by volcanic activity
order of a few decimeters; In active volcanoes hydraulic conditions are affected
– the slope angle is almost constant and high (35–45◦ ) not only by infiltrating rainwater but also by vol-
and the shallow failure surface is parallel to the canic activity, which produces complex changes in the
limestone formation; state variables of pore fluids (i.e. pore fluid pressure).
– the water table is absent and the deposits are initially In particular, volcanic activity can modify pore fluid
unsaturated; pressure as far as to induce slope instability.
– the soil is susceptible to liquefaction. In the geotechnical and geological literature this
effect is well documented only for lava domes with
Starting from these similarities, Olivares & Damiano reference to gas overpressures induced by rainfall
(2007) proposed a framework (Fig. 2) to predict the infiltration (see e.g. Elsworth et al. 2007).
development of landslides induced by rainfall infiltra- In relatively steep pyroclastic cones, slope instabil-
tion in homogeneous deposits under the assumption of ity may also be induced by the increase in water content
an infinite slope. produced by vapour condensation, whose main effect
On steep slopes (i.e. slope dip α greater than is the decrease in suction up to possible development
the saturated friction angle φ  of the soil) failure of positive pore pressures.
can occur in almost complete saturation or in states This mechanism seems to apply to slope stability
still far from saturation, conditioning the failure conditions of the cone of La Fossa at Vulcano Island
mechanisms (drained or undrained) and post-failure where, during the most intense well documented vol-
evolution (flowslide, slide or debris avalanche). canic unrest, large slope movements developed on the
Hence, in this schematization the elements control- NE flank of the cone.
ling different failure and post-failure mechanisms are In order to verify the validity of the proposed
represented by the state properties and in particular by mechanism, geotechnical investigations including lab-
the degree of saturation Sr (or water content w) at the oratory testing and in situ suction measurements were
onset of instability. conducted and used to set up simple seepage models
The weakness of this framework is its ability to and run stability analyses (see section 3). Continu-
predict the hydrologic slope response (i.e. of link- ous monitoring of suction and soil/air temperature at
ing rainfall to the space-time distribution of Sr or w two selected sites was recently started with the aim
and that of the state of stress within the soil deposits). of ascertaining whether suction changes can be corre-
In this complex linkage a fundamental role is played lated to temperature variations induced by an evolution
by the boundary conditions, by the intrinsic hydraulic of the volcanic activity.

204
 The research currently involves the following insti- 100
Ash (B) Upper pumices (A)
90

Percent by weight
tute: CNR-IGAG, CIRIAM and INGV-Palermo. 80 Altered ash (D) Lower pumices (C)
70
60
50
40
30 a) b)
2 CERVINARA FLOWSLIDE 20
10
0
The Cervinara flowslide (3a in Fig. 1) was a typi- .0001.001 .01 .1 1 10 .0001 .001 .01 .1 1 10
d [mm] d [mm]
cal rainwater-induced flowslide that occurred along
a fairly regular slope formed by a primary deposit
Figure 3. Grain size distribution of the tested material.
of unsaturated layered air-fall pyroclastites overlying
fractured limestone. Figure 1 presents a plan-view and
a schematic cross-section obtained from boreholes and
pits carried out along the landslide. At the main scarp, 200μm
the slope is around 40◦ and the average thickness of the 1mm

pyroclastic cover is about 2–2.5 m. From top to bot-

tom under a top soil formed by remoulded volcanic
Intragranular pores
ashes (60 cm thick) the following layers were identi- bridge
fied: A) an upper layer (20 cm) of coarse pumices;
B) a layer (75 cm) of volcanic ashes; C) a horizon
of finer pumices mixed with ash (20 cm); D) a bot- a) b)
tom layer (40 cm) of weathered ash in contact with the
Intergranular pores
fractured limestone bedrock. The total thickness of the
cover increases at the slope toe due to the presence of
Figure 4. SEM photographs: a) pumice; b) fabric of layer B
colluvium. showing large (from Picarelli et al, 2006) intergranular pores,
Table I reports the mean values of physical prop- often filled with silt.
erties of the different materials (Olivares 2001;
Olivares & Picarelli 2003; Picarelli et al. 2006). Poros-
ity is very high and the soils are unsaturated with Sr 2.1 Mechanical and hydraulic characterization
depending on environmental conditions. Within each
layer the grain size distribution is quite uniform (Fig. 3) Mechanical and hydraulic characterization of soils was
indicating a highly selective deposition mechanism carried out on undisturbed and remoulded specimens
(air-fall) (Picarelli et al., 2006). In particular, volcanic recovered from pits. As the layers are cohesionless,
ashes (B) display a high sandy component but also a undisturbed sampling was successfully carried out
significant amount of non-plastic silt; the altered ash only in the ash (layer B), and altered ash (layer D).
in layer D is finer than the ash in layer B and pumices Data on shear strength of pumices are not available.
fall in the domain of sandy gravel (A) and gravelly sand The shear strength of saturated volcanic ashes was
(C). SEM observations show that pumices are vesic- measured through CID and CIU triaxial tests on sat-
ular and present small pores (intragranular) separated urated natural samples (Olivares and Picarelli, 2001).
by glassy diaphragms caused by quick quenching in The ash (B) is characterized by a friction angle φ  of
air after explosion (Fig. 4a). Soil B is characterized by 38◦ and a nil value of cohesion c , while weathered
intergranular and intragranular pores (Fig. 4b, c). The ash has a peak and critical state-strength characterized
intergranular pores and have a size comparable to that by φ  and c equal to 31◦ and 11 kPa and 35◦ and 0
of the coarsest particles. respectively.

Table 1. Average values of properties of pyroclastic soils.

dmax γ w n Sr
Layer Material mm U Gs kN/m3 [%] [%] [%]

A Coarse pumices 20 5 2.5 13 25 52 36

B Volcanic ashes 10 5 2.6 14 67 67 84
C Pumices 20 42 2.6 14 40 50 40
D Weathered ashes 10 25 2.6 16 54 54 75

* w and Sr were obtained from samples taken at the end of wet season.

205
 Since suction largely controls strength and The unsaturated coefficient of permeability of ashes
hydraulic properties of these materials, a significant B obtained from the interpretation of the transient
part of the testing program was carried out using appa- phase of suction equalization (Kunze et al., 1965)
ratus especially suited to unsaturated soils. The shear in SCTX tests is reported in figure 6a. In the same
strength in unsaturated conditions was measured only figure the permeability values versus suction esti-
for soil B by means of suction-controlled triaxial tests mated using the expression proposed by Brooks and
(SCTX) using a mean net stress (p − ua ) between Corey (1964) and the permeability function obtained
20 and 200 kPa and a suction (ua − uw ) between 10 by Gardner’s equation (1958) are plotted. In both cases
and 80 kPa (Olivares and Picarelli 2003). The shear in the range of suction between 0 and 80 kPa, perme-
strength envelope of the saturated material (c = 0; ability decreases by about two orders of magnitude as
φ  = 38◦ ) and strength values of the unsaturated mate- suction increases. Figure 6b reports the correspond-
rial measured at different values of suction are reported ing data in terms of volumetric water content θw versus
in Figure 5b. All data show the significant role of suc- (ua -uw ) from SCTX tests at the end of suction equaliza-
tion, even at small values: in particular, for a suction tion together with the corresponding retention curve
of 4–8 kPa cohesion is 2–6 kPa (Fig. 5a). These val- obtained by means of conventional long-term evapora-
ues are sufficient to justify the stability of steep slope tion and infiltration tests on natural and reconstituted
during dry period. samples. These results are in good agreement with
The coefficient of permeability at saturation mea- the expression proposed by van Genuchten (1980)
sured on natural samples taken from layers B and (θr = 0.3; θs = 0.7; m = 0.2; n = 7; α = 1.7).
D through constant head tests for different effec-
tive stresses (40–70 kPa) ranges between 1.04 × 10−7
and 5.46 × 10−7 m/s for layer B and between
2.2 Results of infiltration tests and comparison
8.5 × 10−8 and 6.1 × 10−7 m/s for layer D. The sat-
with numerical analyses
urated permeability of reconstituted pumices by con-
stant head permeability tests assume the values of Two types of infiltration tests performed on a small
1.1 × 10−5 m/s (layer A) and 5.1 × 10−6 m/s (layer C). scale of slope on pyroclastic soils taken from the
Cervinara site are presented (Tables 2 and 3).
The tests were performed on both homogeneous and
layered slopes: in the first case the infiltration process
in a deposit with a slope of 40◦ lying on an impervious
base is presented; by contrast, in the second case, the
infiltration process on layered deposit lying on a pervi-
ous base is analyzed. In all tests, to simulate a hydraulic
boundary condition of free flow at the toe of the slope
a geotextile drain is positioned (side 3 in figure 7),
with conductivity more than one order of magnitude
higher than that of pumices. In figure 7 a sketch of the
model slope and monitoring system is given. Dimen-
sions of a small scale slope are selected as regards the
thickness/length ratio to consider the assumption of an
infinite slope valid. During the infiltration (constant
rainfall with intensities varying from 18 to 105 mm/h)
Figure 5. Unsaturated shear strength of Cervinara ashes and evaporation stages, the following were monitored:
(Olivares, 2001). water content (by the TDR technique), displacements
(by laser transducers and the PIV technique), suction
(by mini-tensiometers) and positive pore pressures (by
1.E-05
Constant head tests Brooks & Corey 1964
0.8
SCTX tests
miniaturized pore pressure transducers). More details
1.E-06
Kunze et al. 1965 Gardner 1958
infiltration test
evaporation test about tests results, the characteristics of the apparatus
0.6 initial conditions in flume tests
van Genuchten and the monitoring system are reported in Olivares &
Damiano (2007) and Olivares et al. (2008).
w

1.E-07 0.4

1.E-08 0.2

ua-uw [kPa] ua-uw [kPa]

2.2.1 Results of infiltration tests on homogeneous
1.E-09 0.0
artificial deposit
In figure 8 (a and b) results of FL16 and FL10 tests
Figure 6. Unsaturated permeability and retention curve on are select which show, for state of stresses corre-
Cervinara ashes. sponding to 10 cm of thickness, the typical response

206
Table 2. Tests on homogeneous slopes in volcanic ashes
(B): initial and boundary conditions.

initial rainfall boundary

conditions characteristics conditions*
slope

side 1

side 2

side 3
angle
test ni wi [mm/h] t

80 27’
FL7 40 0.71 0.31 0 4’ b b a
80 19’

FL8 40 0.70 0.30 60 31’ b b a

FL9 40 0.73 0.37 60 30’ b b a

FL10 40 0.65 0.43 60 37’ b b a

FL15 40 0.75 0.50 40 36’ b b a

FL16 40 0.75 0.40 80 30’ b b a

* a = pervious; b = impervious.

Table 3. Tests on layered slopes: geometry, initial and

boundary conditions. Figure 7. Sketch of the instrumented flume: a) test on
homogeneous slopes; b) test on layered slopes; c) plan-view.
L hi ni wi
slope layer soil [m] [m] [%] [%]

upper ashes (B) 1.7 0.10 76.0 35.0 respectively of a loose and a dense soil. The geom-
etry, the initial and boundary conditions of the slope
layered

medium pumices (A) 1.7 0.04 72.0 12.6 and the intensity and duration of artificial rainfall are
summarised in Table 2. In both the tests during infil-
lower ashes (B) 1.7 0.06 75.0 35.0
tration a marked suction decrease occurs starting from
the ground surface towards the base of the model as
rainfall characteristics boundary
confirmed by the delay (about 6–8 minutes) of the
test slope
i t
deep tensiometers installed at the bottom of the model
side 1

side 2

side 3

FL20 angle
slope (figure 8a, b). As revealed by settlement mea-
Stage [mm/h]
surements recorded by laser sensors, the volumetric
I_a 0 50 61’ a b a collapse is significant (about 8%) in the case of loose
I_b 0 0 6days 20h 10’ a b a soil (Figure 8a; FL16 test) and negligible in the case
of dense soil (Figure 8b; FL10 test). In both cases,
II_a 0 27 2h 32’ a b a about 10 minutes before instability, the deepest ten-
II_b 0 0 2days a b a siometers record a practically nil value of suction and
the pore pressure transducers record a positive value
III_a 20 27 21’ a b a
III_b 20 0 12days 21h a b a
at the base, while those superficial maintain constant
values of about 2–3 kPa.
IV_a 40 18 4h 50’ a b a Soil water content was measured by an innovative
IV_b 40 0 18days 17h 58’ a b a TDR technique (inverse profiling methods; Greco,
2006) to extract a non-homogeneous moisture pro-
V_a 40 85 1h 45’ a b a file along the probe axis. Measurements are based on
V_b 40 85 5h 7’ a b b the correlation between the bulk dielectric permittiv-
V_c 40 0 3days 17h 24’ a b b
ity of wet soil (εr ) and volumetric water content θw
VI_a 40 105 7h 25’ a b b appropriately determined on Cervinara volcanic ashes
(Damiano et al., 2008) since it differs significantly
* a = pervious; b = impervious. from the relationship proposed by Topp et al. (1980).

207
 0 0 Coupling the measurements of the suction with the
a) laser sensors θw at the same depth, it is possible to extract the

settlements [mm]
-10 -5
water retention curves experienced by the soil during
u a - u w [kPa]

superficial
-20 tensiometers -10 the infiltration process. In Figure 9b these curves are
-30
deep tensiometer
-15 compared with those obtained through conventional
long-term laboratory techniques (see section 2.1). The
-40 failure -20
two experimental results were fitted through the van
-50 -25 Genuchten (1980) expression: the characteristic curve
0 5 10 15 20 25
t [min]
30 extrapolated by flume tests (curve 2 in fig. 9b) clearly
0 -1 shows that for θw > 0.4, suction is lower than the val-
b) laser sensors
-10 0 ues obtained by conventional laboratory tests (curve
1 in fig. 9b). This effect is probably related to the

settlem ents [m m ]
-20 1
u a - u w [kPa]

superficial failure different rate of suction decrease imposed in the two

-30 tensiometers 2
procedures accounting for the pyroclastic nature of
deep tensiometers
-40 3 the particles. The size of the pores between the par-
-50 4 ticles (intergranular pores) is significantly higher than
-60 5 that of internal pores (intragranular pores) (see fig. 4).
0 5 10 15 20 25 30 35 Consequently the amount of retained water required
t [min]
to cause a given suction decrease is a function of the
exposure time. Indeed only long exposure times (as in
Figure 8. Suction and settlement measurements during:
the conventional tests) enable the intragranular pores
a) FL16 test; b) FL10 test.
to absorb water.

2.2.2 Results of infiltration tests on layered

t = 00'00 t = 05'15 t = 09'28 0.8
artificial deposits
SCTX tests
t = 11'40 t = 13'59 t = 18'20
infiltration test In this section we discuss the results of phase IVa of
t = 20'49 t = 23'37 evaporation test
10
0.6 initial conditions in flume tests the layered FL20 test with a pumiceous layer of soil
z [cm]

flume infiltration test

8 van Genuchten (1) A interbedded between two ashy layers of soil B. In
this test the slope angle was of 40◦ and a pervious
van Genuchten (2)
0.4
6
θw

4
boundary was reproduced at the base (side 1); rainfall
2
0.2
van Genuchten (1): m=0.2 n=7 α=1.7 θr=0.3 θs=0.7
intensity was kept constant at 18 mm/h for 4 hours and
van Genuchten (2): m=0.08 n=20 α=5.0 θr=0.3 θs=0.7 50 minutes.
0 0.0
0.20 0.30 0.40 0.50 0.60 0.70 0 20 40 60 80 The results in terms of suction against time are
θw ua-uw [kPa]
shown in figure 10. As in the case of homogeneous
Figure 9. Volumetric water content profiles retrieved by
deposits a marked suction decrease appears throughout
TDR during experiment FL16 (a) and Water Retention Curve the soil mass, but in this case an instability condition
derived from TDR compared with WRC obtained from was not reached. The comparison between the mea-
laboratory tests. surements in the two ash layers across contact with

0
As an example of this application, in Figure 9 are
plotted the results of the FL16 test are plotted in terms 5
of θw profiles at different times extracted along a probe
installed normally to the ground surface. At 11 40 10
from the beginning of test, the θw profile shows a
strong curvature due to the wetting of the uppermost 15
soil layer, in line with a decrease in suction recorded
by superficial tensiometers (fig. 8a). Subsequently, 20
section A-A'
as the deep tensiometers start to record significant
25 section B-B'
decreases in suction, the form of the profile changes
(t = 18 20 min) due to the progressive wetting of 30
section C-C'

the lower part of the slope. As the deep tensiometers 0 40 80 120 160 200 240 280
record a suction of 1 kPa the θw profiles (23 37 min)
show that an almost uniform moisture distribution is
reached, highlighted by a value of θw approaching the Figure 10. Suction measurements during FL20 at three
porosity of soil. different depths.

208
 t = 00' 00 t = 44' 03
0.8
flume infiltration test
section 2.1. The unsaturated permeability functions
t = 92' 44 t = 175' 33
0.2
t = 293' 38 van Genuchten, 1980 are described by the Brooks and Corey (1964) expres-
0.6
ashes vG: m=0.3 n=7 α=3.3 sion. The adopted retention curves are described by

θw
0.15 θr=0.25 θs=0.55
the van Genuchten (1980) expressions: for both top-
pumices 0.4
0.1 soil and ash (B) the retention curves of figure 9b were
ashes
0.2
adopted; for pumices (A and C) the retention curve of
0.05
figure 11b; for altered ash (D) data from the literature
z [cm]

θw ua-uw [kPa]
0 0.0 (Fredlund and Rahardjo, 1993).
0.0 0.2 0.4 0.6 0.8 0 20 40 60

2.3.1 Calibration of numerical model trough

Figure 11. Volumetric water content profiles retrieved by a back-analysis of infiltration tests
TDR during experiment FL20 (a) and Water Retention Curve Calibration of numerical model was carried out on the
of pumices derived from TDR. basis of results of infiltration tests on homogeneous
(FL10) and layered slopes (FL20). A 3D analysis was
the pumiceous layer shows an immediate response performed schematizing the slope model with a mesh
in the top layer with a different trend and a marked characterized by dx = dy = dz = 1 cm, consid-
delay in the bottom layer. Furthermore, the compari- ering the boundary conditions indicated in tables 2
son between the two suction measurement performed and 3 and assuming for the initial condition a constant
in the same bottom ash layer highlights a negligible value of suction equal to the mean value recorded at
difference despite the different depth. the beginning of tests.
Figure 11a gives the θw profiles at different times As an example of this calibration stage, in figure 12
obtained with the TDR inverse profiling methods the experimental data of FL10 test are compared with
using a single probe installed through the three layers. the results of two numerical simulations, considering
Since the suction assumes the same value at the ash- for the ashes B the van Genuchten parameters obtained
pumice interface, coupling the measurements of the either in conventional laboratory tests (simulation 1
suction in the ash with the θw evaluated in the pumices, with curve 1 in Fig. 9b) or in the flume infiltra-
it is possible to extract the water retention curves of tion tests (simulation 2 with curve 2 in Fig. 9b). As
pumices experienced by the soil during the infiltration expected, the best fitting was obtained by using reten-
process (Figure 11b). The experimental points were tion curve 2 that allows both the trend and the final
fitted by the expression proposed by van Genuchten values of suction at different depths to be captured.
(1980) (θr = 0.25; θs = 0.55; m = 0.3; n = 7; Conversely, simulation 1 cannot reproduce the trend
α = 3.3). of the deepest tensiometers and the value of suction at
the end of the test. Retention curve 2 probably leads
to higher suction than measured since it was obtained
2.3 Analysis of infiltration
for equalization times longer than the test duration,
The complexity of infiltration process on unsaturated which was insufficient to involve the water inside the
soils requires the use of a numerical model. The intragranular pores.
reliability of the model can be improved through a
preliminary calibration (given the variability in soil
parameters) and a validation through a back-analysis
0
of in situ monitoring. In this paper the former one
was obtained using data from laboratory tests on nat-
ural samples and simulating infiltration tests on the
slope model; the latter through back-analysis of in 2
20
situ suction measurements using initial and boundary
conditions derived from monitoring.
Numerical analysis was carried out using an
I-MOD3D (volume finite method) program developed 40
4
in VBA application for ARCOBJECTTM /ARCGIS measured data
9.2TM to automate the mesh-generation starting from simulation 1
a Digital Terrain Model. simulation 2

The analysis was performed under isothermal con- 60

6
ditions for a undeformable unsaturated porous medium 0 10 20 30
neglecting the flux of the gas phase. The saturated per-
meability of volcanic ashes (B and D) and pumices Figure 12. Test FL10: comparison between suction mea-
(A and D) was assumed equal to the maximum surements and numerical simulation at section A-A in
values obtained in laboratory tests and reported in fig. 7.

209
 80 0

70 40

60 80

50 superficial devices nest 1 120

depth 0.60m nest 2

h [mm]
40 nest 3
160

30 nest 4 200
nest 5
20 240

10 280

0 320

May-02

Mar-03
Jul-02
Apr-02

Oct-02

Feb-03
Jun-02

Aug-02

Sep-02

Dec-02

Jan-03
Nov-02
70 0

60 40

50 80

intermediate devices

h [mm]
40 120
depth 0.90-1.00m nest 3
nest 3
30 160
nest 4
Figure 13. Test FL20: comparison between suction mea- nest 4
20 200
surements and numerical simulation at three depths at section
A-A in fig. 7. 10 240

0 280

Jun-02

Jan-03
Aug-02
Jul-02

Feb-03

Mar-03
May-02
Apr-02

Sep-02

Oct-02

Nov-02

Dec-02
In the figure 13 the results of the layered FL20
test are compared with the numerical simulation con-
sidering, for both the soils, pumices and ashes, the Figure 14. Suction measured at the five tensiometer stations
as a function of daily rainfall.
retention curves obtained in flume tests. Numeri-
cal analysis correctly simulated the response of both
points located across the pumice layer and points the previous 30 years maximum (180 mm in Decem-
located within the ash layer below the pumices, repro- ber 19th 1968). Furthermore the total rainfall during
ducing the different trend recorded above and below the summer (550 mm) was almost half the annual pre-
the pumice. The results show the importance of model cipitation. Due to the summer rains, suction fell from
parameter calibration based on flume infiltration test the end of July with a faster decrease at the end of
results, since infiltration rates and boundary condi- September. In the months of October and November,
tions applied to the model slope resemble real slope it fluctuated around values of a few kPa.
conditions. Monitoring also shows that during summer a signif-
icant reduction in suction occurred as a consequence
2.3.2 Validation of model: comparison between of prolonged rainfall, as after July, 19th and not dur-
field data and numerical analysis ing the most intense and short events, as on August
At 2002 a complex monitoring system was installed 7th. This is a consequence of the low hydraulic
on the natural grassed area surrounding the catas- conductivity of the shallow unsaturated soils, which
trophic flowslide of Cervinara (on the right side of prevents water infiltration during short precipitations.
the landslide 3a in fig. 1). The equipment con- In contrast, infiltration is facilitated by the increase
sists of a pluviometer and five tensiometer stations. of degree of saturation occurring for long lasting
The tensiometers (jet-fill type manufactured by Soil- events.
Moisture) were installed at different dephts within the The period between 13/7/2002 to 8/8/2002, char-
top soil and the volcanic ashes (B and D). Figure 14 acterized by abnormal rainfall after a prolonged dry
shows the suction measured at different depth from period, was selected to validate the numerical model.
April 2002 to April 2003. During the winter sea- The slope has been schematized with a 3D mesh
son, suction reached very low values, ranging between derived by DEM (with dx=dy=0.5 m) and by stratig-
2 kPa (in the top soil) and 15 kPa (in layer B). The dry raphy of fig. 1 using a dz=0.12 m. At the ground
season started in June as indicated by the lack of rain- surface two conditions have been used: average daily
fall during the period June-July. In this dry period the rainfall intensity from monitoring or evaporation flux
suction increased up to values of some tens of kPa with during dry period considering the minimum value sug-
a peak in middle July, when the shallow device of nest gested by Wilson (1990) for unsaturated soil (ranging
1, installed at depth of 60 cm in the evapotranspiration between 0.33 mm/h (first day) and 0.012 mm/h (after
zone, indicated a maximum value near to 50 kPa. three days)). For the lateral and base surfaces have
Subsequently the summer had an abnormal rainfall been considered a condition of free flow. The initial
regime for the Mediterranean area. On August 7th the condition in terms of suction has been established from
daily rainfall was of 165 mm only slightly lower than field measurements.

210
 80 0

70 40 1988 landslide Forgia Vecchia

60 nest 1 0.6m 80
nest 2 0.6m
ua - uw [kPa]

50 120

h [mm]
simul. 0.24m
40 160
simul. 0.36m
30 200
simul. 0.6m
20 240

10 280

0 320

Aug-02

Aug-02

Aug-02
Jul-02

Jul-02

Jul-02

Jul-02

80 0

70 nest 3 0.9m 40
60 nest 3 1m
80
nest 3 1.9m
ua - uw [kPa]

50
m
500

h [mm]
120
simul 1.08m
40
simul 2.28m 160
30

20
200 Figure 16. Shadow relief of La Fossa edifice.
10 240

0 280
10-Aug-02

20-Aug-02

30-Aug-02
01-Jul-02

11-Jul-02

21-Jul-02

31-Jul-02

400 SE wet surge deposit (C) NW

Punte Nere fall products with
fumarole dry surge intercalations (B)
300
P.te Nere basal
(C) dry surge (A)
Figure 15. Comparison between in situ measurements and 200
numerical analysis. (B)

100 Punte Nere

(A) lava flow
A comparison between the monitoring results and
0
numeric simulations at three depths is shown in Punta Roia
lavas
figure 15. During this wet period a adequate agree- -100
ment has been obtained in both top soil and volcanic
ash layer B, although the measured values are discon-
tinuos and a higher decrease of suction is predicted Figure 17. Schematic SW–NE profile of La Fossa cone
during rainfall. As observed by field measurements based on data from Dellino et al. (1990). The dashed line
and confirmed by numerical analysis prolonged rain- indicates the approximated location in the sequence of the
fall (13/07–19/07) have an effect on suction higher 1988 slide.
than a shorter and heavier rainstorm (7/08) in the shal-
lowest layers. In layer D the suction variation does not
seem to be strictly influenced by rainfall events and the – layers of coarse grained (up to 200 mm in diameter)
inadequate prediction of numerical analysis is proba- fall products with thin intercalations of sands from
bly related to the assumption on the initial condition. dry surge deposition (unit B);
During dry period the predict trend is qualitatively in – a lava flow closing the eruptive cycle.
agreement with measurements even if characterized
This sequence is overlaid by a wet-surge deposit
by lower value. This is probably due to the effect of
(varicoloured tuffs) up to 20 m thick (unit C).
transpiration at the upper boundary due to the pres-
At present volcanic activity is essentially hydrother-
ence of a grassed surface that has not been considered
mal and concentrates at La Fossa edifice. Degassing is
in the analysis.
both concentrated at fumaroles (i.e. particular points
where structural, stratigraphic and stress conditions
created preferential paths) or is diffused throughout
3 LANDSLIDES AT ‘‘LA FOSSA’’ CONE
the whole cone, favoured by the high permeability of
the pyroclastic formations.
Instabilities occur along the flanks of La Fossa pyro-
Around fumaroles pyroclastic materials are altered,
clastic cone that has a base diameter of 1200 m at sea
often up to the complete argillification; volumes and
level and is 400 m high (Fig. 16).
geometry of the argillified material (material D) are
Major instabilities affected the NE sector (where
highly variable, depending on flow paths and physico-
investigations concentrated) that is formed by a
chemical characteristics of the hydrothermal fluids.
sequence (Fig. 17) consisting from bottom to top of
Small sheet slides occur in the Varicoloured tuffs
(Dellino et al. 1990):
but the largest documented instability was the transla-
– laminated sands from dry surge deposition, up to tional slide occurred along a partly altered dry surge
150 m thick (unit A); horizon within the Punte Nere fall deposit (unit B).

211
Table 4. Physical properties of pyroclastic materials.

Mat. γ kN/m3 wn % ρs Mg/m3 Sr % e0 CF % w L , IP %

C 14.70 13.52 2.52 36–45 0.9 10

A 13.84 13.00 2.67 28.2 1.13 –
B 14.62 20.8 2.66 47.5 1.16 –
D1 15.70 74.5 2.74 100 2.04 57 75, 23
D2 14.91 68.7 2.61 95.0 1.89 63 85, 41

D1: 1988 slip surface; D2: outer cone rim.

400
dry surge normal to bedding
dry surge parallel to bedding
300 wet surge
argillified pyroclastite

200

100

0
0 100 200 300 400 500

Figure 18. Direct shear tests on pyroclastites. Figure 19. View of the 1988 slide. At the top on the right
the main scarp white altered materials is apparent.

3.1 Geotechnical characterization

Hydrothermal alteration around fumaroles has
Undisturbed samples of dry surge materials (from completely changed mineralogy, texture and hence
units A and B), varicoloured tuffs (unit C) and argilli- mechanical behaviour of the parent pyroclastites.
fied pyroclastites (material D) were taken from cuts. Argillified materials are fully saturated and have
The average values of the physical properties are medium-to-high plasticity deriving from a high smec-
reported in Table 4. tite content (D1 and D2 in Tab. 4). Materials sampled
The material forming the varicoloured tuffs is a silt at different sites show unexpected similarity, from
with sand and traces of clay. It has low saturation, apart both mineralogical and geotechnical point of view,
from local increases around fumaroles. The deposit is which indicates that alteration has the same effect
homogeneous and has the mechanical behaviour of on pyroclastites with different texture. Argillifica-
a non-cohesive soils. Direct shear tests on saturated tion drastically reduces shear strength: friction angle
specimens performed at normal stresses lower than the decreases to 26 or 21◦ depending on whether null inter-
fragile-ductile transition provide, similarly to triaxial cept is imposed or not (in the latter case a cohesion of
CIU tests, a friction angle ϕ  of 36◦ with negligible 21 kPa is obtained). Alteration also affects hydraulic
cohesion (Fig. 18). conductivity which reduces to 3.5 × 10−10 m/s.
Dry surge deposits at the base of the Punte Nere
sequence and those intercalated in the overlying fall
3.2 Instability phenomena
products are both uniform medium sands formed by
extremely irregular, clean particles at times bonded Small slips occur around the cone within the Vari-
together. coloured tuffs but the largest instability recorded at
Strength data from direct shear tests, reported in Vulcano is the 1988 slide (Fig. 19). On April 20th
Figure 18, show slight non-linearity and anisotropy about 2 × 105 m3 of pyroclastic material detached on
(ϕ  from linear regressions forced through the origin the eastern side of the cone along a dry surge horizon of
varies between 35◦ and 38◦ depending on shearing the fall deposit (B) (Tinti et al., 1999). The main scarp
direction). Otherwise if a cohesion intercept c’ is con- developed at an active fumarole and the uppermost part
sidered regression yields values of c = 30 kPa and of the sliding surface was altered up to argillification.
ϕ  = 30◦ . Seepage of water condensed from fumarole vapours

212
 3.3 Monitoring
In order to collect experimental evidence of the link
between soil suction and degassing activity, in 2007
continuous monitoring of soil temperature and suc-
tion was started by INGV-Palermo and 2nd University
of Naples at two stations (figure 16). Increase in soil
temperature revealed, in fact, to be associated to the
rise of steam output. Probes were installed in altered
pyroclastites, close to a degassing point (VGO station),
and far from fumaroles, in fresh pyroclastites (VCD
Figure 20. Steam output at fumaroles located at La Fossa station). During the preceding years, suction was also
crater (after Bukumirovic et al. 1997, re-drawn).
measured manually by means of portable tensiometres
at other locations.
1000 Soil temperature and suction are plotted versus time
April 20th landslide
1972-73 in Figure 22. In order to separate the effect of degassing
from those of evaporation and rainwater infiltration,
800 1961-62
air temperature and rainfall plots were also included.
Cumulative rainfall (mm)

1987-88
Data allow to draw some preliminary considera-
1969-70
600 tions. Soil temperature markedly increases in middle
July 2007 and maintains at the same level over all sum-
mer. During this period of more intense volcanic activ-
400
ity suction at VGO station progressively decreases in
spite of the absence of rainfall and of high air tem-
200 perature (i.e. higher evaporation). In this respect short
suction increments only occur at air temperature peaks.
The effect of the increased degassing lasts until the
0
J A S O N D J F M A M J end of November as the insensibility of suction to the
1987 1988
increased rainfall demonstrates. Successively suction
slightly increases as an effect of scarce rainfall and
Figure 21. Average cumulative rainfall between 1965 and
1995 (bar chart) and actual cumulative rainfall in particularly
normalization of degassing.
wet years (line plots). Finally, far from degassing points (station VDC),
where soil and air temperature are similar, suction is
virtually influenced only by rainfall.
was observed along fractures forming the scarp even
some years after the slide.
The slide occurred during a major period of unrest, Sudden increase Progressive normalization
in volcanic activity
characterized by intensification of seismic activity and 78
of volcanic activity
so
degassing (Fig. 20). However failure does not seem to d e i l tem
ga
s s p. c
74 ing lo
have been triggered by dynamic actions because the po s e t o
int
(VGa
strongest seismic event (M = 4, Neri et al. 1991) 70
O )
66
occurred 20 days before the slide with epicentre at 40

20 km from the island. Rasà & Villari (1991) invoke 62

30
air tem
also reiterate inflation-deflation cycles and localized 58 peratu
re
action of high-temperature hydrothermal fluids as 54
20

general causes of strength reduction in the volcano 10

flank. soil temp. far from

fumarole (VCD) 0
The analysis of hydraulic conditions (groundwater 0 20
circulation and suction), degassing and rainfall regime 0.5
VC on

at La Fossa cone suggest that failure could be initi- 16

D
at ucti

1 rainfall
s

1.5
ated by a sharp change in pore fluid pressures due to 2
12
the volcano unrest which was possibly favoured by a 2.5
8
preceding rainy wet season. The sole rainwater infil- 3 suction at VGO
3.5
tration seems instead to be not sufficient to induce 4
4

instability. Analysis of hydrologic data between 1958 J A S O N D J

and 1988 indicate that other three histories of cumu-

lative rainfall were equal or higher than that recorded Figure 22. Soil/air temperature, suction and daily rainfall
during the months preceding the slide (Fig. 21). versus time.

213
3.4 Hydraulic conditions and slope stability
Monitoring data seem to support the hypothesis that
condensed vapour increases water content of pyro-
clastites. Furthermore, observations around fumaroles
and in wells drilled at the cone foot (Madonia, pers.
com.) suggest that condensed vapour could change
seepage in the cone and hence shallow groundwa-
ter circulation. In fact continuous seepage of warm
water from fractures forming the 1988 slide scarp in t
e rv
was noticed even during the driest season of the last al
of
su
fifty years (Sept. 2003). Furthermore in fboreholes at c ti
on
co
n to
the cone foot (i.e. BL1 borehole in Figure 16) high normal inflow of urs
20
pore water temperature was measured in the dry sea- condensed vapour kP
a
son within the first 15 m, accompanied by intense
alteration. saturated a)
zone
These two elements seem to indicate that warm
water from the cone permanently circulates in the shal- inflow of condensed
lower layers. The complete displacement of the colder in t
e rv
vapour during unrest
al
sea water (the borehole top is at sea level) also sug- of
su
c ti
on
gests that flow is continuous and seepage velocity is co
nto
significant. u rs
10
kP
Evaluating amount and distribution of saturation a

changes and seepage processes in the pyroclastic ANISOTROPIC

deposits is extremely difficult. In order to make
preliminary considerations on flow paths and distribu-
b)
tion of pore pressures, steady-state and transient 2D
seepage analyses were carried out under simplified 100 m
hypotheses using water retention curves determined
in the laboratory. Analyses wee conducted through
the SEEP/W code (Geo-Slope International, 2004). Figure 23. Increase in concentrated degassing on seepage
and suction distribution in the 1988 slide area (a) consider-
Stratigraphy and boundary conditions of the model
ing or (b) otherwise anisotropy of the pyroclastic materials;
are reported in Figure 23: the seepage domain was suction contours start at zero value on left boundary.
delimited by the ground surface at the top, the slip
surface at the bottom, and the fumarole fracture at
the upper side where inflow of condensed vapour is 1988 slide mobilized a shear strength which is inter-
imposed (Fig. 23). Along the slip surface the argilli- mediate between that of the fresh and of the argillified
fied portion is fully saturated whilst downslope suction dry surge material. In the case of dry slope the percent-
is equal to average values measured in situ. Inflow age of argillified slip surface that causes failure does
is evaluated from average flow measurements in the not match in situ observations. Therefore positive pore
shallow groundwater (Madonia pers. com). pressures and/or partial filling of the tension crack at
Seepage parallel to layering establishes throughout the main scarp are to be considered.
the partially saturated varicoloured tuffs and in most Finally the drop in suction evaluated from seepage
of the underlying B unit, being more defined at the analyses in the partially saturated Varicoloured tuff
interface with the argillified horizon. The influence of blanket could be sufficient to trigger the small slips
the fumarole inflow is greatly increased accounting and sheet slides observed around fumaroles.
for material anisotropy (kh /kn = 10); under this
assumption a maximum matric suction of 70 kPa at the
slope top is calculated. If during the unrest the inflow 4 CONCLUSIVE REMARKS
is increased (four times according to the steam output
reported by Bukumirovic, 1997), suction decreases by Comparison between the results of model tests, numer-
some 20 kPa in the upper third of the slope and a con- ical simulations and in situ measurements allowed us
tinuous zone of positive pore pressures forms within to set up a reliable tool for predicting slope response
the varicoloured tuffs, which extends downslope for to rainfall infiltration for shallow layered pyroclastic
few tens of metres. deposits. The analysis of tests on Cervinara ash showed
Limit equilibrium analyses run using strength that analysis is correct if the highest saturated perme-
parameters reported in section 3.2. indicate that the ability obtained from laboratory tests is adopted and

214
the retention curve is extracted from the results of infil- Elsworth, D., Voight, B. & Taron, J. 2007. Contempo-
tration test characterized by boundary conditions and rary views of slope instability on active Volcanoes. Vol-
infiltration rates similar to real slope conditions. canic Rocks, Proc. Workshop W2—11thISRM Congress,
Comparison between homogeneous and layered Malheiro A.M. & Nunes J.C. (Eds.), Azores, 3–9. Taylor
slope models indicates the strong influence of the & Francis.
Fredlund, D.G. & Rahardjo, H. 1993. Soil Mechanics for
pumice layer interbedded between ash layers that sub- Unsaturated Soils. In Wiley-Interscience Pubblication,
stantially modify the flow regime in the lowermost ash John Wiley & sons, inc.
layer, preventing soil wetting. This result is confirmed Gardner, W.R. 1958. Some steady state solutions of the
by field monitoring. unsaturated moisture flow equation with application to
On the flanks of pyroclastic cones of volcanoes evaporation from water table. Soil Sci., 85(4): 228–232.
characterized by active hydrothermalism, intensifica- Geo-slope International 2004. SEEP/W Finite-element code
tion of degassing can result in a decrease in suction due for groundwater seepage analyses.
to localized or diffuse release of condensed vapour. At Greco, R. 2006. Soil water content inverse profiling from
La Fossa crater the phenomenon was evidenced by single TDR waveforms. Journal Hydrol., 317: 325–339.
Kunze, R.J., Uehara, G. & Graham, K. 1968. Factors impor-
in situ monitoring of soil suction and soil tempera- tant in the calculation of hydraulic conductivity. Proc. Soil
ture. In situ observations and measurements indicate Sci. Soc. Amer., 32: 760–765.
that seepage of condensed vapour is appreciable. Sim- Neri, G., Montalto, A., Patanè, D. & Privitera, E. 1991.
ple models based on the geotechnical characterization Earthquake space-time-magnitude patterns at Aeolian
of pyroclastic materials suggest the hypothesis that Islands (Southern Italy) and implications for the volcanic
variations in suction can be significant to stability surveillance of Vulcano. Acta Vulcanologica, 1: 163–169.
of volcano slopes when these are very close to limit Olivares, L. & Picarelli, L. 2001. Susceptibility of loose
conditions and if material hydraulic anisotropy is pyroclastic soils to static liquefaction—Some preliminary
considered. The validation of this hypothesis requires data. Proc. int. conf. Landslides—Causes, countermea-
sures and impacts. Davos
further monitoring data during periods of intense Olivares, L. & Picarelli, L. 2003. Shallow flowslides trig-
unrest and more comprehensive models that account gered by intense rainfalls on natural slopes covered by
for non-isothermal multiphase pore fluid pressure and loose unsaturated pyroclastic soils. Géotechnique, 53(2):
groundwater circulation, influencing the state of stress 283–288.
and hence stability. Olivares, L. & Damiano, E. 2007. Post-failure mechanics
of landslides: laboratory investigation of flowslides in
pyroclastic soils. Journal of Geotechnical and Geoenvi-
ACKNOWLEDGEMENTS ronmental Engineering ASCE, 133(1): 51–62.
Olivares, L., Damiano, E., Greco, R., Zeni, L., Picarelli, L.,
Minardo, A., Guida, A. & Bernini, R. 2008. An instru-
The research was supported by the INGV-DPCI con- mented flume for investigation of the mechanics of
tract V5/12 (2005–2007) and PRIN 2006 Project. rainfall-induced landslides in unsaturated granular soils.
Special thanks are due to P. Madonia (INGV- Submitted to ASTM Geotechnical Testing Journal.
Palermo) for monitoring data and suggestions on Picarelli, L., Evangelista, A., Rolandi, G., Paone, A.,
volcanic activity, to E. Damiano (2nd Univ. of Naples) Nicotera, M.V., Olivares, L., Scotto di Santolo, A.,
for model slope tests and to V. Savastano (AMRA) and Lampitiello, S. & Rolandi, M. 2006. Mechanical proper-
V. Grana (CNR-IGAG) for numerical simulations. ties of pyroclastic soils in Campania Region. Proc. 2nd int.
work. on Characterisation and Engineering Properties of
Natural Soils, Singapore, 3: 2331–2383.
Rasà, R. & Villari, R. 1991. Geomorphological and morpho-
REFERENCES structural investigations on the Fossa cone (Vulcano,
Aeolian Islands):a first outline. Acta Vulcanologica,
Brooks, R.H. & Corey, A.T. 1964. Hydraulic properties of 1: 27–133.
porous media. Hydrology Paper No. 3, Colorado State Tinti, S., Bortolucci, E. & Armigliato, A. 1999. Numeri-
Univ., Fort Collins, Colorado cal Simulation of the landslide-induced tsunami of 1988
Bukumirovic, T., Italiano, F. & Nuccio, P.M. 1997. The on Vulcano Island, Italy. Bulletin of Volcanology, 61:
evolution of a dynamic geological system: the support 127–137.
of a GIS for geochemical measurements at the fumarole Topp, G.C., Davis, J.L. & Annan, A.P. 1980. Electromag-
field of Vulcano, Italy. J. Volcanol. Geotherm. Res., 79: netic determination of soil water content: measurement
253–262. in coaxial transmission lines. Water Resour. Res., 16:
Calcaterra, D. & Santo, A. 2004. The January, 10, 1997 574–582.
Pozzano landslide, Sorrento Peninsula, Italy. Engineering van Genuchten, M. Th. 1980. A closed-form equation for
Geology, 75: 181–200. predicting the hydraulic conductivity of unsaturated soil.
Dellino, P., Frazzetta, G. & La Volpe, L. 1990. Wet Soil Sci. Soc. Am. J., 44: 615–628.
surge deposits at La Fossa di Vulcano: depositional and
eruptive mechanisms. J. Volcanol. Geotherm. Res., 43:
215–233.

215
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Prediction of landslide movements caused by climate change:

Modelling the behaviour of a mean elevation large slide in the Alps
and assessing its uncertainties

Ch. Bonnard
Formerly Soil Mechanics Laboratory, Swiss Federal Institute of Technology, Lausanne, Switzerland

L. Tacher
Engineering and Environmental Geology Laboratory, Swiss Federal Institute of Technology, Lausanne, Switzerland

M. Beniston
University of Geneva, Caroug, Switzerland

ABSTRACT: The consideration of predicted climate change conditions in the hydrogeological and geo-
mechanical modelling of a large landslide allows the assessment of its future behaviour in case of crisis. This
application shows that the predictions are not necessarily pessimistic, despite of the uncertainties of the needed
assumptions.

1 INTRODUCTION AND OBJECTIVES In terms of land-surface processes, and particu-

larly slope instability events, climatic factors are often
For many decades there has been a clear consensus assessed as a key factor in the triggering and/or the
within the scientific community to express various amplification of various forms of landslides, rock-
quantitative or semi-quantitative relations between cli- falls, debris flows, etc. Precipitation is certainly today
matic conditions and general landslide movements of the dominant driving mechanism for many forms
different kinds, as well as to use these experimental or of slope instabilities, through water loading in soils
empirical relations to try to predict future movements beyond a critical threshold, or through excessive runoff
(Terzaghi 1950, Wieczorek 1996). However such rela- that will lead to rapid surface erosion and debris flows.
tions have often proved to be deceiving as many Both heavy but short-lived precipitation or moderate
short-term or long-term complex factors influence the but continuous rainfall can thus provoke various forms
crises of landslides. of slope response, either in a natural or a man-made
One of the possible predictive approaches with surrounding. Extremes of temperature can also con-
neural networks has tried to combine observed past tribute to slope instability, notably through repetitive
movement data and climatic information to predict freeze-thaw mechanisms at high elevations that tend
future movements (Vulliet et al. 2000) Such a pre- to weaken rocks by progressively enlarging fractures.
diction is nevertheless reliable only in a short-term In addition, permafrost degradation in high mountains
perspective and when no critical situation is likely to resulting from milder atmospheric temperatures can
occur (Bonnard 2006), which is indeed the case when also contribute to slope instabilities by reducing the
a reliable forecast is wished! cohesion of slope material currently embedded within
On the other hand, because of the numerous impacts subsurface ice.
that a changing climate can have on many elements In a changing global climate, it is thus of interest
of the planetary environment, it is of essence to pre- to know how temperature and precipitation patterns
dict the future course of climate forced by enhanced may change, both in space and time, and also in terms
concentrations of greenhouse gases in the atmosphere. of mean and extreme conditions. Such changing con-
Predicting the speed and amplitude of climatic change ditions may result in increased or even new forms of
can thus provide a measure of guidance to decision slope hazards compared to those encountered under
makers and climate-impact specialists. current climate.

217
 It is also essential to model the complex infiltration
process of rainfall and snowmelt in large slides, in
order to be able to establish a transient distribution
of groundwater pressure at any point of the landslide
mass and within the slip surface, so as to model the
movements induced by these pressure changes in a
FEM.
The objective of this paper is thus to present the
global trend of climate change and then to illustrate its
possible long-term effect in the case of a large slide.
The obtained results will show that in some cases, the
so-called evidences are not granted for sure. It is also
important to trace and quantify all kinds of possible
uncertainties in this multiple process in order to assess
the reliability of long-term predictions.
Figure 1. Range of climate futures according to
greenhouse-gas emission scenarios; the dark gray zone
represents the more likely range of global temperature
2 TREND OF GLOBAL CLIMATE CHANGE
change (IPCC 2007). For the emission scenarios, see
Nakicenovic et al. (2000).
Perhaps the most exhaustive source of information
concerning future climatic change is provided by the
Intergovernmental Panel on Climate Change (IPCC).
In the succession of reports published in 1996, 2001 may decrease (i.e., many semi-arid and arid regions
and 2007 (IPCC 2007), a number of global climate could become even drier in the future). Precipitation
simulation models have been applied to assess the totals will probably increase by 2100 in the Monsoon
response of the climate system to anthropogenic climates of India and China, and in the inter-tropical
forcing in the 21st century, based on a number of convergence zone around the equatorial belt.
greenhouse-gas emission scenarios developed by Temperature in Europe will increase on average
Nakicenovic et al. (2000). According to the scenario, by 4–6◦ C, with strong regional and seasonal differ-
itself a function of assumptions on population growth, ences; for example, summertime warming in southern
economic growth, technological choice, and policy Europe is expected to be greater than during the winter,
decisions, the global mean temperature change over because of the positive feedback effect of dry soils dur-
present ranges from 1.5–5.8◦ C, as illustrated in ing this season (Seneviratne et al. 2006). In the Alps,
Figure 1. This represents an amplitude of change that wintertime temperatures will rise by 3–4◦ C by 2100
is probably one order of magnitude larger than changes compared to current climate, according to the level of
reconstructed for the past 20, 000 years (i.e., since the greenhouse gases. Summer temperatures may rise by
last glacial maximum), and a speed of change that is up more than 6◦ C during the same period, as a result of
to two orders of magnitude greater than typical natural positive feedback effects from dry soils and reduced
fluctuations of climate. snow and ice cover in the Alps (Beniston et al. 2007).
Climate model solutions suggest that the change Figure 2 shows the difference between summer tem-
in temperature will be stronger in the high latitudes peratures for current (1961–1990) and future climates
compared to the equatorial region. This is because of in Basel, Switzerland, not only for mean conditions,
the strong positive feedback that can be expected as a but also in terms of the upper quantiles that essentially
result of smaller areas covered by snow in the north- represent heat-wave conditions.
ern continents, a shorter winter period, and reduced Simulated results for the low emissions B2 scenario
sea-ice cover in the Arctic Ocean. Reduced snow and and the high emissions A2 scenario are shown; inter-
ice cover will substantially modify the surface energy estingly, the difference in temperature between the
balance, particularly through increased absorption of high and low emissions scenarios is less than between
solar energy. Temperature change will also be greater the B2 scenario and current climate. This implies that
over the continents than over the oceans, because of even with rather stringent policies to abate greenhouse
the larger heat capacity of the oceans. gas emissions, the increase in temperatures as seen
While a warmer climate will enhance the hydro- for the B2 scenario will result in summer heat waves
logic cycle, precipitation will not necessarily increase that are as intense, or even stronger, than the 2003
everywhere. The latest climate models published by heat wave, with an even greater potential for strong
the IPCC (2007) suggest that the northern latitudes heat waves in the A2 scenario. Indeed, statistically-
may experience greater precipitation than currently, speaking, the 2003 heat wave could occur one summer
but that rainfall in Mediterranean and arid climates out of two in a future climate (Schär et al., 2004).

218
Figure 2. Observed summer temperatures for 1961–1990
(means and 90% quantile values), and simulations for Figure 3. Percentage change in average seasonal precipita-
2071–2100 for the low emissions B2 and the high A2 emis- tion for the control and scenario climates as simulated by 4
sions scenarios established by the IPCC (2007). See text for different regional climate models; see text for further details.
further details.

events also changes in seasonality compared to current

climate. Model simulations suggest that springtime
In a future climate, whatever the emissions scenario extremes will increase the most, while summer events
considered, the freezing level will thus rise by about may decline by as much as 50%. However, when heavy
500–600 m in winter and close to 1, 000 m in summer, precipitation occurs in a warmer climate, it may be
thereby accelerating glacier retreat and exacerbat- even more intense because of the additional energy
ing the natural hazards associated with deglacierized provided by a warmer atmosphere than today.
landscapes. Positive temperatures at increasingly high In terms of the potential for floods, natural hazards
elevations will penetrate into permafrost layers, lead- and damage that the changes in means and extremes
ing to its progressive melting and thus reducing the of precipitation may trigger in the Alps, it should be
cohesion of soils. emphasized that heavy precipitation is a necessary but
Christensen & Christensen (2003) have shown that not sufficient condition for strong impacts. For exam-
northern Europe will experience more precipitation on ple, Stoffel and Beniston (2007) have shown that while
average, while in a large band stretching from France debris-flows of the past have occurred mostly during
to the Black Sea and beyond, summer precipitation is wet summers, it is conceivable that in a greenhouse
projected to diminish by as much as 40%. Simultane- climate the frequency of such events could decrease
ously, many regional climate models point to a strong because of the shifts in the occurrence of extreme pre-
increase of short-lived but very intense precipitation cipitation from summer to spring or fall by 2100. The
events in certain regions that are already prone to such response of slopes and watersheds to high precipi-
hazards, such as parts of central Europe, the Alps, tation levels varies from one event to another for a
southern France, and northern Spain. number of reasons, in particular the prior history of
In Switzerland, simulations of climate forced precipitation, evaporation, permeability of soils and
by high greenhouse-gas emissions for the period the buffering effect of snow during an event. That may
2071–2100 compared to the reference 1961–1990 lead to decrease at mean altitude. The more elevated
period shows a marked shift of the seasonality of the freezing level, the greater the potential for strong
mean precipitation (e.g., Beniston 2006), with strong runoff and high intensity of erosion since there is a
increases in winter and spring, and substantial reduc- larger surface area upon which water can flow off the
tions in summer and fall, as seen in Figure 3 for four slopes. Under current climate, the most intense events
different regional climate model simulations (the Dan- are observed to occur during the summer months,
ish HIRHAM; the Swiss CHRM; the Italian ICTP; and where the freezing level is higher than 3,500 m above
the Swedish RCAO models); while the absolute value sea level. In a future climate, on the other hand, the
of change differs from one model to another, all mod- freezing level associated with heavy rainfall will be on
els agree on the seasonal sign of change. The principal average 500 m lower because many events will take
cause of these changing patterns is related to the strong place either in spring or in autumn, at a time of the
summer warming and drying in the Mediterranean year where conditions will be cooler than for current
zone that would also affect the Alps and regions to summers. However, even if their frequency is likely
the north, and the enhanced winter precipitation that a to decrease, the magnitude and impacts of future sum-
milder climate may bring to the region, rather under the mertime debris flows, mudslides or rock-falls could be
form of rain than of snow. As a result of the change in greater than currently because of warmer temperatures
mean precipitation, the frequency of extreme rainfall and higher precipitation intensities.

219
3 TYPES OF LANDSLIDES AND SLOPES
LIKELY TO BE AFFECTED BY CLIMATE
CHANGE

It is evident that all kinds of shallow slides and impro-

perly drained engineered slopes, as well as poten-
tial debris flow creeks are likely to present a higher
hazard level in the future, as one of the major char-
acteristics of climatic conditions in the 21st century
is to display more intense storms occurring probably
with an increased frequency. Another reason for this Figure 4. Relationships between horizontal displacements
increased hazard is the always extending area of imper- and computed flux entering the La Frasse landslide in a
vious zones in the concerned watersheds, due to the cumulated representation. Thin line: infiltration issued from
development of urbanization and roadways. This situ- COUP model (considering temperature, snowmelt, vege-
ation causes higher peak floods in streams or excessive tation, soil, sun exposure . . . ). Thick line: weighted and
discharges in inappropriate sewage or drainage ducts truncated infiltration over several years in the past. The best
that are likely to divert sudden flows at the surface fit of the thick curve with the displacements (grey area) is
concluded from the shape of the curves rather than from the
of slopes through manholes or pipe failure and thus
value of the corresponding points.
generate destructive mud flows, if specific retention
works are not foreseen.
In the case of high mountain slopes, the increasing – distribution of permeability fields at the surface of
melting of the permafrost zone due to higher sum- the slide and in the landslide mass
mer temperatures can also be a cause of unexpected – possibilities of differentiated infiltration along
debris flows, even outside of a rainfall event. The loos- streams flowing on the slide as well as in cracks
ened fan material at the toe of mountain cliffs provides – occurrence of floods in the stream flowing at the toe
more sediments for the debris flows that can generate of the slide and likely to cause erosion
more extensive impacts in the valley floor; this crit- – development of urbanization and collecting water
ical situation is especially due to the intense tourist pipes and ducts
development of chalets in zones providing space with – differences of reaction of the landslide mass along
a view and easy access. the slope (swelling and depletion zones).
In all these cases there is a nearly simultaneous
occurrence between the storm triggering the land- In such cases it is also necessary to consider the
slide and the development of the landslide process effect of climatic impact not only for one specific
leading to severe visible impacts on buildings or agri- event, but for a crisis period that can be caused
cultural land, so that there is no doubt about the by antecedent conditions extending to several years
relation between the climatic conditions and the con- before the observed crisis. In the case of La Frasse
secutive damage. However in the cases of larger slides Landslide, in Switzerland, it has thus been shown
such a correlation is not evident to demonstrate and (Tacher et al. 2005) that several factors like intercep-
a nearly similar rainfall pattern can cause a slight tion of modest rainfall and evapotranspiration had to
increase of the movements of a slide during one win- be considered in order to express a good correlation
ter and a severe crisis during the next winter, as it was between infiltration flow and movements (Fig. 4).
observed and monitored at La Chenaula landslide in It is then necessary to analyse the changing climatic
Switzerland in 1982 and 1983 (Noverraz & Bonnard situation through the formulation of scenarios imply-
1992). ing different episodes along the yearly cycle. It is of
In the case of large to very large slides, extending course essential to gather enough long-term data on
on an area of one to several km2 , with differences of the movement of the slide as well as continuous dis-
elevations between scarp and toe that can reach several placement information on so as to understand its global
hundreds of meters, the situation is even more com- behaviour and to confront the observed and modelled
plex, as several factors may influence the reaction of displacement vectors.
the landslide mass, namely:

– variation of rainfall amount with elevation 4 LONG-TERM MONITORING OF THE

– offset of snowmelt episodes along the slope IMPACT OF CLIMATE CHANGE ON
– capacity of snow cover to absorb a large amount of LANDSLIDE BEHAVIOUR
rainfall before infiltration occurs
– variation in vegetation implying different intercep- In order to determine the possible effect of climate
tion and evapotranspiration patterns change conditions on the behaviour of large slides,

220
it is first necessary to gather available data provid- In a few cases, indeed, some monitored points of
ing information on their long-term movements. Such these large slides have displayed an acceleration phase
an investigation has been carried out in Switzerland that may last from several months to a year, like in the
within a national research project (PNR31), consider- case of La Frasse landslide, for which crisis periods
ing a dozen of very large slides, extending over areas of a few months duration have been recorded in 1966,
from one to some 40 km2 , and for which ancient geode- 1982–83 and 1993–94 (Tacher et al. 2005). In most
tic survey data had been collected (Noverraz et al. of the duly monitored slides, this major acceleration
1998). phase does not imply the whole landslide mass, but
In the specific case of Lumnez Landslide, in the a part of it, generally located at its toe or eventually
Canton of Graubünden in Eastern Switzerland, the in an area in which the depth of the slide is reduced.
position of the spires of 7 village churches has been Such a situation was clearly put forward in the case
regularly monitored for more than a century (up to 17 of Chlöwena Landslide in Switzerland, in 1994: the
monitoring campaigns beginning in 1887). The results crisis lasted for 4 to 5 months, with a peak velocity of
showed in general a very constant average velocity, 6 m/day at the end of July that was reduced to a few
varying from 3 to 20 cm/year (Fig. 5). cm/year at the end of September (Vulliet & Bonnard
Only one point located near the toe of the slide (vil- 1996).
lage of Peiden) displayed a clear reduction of velocity A comprehensive approach of such complex phe-
after the years 1940, which can be partially explained nomena therefore requires first a long-term mon-
by a series of dry years and then by the construction of itoring and then a detailed modelling in order to
a dam on the river Glenner flowing at its toe, upstream understand the hydrogeological and geomechanical
of the slide. As far as the annual rainfall is concerned, conditions that explain the crisis episodes. In a second
the long-term trend is not so marked in this region step it is possible to determine the probable effects
(average value of 950 mm/year—Figure 5) as in west- of climate change in a quantitative way, consider-
ern Switzerland, in which a clear increase of annual ing several crisis scenarios. Such an approach has
rainfall by some 10% has been observed since the been applied to various slides and in particular to the
years 1980 (this fact has induced the Swiss hydrolog- Triesenberg landslide.
ical service to change the long-term reference rainfall
value from the period 1901–1960 to that extending
from 1961–1990 for all rain gauge stations).
It can thus be observed that most of the very large 5 MODELLING THE CLIMATE CHANGE
slides monitored display a fairly constant velocity even CONDITIONS : THE TRIESENBERG
if the are affected by long periods of higher precipita- LANDSLIDE
tion. This fact can of course be due to their size and
depth, inducing a certain mechanical inertia, as well The Triesenberg landslide extends over a significant
as to the large storativity of their hydrogeological con- part (i.e. 5 km2 ) of the Principality of Liechtenstein
ditions. But this observation is not necessarily valid (160 km2 ), located to the East of Switzerland
for all slides. (Figure 6). It also includes two villages, Triesen at
its toe and Triesenberg at mid-slope, the infrastruc-
tures of which incur occasional damage, in particular
during crisis episodes.
The movements of this landslide are quite ancient
and date back to the end of the Wurmian period;
presently they are generally slow (i.e. some mm/year to
cm/year) in normal conditions and locally may reach
velocities of a few dozens of cm/year during severe
crisis periods. As the slide displays a relatively slow
movement, many buildings have spread on the slope
in particular during these last decades, due to the real
estate development.
The objectives of the research were the following:

– to determine the critical hydrogeological conditions

that cause an acceleration of the slide and that may
justify the triggering of an alarm system;
Figure 5. Cumulated displacements of several topographic – to foresee the behaviour of the slide under the possi-
points at Lumnez landslide over a period of more than one ble effect of climate change, so as to establish bases
hundred years. for the sustainable development of the slope.

221
 Figure 7. Geological model of the Triesenberg landslide.
The draped topographic map is lifted to display the geological
units.

Figure 6. Location of the Principality of Liechtenstein and

of the Triesenberg landslide.

The first aim of the models developed does not con-

sist in determining the possibilities of stabilizing the
overall slope, as it can be expected that such works
would by far pass the planned investments by the Figure 8. Geological vertical cross-section.
authorities of the Principality. What is aimed at is to
live with the slide, and not to slow it down.
The specific difficulties presented by the Triesen-
berg landslide mainly refer to its large area, to its the so-called Prehistoric landslide. This upper zone,
essentially unsaturated hydrogeological conditions largely inactive, is not considered to cause a driving
and to the slow movement velocities. force on the slope. Approximately, it covers 1.7 km2
After calibrating the model parameters with respect with a volume of 74 million m3 .
to the crisis of 2000, the impact of climate change – The prehistoric landslide is known by some bore-
has been analysed by modifying the boundary con- holes. It is more than 80 m deep and is made of
ditions of the hydrogeological model, on the basis flysch (clayey shales). This zone is today stabilised;
of the relevant climatic scenarios, as set up by the moreover, no movement has been observed at the
Swiss Commission for the assessment of climate toe of the landslide, where it lies under the Rhine
change (OcCC, 2004). Then the respective computed river alluvia (gravels).
groundwater pressures have been introduced in the – The active landslide (Table 1) covers the prehis-
geomechanical model, as it was done for the year 2000. toric one. It is also composed of flysch and takes
place on a slip surface located at an average depth
of 10–20 m. According to the inclinometer data,
5.1 Main features of the landslide available from 1995 to 2002, the slip surface is
approximately one meter thick.
5.1.1 Morphology and geology
The slope is oriented from North-East (up) to South- The analysis of the observed intensities and direc-
West (down). It presents some small undulations but tions of the movements (Figure 9) showed that the area
is generally fairly regular. Based on a digital terrain is indeed composed of three instability zones that can
model, the mean slope is 24◦ (Figures 7 and 8). Three be considered as independent.
parts are distinguished: This is confirmed by the reduced depth of the slip
surface close to the assessed boundaries of the three
– In the upper part, deep-seated slope movements areas and by the spatial distribution of damage to
occurred, probably at the end of the Wurmian glacial infrastructures (Frommelt AG 1997).
retreat (14,000 years); they are now underlined by The practical consequence of the decomposition of
a terrace in the topography at the top of the slope the landslide into three distinct systems is to allow
(Figure 8) and were triggered by a deep landslide, defining three different modelling areas for the 3D

222
Table 1. Main features of the Triesenberg active landslide. The yearly observation of displacements shows a
close dependence of the movements on the seasons.
Aspect Characteristics A reactivation is generally perceived in the spring,
which corresponds to the snowmelt period. This indi-
Area 3.1 km2 cates that the main driving force of the movements is
Altitudes min. 460 m, max. 1500 m a.s.l.
Length 2300 m
the variation of pore water pressure in the slope. How-
Width 1500–3200 m ever, a reactivation may also occur following a storm
Mean depth 10–20 m event.
Volume 37 millions m3 The tectonic Arosa zone (Fig. 8) is a very impor-
Mean slope 24◦ tant feature of the hydrogeological system due to its
Mean velocity 0 to 3 cm/year low permeability: a part of the Valüna Valley ground-
Soil Flysch (clayey shales) including water (Fig. 7) flows on the Arosa zone and feeds the
elements of limestone and sandstone basal surface of the landslide, causing the Triesen-
Vegetation Pasture land and some berg groundwater basin to be much larger than its
wooded zones
topographic watershed. This mechanism is proven by
Investigations Hydrogeology, boreholes with
inclinometers, GPS, RMT geophysical several observations (Tacher & Bonnard 2007).
methods, laboratory tests, modelling Such a double feeding is also effective outside inten-
Possible damage Infrastructures of two villages sive infiltration periods. Both a hydraulic balance of
the Triesenberg slope (Bernasconi 2002) and a numer-
ical model calibration suggest that about one half of
the inflow in the landslide is supplied by a base flow
from the Valüna Valley through the sandstones cover-
ing the Arosa zone (ca. 9 mio m3 /year). Groundwater
discharge occurs through some one hundred springs
distributed over the landslide, as well as at its toe, in
the River Rhine alluvia. The water table is located
about 20 to 30 m below the soil surface at the top of
the landslide, whereas at the bottom, it almost reaches
the ground surface.

5.2 Hydrogeological modelling of the year 2000

The year 2000 was chosen to perform the modelling;
during this year, a critical phase with a reactivation
of the movements was observed, showing a good cor-
relation with the snowmelt phase in April. A violent
thunderstorm also occurred on August 6th. This year
has a return period of the annual rainfall of 42 years,
according to the Gumbel law. Another reason of this
choice is the availability of calibration data for both
hydrogeological and geomechanical models. As the
slide is very thin, the unsaturated zone is of relatively
Figure 9. Total displacements between 1976/1981 and
1996/1997 (i.e. during some 20 years) in the whole insta-
high importance, which justifies computing ground-
bility area, and boundaries of the three so called independent water flows in unsaturated regime, i.e. the flows are
landslides. governed by Richards’ equation (Hillel 1980).
From the model results, in terms of volumes,
the direct infiltration reached 7.52 mio m3 in 2000,
mechanical modelling, supposing negligible kine- while the inflow through the Arosa zone was about
matic interactions between them (the hydrogeological 9.86 mio m3 (Figure 10). The cumulated rate of the
model involves a single regional area). springs reached 1.06 mio m3 , which represents only
a few percent of the total outflow; the balance flow
seeps in the Rhine river alluvia.
5.1.2 Hydrogeology The outflow curve is smoother than the inflow
Besides supplying the geomechanical models with events, due to the capacitive function of the land-
hydraulic pressures, the hydrogeological analysis and slide. Typically, the August 6th storm response was
modelling aims at a better understanding of the partic- absorbed and delayed. In May, the snowmelt episode
ular hydrogeological behaviour of the slope. did not lead to spectacular changes in the hydraulic

223
balance because, due to the slope topography, the the monitoring data. Local pervious heterogeneities
melting occurred progressively from the bottom to the that were not considered in the model accelerated the
top. For example, when the snowmelt occurred in the piezometer response to inflows in the Valüna Valley.
Valüna Valley, it had already finished on the landslide Such a delay did not occur at the beginning of August
several days to weeks earlier. since inflows concern both the Triesenberg and Valüna
More relevant from the geomechanical point of basins.
view is the piezometric behaviour. It is illustrated The numerical results suggest that the model glob-
by piezometer B8 (Fig.11). The respective calibration ally fits with reality, despite a simplification of the
was carried out by comparing the water table data with parameter fields, a rough estimation of the unsat-
the hydraulic head computed at the bottom node at this urated parameters and a minimal knowledge of the
site, i.e. at the slip surface. Both main events of the year real hydraulic balance. Computed hydraulic pressures
2000 led to a peak more than 2 m high. Just after these are thus suitable as an input in the hydro-mechanical
peaks, the head decrease was slower in the model than models in order to describe the direct causes of the
in the reality. This can be explained by the relative movements during crises.
smoothing of the parameter field, mainly over depth.
Heterogeneities are also responsible for another
observation: during the snowmelt event, the model
reacted with a delay of some days with respect to 5.3 Geomechanical modelling of the year 2000
The effect of the hydraulic head variation with time,
as determined by the hydrogeological modelling, on
the mechanical behaviour of the whole slide, has been
modelled by a FE code, Z-SOIL (2-D and 3-D), using
a Biot-type formulation, implying the conservation of
mass and momentum of both fluid and solid phases
(François et al. 2007).
In the 3-D model, the maximum displacement val-
ues are in general slightly lower, but they appear within
zones where damage has been reported (Fig. 12).
Parametric studies have also been carried out to
evaluate the effect of the selected friction angles
(between 30 and 21◦ ) and of the range of water
pressure variation (the computed data through the
hydrogeological model were multiplied by 1.25 and
Figure 10. Hydraulic balance of the model for the year 1.5 respectively).
2000. Thin line: Feeding by the Arosa zone. Dot line: Direct Both simulations display nearly linear variations
infiltration on the slope. Thick line: Outflow rates through and prove that, even in extreme conditions, it is not
springs and the River Rhine valley. expected that the movements will lead to a catastrophic
behaviour of the whole slope.

Figure 11. Measured and computed hydraulic heads in Figure 12. Distribution of the obtained displacements after
piezometers B4 and B8. Thin line: B4 measured. Thick line: 291 days of simulation, from January 1 to October 18, 2000
B4 computed. Thin dotted line: B8 measured. Thick dotted (water pressure data multiplied by 1.25) and location of the
line: B8 computed. more active zone in the central slide (3 cm).

224
5.4 Modelling of Climate change impact The target of the models is here to consider the most
and related uncertainties unfavourable scenario as far as the landslide move-
ments are concerned. Thus these worst case infiltration
According to (IPCC 2007), the air temperature should
conditions for 2050 are as follows, even if they are not
increase in the medium term, especially in summer,
the most plausible:
and the rainfall should increase in winter, but decrease
in summer. The climatic scenario for 2050 used in this – No consideration of the decreasing of gross rainfall
study is issued from the Swiss ‘‘Organe consultatif sur in summer. The infiltration curve is left intact from
les Changements Climatiques’’ (OcCC 2004), more May 1st,
specific to the North of the Swiss Alp context (Fig. 13). – Keeping the snowmelt event of the end of April,
According to this scenario, it can thus be expected – In winter, adding infiltration periods without de-
that: creasing the accumulated snow height.
– In winter, the total infiltration would increase and
In practice, the 2050 infiltration scenario implies
rain would partly replace snow accumulation. On
to add infiltration days between January 1 and April
the other hand, snowmelt at the beginning of the
20 to the year 2000 conditions. For all altitude
spring would be less important.
classes of infiltration, a 5 mm/day event is introduced
– In summer, the storm events would remain simi-
each ten days (Fig. 15). This represents an additional
lar, if not slightly worse, but the total infiltration
infiltration of 55 mm/year.
would be smaller than today because of higher
evapotranspiration.

In this study, the rainfall in 2050 is considered to

increase by 2 mm/day in winter and to decrease by
2 mm/day in summer. Those values are added as a one
year sinusoidal transformation to the records for the
year 2000. Similarly, the temperature curve for 2050
is obtained by adding a one year sinusoidal function to
the records of the year 2000, considering a warming
of 1.5◦ C in winter and 3.5◦ C in summer (Figure 14).
Considering the impact on landslides, such a sce-
nario is not obviously more severe, mainly for the
landslide zones in altitude. Indeed, besides the total
infiltration, the groundwater pressure fluctuations
have a major effect on the movements. By diminish-
ing the rather massive infiltration period of snowmelt,
the 2050 scenario smoothes out the groundwater head
Figure 14. Climatic scenario for 2050 in the Valüna valley.
curve at spring time. In particular in the Valüna valley, Upper curves: Temperatures in 2000 (solid line) and in 2050
the fast snowmelt at the beginning of May 2000 might (dotted line). Lower curves: Gross rainfall in 2000 (solid line)
be replaced by a succession of less important episodes and in 2050 (dotted line).
of rain, falling on a thin accumulation of snow.

Figure 13. Climatic scenario for 2050 for the North of the
Alps, after OcCC. Horizontal axis: months of the year. Black
bars: Rainfall change in % with respect to present average Figure 15. Scenario 2050. Infiltration conditions for alti-
seasonal values. Grey bars: Temperature change in ◦ C. tudes below 625 m.

225
 The results of both hydrogeological and geo- Dr Riccardo Bernasconi geological office for supply-
mechanical models with such modified boundary con- ing data and advice, as well as all the colleagues who
ditions are very similar to those obtained for the year participated to the modelling.
2000. Typically, the hydraulic heads in piezometer B8
(Fig. 11) are changed by some centimetres only.
Considering uncertainties on climatic changes, the
REFERENCES
modelled scenario appears to be probably the worst
case. In such conditions, the computed velocity field Beniston, M. 2006. The August 2005 intense rainfall event in
for 2050, if the parameter calibration on year 2000 is Switzerland: not necessarily an analog for strong convec-
considered as reliable, is a rather pessimistic global tive events in a greenhouse climate. Geophysical Research
assessment of the slope. However, the transition to a Letters 33, L5701.
stormier climatic regime may have local consequences Beniston, M., Stephenson, D.B, Christensen, O.B., Ferro,
(hectometric slides, mudflows) not considered in this C.A. T., Frei, C., Goyette, S., Halsnaes, K., Holt, T.,
regional modelling. Jylhä, K., Koffi, B., Palutikof, J., Schöll, R., Semmler,
T., and Woth, K., 2006. Future extreme events in Euro-
pean climate; an exploration of regional climate model
projections. Climatic Change, 81, 71–95.
6 CONCLUSIONS Bernasconi, R. 2002. Tiefbauamt des Fürstentums Liecht-
enstein - Hangsanierung Triesenberg—Hydrogeologische
A detailed hydrogeological and geomechanical mod- Überwachung—Ergebnisse der Markierversuche Valüna-
elling as it was recently applied at the Triesenberg tal 1999/2000, Hydrogeologischer Bericht Nr. 1124–04,
August 2002.
and La Frasse landslides allows a significant mod- Bonnard, C., Forlati, F. & Scavia, C. 2004. Identification and
elling of large landslide movements during crises, mitigation of large landslide risks in Europe : advances
provided sufficient information is available. The appli- in risk assessment. IMIRILAND Project. 317 p. Leiden:
cation of predicted climatological conditions in the Bal-kema, ISBN 90 5809 598 3.
future then supplies quantitative values of possible Christensen, J.H. & Christensen, O.B. 2003. Severe summer-
movements, considering appropriate scenarios. How- time flooding in Europe. Nature, 421, 805–806.
ever, extremely rare conditions with a very remote COUP model software, Per-Erik Jansson, Department of
probability cannot be modelled reliably, as the bound- Land and Water Resources Engineering, Royal Institute
ary conditions may significantly differ from the ones of Technology, Stockholm.
François, B., Tacher, L., Bonnard, Ch., Laloui, L. &
considered in the original model. Triguero, V. 2007. Numerical modelling of the hydroge-
The analysis of several large landslides in other con- ological and geomechanical behaviour of a large slope
texts (Bonnard et al. 2004) also shows that the effect of movement: The Triesenberg landslide (Liechtenstein).
climate change on landslides within the next 50 years Canadian Journal of Geotechnics 44:840–857.
or so must not be overemphasized. Indeed, as shown Frommelt, A.G., Ingenieurbüro, Vaduz. & Hangbewegun-
here, the progressive snowmelt that will begin earlier gen, Triesenberg. Verschiebungsmessungen 1978/81–
than before tends to reduce the occurrence of critical 1996/97, 1997.
situations in the spring or summer. On the contrary, it Hillel, D. 1980. Fundamentals of soil physics. Academic
is clear that the expected increase of storm intensity, Press, New York.
IPCC, 2007. Climate Change. The IPCC Fourth Assessment
as foreseen by some climatologists, may produce more Report. Cambridge University Press, Cambridge, UK.
violent and frequent small slides and debris flows; but Nakiæenoviæ, N., et al., 2000: IPCC Special Report on Emis-
this specific prediction is not relevant for large land- sions Scenarios, Cambridge University Press, Cambridge,
slides and cannot justify a development of more severe UK.
disasters related to this type of phenomena. Indeed, Noverraz, F. & Bonnard, Ch. 1992. Le glissement rapide
due to the heterogeneity of the material at a large scale, de la Chenaula. Proc. Symp. INTERPRAEVENT, Berne.
to the increased range of altitude where infiltration 2:65–76.
occurs, to the capacitive function of the landslide mass Noverraz, F., Bonnard, Ch., Huguenin, L., Dupraz, H. 1998.
and to the more complex hydraulic relationships with Grands glissements de versants et climat. Projet VERSIN-
CLIM, Comportement passé, présent et futur des grands
the bedrock, the response to climatic events may be versants instables subactifs en fonction de l’évolution
significantly smoothed and delayed, which explains climatique, et évolution en continu des mouvements en
this relatively optimistic vision. profondeur. Rapport final PNR 31, FNRS, Berne, 1998.
Zürich : V/d/f, 314 p.
OcCC, 2004. Die Klimazukunft der Schweiz—Eine prob-
abilistische Projektion, Christoph Frei, Institut für
ACKNOWLEDGMENTS Atmosphäre und Klima, ETH Zürich.
Schaer, C., et al. 2004. The role of increasing tempera-
The authors wish to thank the authorities of the Prin- ture variability in European summer heat waves. Nature
cipality of Liechtenstein for supporting this research, 427:332–336.

226
Seneviratne, S.I., et al. 2006. Land-atmosphere coupling and Tacher, L. & Bonnard, Ch. 2007. Hydromechanical mod-
climate change in Europe. Nature 443:205–209. elling of a large landslide considering climate change
Stoffel, M. & Beniston, M. 2006. On the incidence of debris conditions. Lecture at International conference on ‘Land-
flows in the Swiss Alps since the early Little Ice Age slides and Climate Change—Challenges and Solutions’.
and in a future climate. Geophysical Research Letters, 33, Ventnor, Isle of Wight, UK. 21–24 May 2007.
L16404. Vulliet, L. & Bonnard, Ch. 1996. The Chlöwena landslide:
Tacher, L., Bonnard, C., Laloui, L. & Parriaux, A. 2005. Prediction with a viscous model. Proc. VIIth Int. Symp.
Modelling the behaviour of a large landslide with respect on Landslides, Trondheim Vol. 1:397–402.
to hydrogeological and geomechanical parameter hetero-
geneity. Landslides Journal 2 (1):3–14.

227
Geology, geotechnical properties
and site characterization
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Geotechnical appraisal of the Sonapur landslide area, Jainita hills,

Meghalya, India

R.C. Bhandari, P. Srinivasa Gopalan & V.V.R.S. Krishna Murty

Intercontinental Consultants and Technocrats Private Limited, New Delhi, India

ABSTRACT: The occurrence of landslides particularly on cut slopes along the roads with in Jainita hills in
Northern Eastern part of Himalayas of Meghalaya state in India are common features. These slope failures
causes considerable loss of life and property along with many inconveniences such as disruption of traffic along
highways. The paper present deals with geotechnical synthesis of slides as per recommendations of RHRS
system and stabilization measures suggested. The landslide is located on North South trending ridge on eastern
bank of river Sonapur. The rocks involved in sliding are highly jointed sandstone shale of Oligocene series. The
total inclined length of the affected slopes is around 800 meter. The jointing in rock is attributed to the nearness
of the area to a major thrust. These studies indicate the failure along the slopes is ‘‘Rock fall-cum debris flow’’.

1 INTRODUCTION October. The month of June and July experiences max-

imum precipitation. The annual precipitation varies
1.1 Occurrence of land slide from 4000 mm to 8000 mm during the year maxi-
mum precipitation of 1200 mm is recorded in month of
A land slide occurs when due to gravity forces the
June and July. Maximum temperature reaches to 24◦ C
rock/soil mass moves down wards due to heavy pre-
in month of September while minimum temperature
cipitation, run off or ground saturation. The flow
recorded is 10◦ C in month of February.
occurs generally during period of intense rainfall, on
steep hill slopes where the rocks are tectonically dis-
turbed. The flow/fall from many different sources can
combine in channels, and their destructive power is 2 REGIONAL GEOLOGY OF THE AREA
greatly increased. They continue flowing down hills
and through channels, growing in volume in addition 2.1 Rock types
of water sand, mud, boulders, up rooted trees and The rock types in the region available, varies in age
other material. When the flows reach flatter ground, from Archeans to Tertiary the general sequence of the
the debris spread over a broad area thus affecting the rocks available in the area is as follows (Figure 1).
considerable length of road.

1.2 Location of area 3 LOCAL GEOLOGY

The occurrence of land slides, particularly along the AND GEOMORPHOLOGY
road in the Jaintia hills of Meghalaya state is common
feature during rainy season. This cause considerable 3.1 Geology
loss of life and property along with many inconve- The Sonapur landslide is located on North–South
niences such as disruption of traffic between Shillong trending ridge, on east bank of river Sonapur/Lubah.
and Silcher every year. Sonapur land slide occur- The river Lubah flows at elevation 33.0 m from mean
ring near vicinity of township and is located between sea level, and the present road elevation 44.0 meter.
KM 141.100 to KM 141.350 on Shillong—Silcher The highest elevation at top of the crown of slide is
highway. 481.25 meters. Thus the total height affected by slide
along the slopes are also observed slopes is around
440 meters. The rock exposed in the area are sand
1.3 Climate
stones and shales of Oligocene age. The rocks are
The climate of the area is moderate. The rainfall occur highly jointed and local folding along the slopes are
during monsoon period which extends from May to also observed.

231
Table 1. Regional geology. 4 HISTORY OF SLIDE

Rock types Group Age For information see table on next page.
Feldspathic sandstones,
pebbles and conglomerates Brail Oligocene
Shales Sandstone and marls 5 STRUCTURAL ANALYSIS
(Kopiliformation) Garo Miocene
Limestone and Sandstone 5.1 Geometry of slide
(Sylhet formation) Jaintia Eocene
Gneissses and granites Gniessic Archeans The geometry of the slide is as under. (For plan and
complex section of land slide area please refer to Figures 2
and 3).
The area is being drained by two prominent
drainages. The drainage, which is located on southern
side is being continuously recharged through minor

Figure 1. Regional geological map of the area.

Figure 2. Geological plan of slide area.

3.2 Geomorphology
The area is occupied by undulating topography char-
acterized by presence on hills and valleys. The area
is dissected by number of streams and network of
their tributaries. The drainage pattern in the entire
district represent a most spectacular feature revealing
extra ordinary straight course of rivers and streams,
evidently along master joints and faults which are
impressions of major geological activity in the area.
The magnificent gorges scooped out by the river in the
southern part on Jainitia district are result of massive
headword erosion by anticedent streams along joints
of sedimentary rocks, exposed in the area. Figure 3. Section along the central line of slide area.

232
 Table 2. History of slide.

Date of occurrence Possible cause Remedial steps taken Remarks

1998 July i) Loss of cohesion in soil along slopes due to high Clearance of debris Slide of high intensity
precipitation Construction of gabion walls and road level Road closed for more than one month
ii) Earth quake of 1987, possibly resulted loosen- diversion the drain
ing of shear strength between jointed blocks
1989 September Flow of debris material from higher slopes Clearance of debris and maintenance of remedial Intensity was less; road was blocked daily for
measures taken were same as in 1998 1–2 hours for 15 days

1999 June–July Rolling down of boulders from top 200 m above Clearance of debris Road blocked 2–3 days
road level along slopes Diversion of water at road level Land slide of less magnitude but size of boulders
ii) Repairs of gabion structures at road level was 5.5 m × 1.8 m × 4.5 m

233
2000 Aug–Sep Rolling down of boulders from top 200 m above Clearance of debris Land slide frequently occurred. The road was
road level along slopes Diversion of water at road level blocked for total period of 22 days
ii) Repairs of gabion structures at road level
2001 01 to 15 June Loss of cohesion in soil along slopes due to heavy Construction of Retaining wall, Check wall water High-magnitude slide, Boulders of 3.75 m ×
rains chute 2.5 m × 8.8 m rolled down along slopes
Concrete pavement and geotextiles Road blocked for 10 days one oil tanker, truck
with loaded goods, one passenger bus buried
down. Number of casualties not known

2004 July Rolling down of boulders from top 200 m above Construction of culvert Retaining wall, Wire Total number of incidence occurred. Four road
road level along slopes crated wall with jute mesh was closed for 5–6 days. Size of boulders
2 × 2 × 3 and 7 × 10 × 3
Table 3. Geometrical parameters of slide. 6 SLOPE STABILITY ANALYSIS
AND ROCK FALL MECHANISM
Base—Width at road level 300–350 m
The geological structures of the study area are charac-
Width at center 180 m
terized by strongly jointed rocks of variable strength
Crown—Width 70 m
Inclined Length (from River Bed to mainly sand stone, and shales of Oligocene age. The
Crown of slide) 800 m (approx) synthesis of geological date indicates that the slope
Slope from river bed level to point A 35◦ stability problems in the area are associated with sur-
Slope from Point A to B 25◦ charging of slopes during heavy precipitation and with
Slope from point B to C (Crown) >60◦ inflow of ground water from fissures. A great loss of
strength of the rock mass results, particularly in zone
of weathering and causes its subsequent displacement
Table 4. Wedge analysis. along slopes.

Amount and Relationship with

Wedge Intersection of direction of slope stability (on 1. The dislodged blocks from higher reaches rolls
No. Joint set plunge southern face) down along the slopes are mixed with debris and
moves down along slopes, thus forming the slides
1 1 and 2 50◦ S 50◦ W Forms slope as ‘‘Complex Slide’’. The schematic geological
most unstable plan and sections are shown as Figures 2 and 3
as the axis of respectively.
wedge plunge
towards open
2. The slope analysis is divided in two parts
face
2 2 and 3 44◦ : S 33◦ W Forms i. Rockfall along the exposed rock faces of slide
unstable above 200 meter high along slopes with respect
slopes as axis to road level.
of wedge ii. Movement of pre-existing slide debris along the
plunge slopes due to high degree of saturation during
towards face high rainfall from June to Oct every year in the
by angle of
area. The total rainfall in the year 2005 was
<45◦
3 1 and 3 50◦ : S Forms stable 1662.05 mm with maximum rainfall alone in
slopes as the month of Aug 2005 718.
wedge is
dipping inside 3. The Southern and Eastern part of slide at higher
the face levels are having vertical to sub vertical slopes. At
places escarpments of 20–30 meter high are seen.
Along the vertical scarp faces, the sand stone with
minor bands of shale are exposed. The shale bands
springs emerging from the escarpments on southern are having a thickness of 25 cm to 50 cm and are
face of slide. The spring mostly emerges from the very soft and fragile, where as sand stone are hard
contact plane between shale and sand stone. and massive.
The rock falls in the area are guided by following
factors.
5.2 Discontinuity analysis
The prominent sets of geological discontinuities i. Orientation of geological discontinuities, and
observed at road level are: their relation with respect to exposed faces.
Joint set (1) Strike—N 50◦ W–S 50◦ E—Dip 50◦ : ii. Increase in pore—water pressure, the heavy
S 40◦ W rainfall in area, and infiltration of water
Joint set (2) Strike—N 20◦ E–S 20◦ W—Dip 65◦ : through open joints—causes changes in forces
N 70◦ W acting on a rock mass.
Joint set (3) Strike—N 70◦ E–S 70◦ W—Dip 50◦ : iii. Erosion of surrounding materials during heavy
S 20◦ E rainfall, chemical degradation, or weathering
The simple stereographic projection method used of rock.
for ‘‘Wedge—analysis’’ along slopes for evaluating
slopes stability due to intersection of above mentioned 4. The field observations shows that the wedges
joint planes indicates, that the following wedges are formed by intersection of Joint set 1and 2, 2 and 3
formed. have plunge towards SW, i.e, towards exposed face,

234
 these joints are clay filled and are prominent with Table 5. Analysis as per—RHRS—parameters.
shales.
Parameter Observation Points
The blocks rolling from top of slopes, particu-
larly from southern face causes rock fall hazards. Slope height 100 ft 81
The rock fall occur when a block is suddenly Ditch effectiveness Good catchment 3
released from apparently sound face, of sand stone Average vehicle risk 25% of time 3
with thick layers of shale this is due to small Percent decision of Adequate sight 3
deformities on surrounding rock mass. This phe- sight distance distance 100% of
nomenon is anticipated when the forces acting low design value
Road way width 9
across discontinuity planes, changes due to pore
widening paved
water pressure developed on account of presence shoulders 36 feet
of water with in joint planes which isolates a block Geological character Discontinuous joints 9
from its original position due to reduction of shear with random
strength between the contact planes of shale and orientation
sand stone. This phenomenon is attributed to long Rock friction Clay-filling or 81
term detoriation due weathering along joint planes, slickensided joints
dipping towards slopes and results the release of Structural condition Occasional erosion 9
‘‘Key-Blocks’’ causing rock falls of significant features
Difference in erosion Moderate difference 9
size, at some time larger blocks having diameters
rates
7 × 3 × 3 m are also released from higher slopes. Block size >4 81
Once the movement of a rock, on top of crown Quantity of rock fall/ 12 cubicyard 27
of slide scar, has been initiated, its trajectory is event
controlled by geometry of the slope. As the slopes Climate and presence High precipitation 27
are almost vertical, the southern face acts as ‘‘SKY of water in slope period or continual
JUMP’’, and imparts a high horizontal velocity to water on slope
falling rock, causing it to bounce a very long way. Rock Fall history Many falls 27
The downward slopes from 250 m up to river Total score 369
bed are covered by boulders of gravels, mixed with
clay and this bed absorbs a considerable amount of
energy of falling rock, and movement of boulders
is reduced in many cases are even stops completely. less than 300 are assigned, a very low priority where
The slopes from river bed level to 200 m high as the slopes with rating excess 500 are identified as
are covered with loose pre-existing slided mate- areas requiring urgent remedial measures. Since in
rials with gravels and boulders mixed with clay. the present case the total rating is more than 300 but
Such type of materials is having very low cohe- less then 500, the area falls under category requiring
sion and angle of internal friction. The material long term planning for slope stabilization after detailed
when charged with water forms a slurry, and has geological investigation.
a tendency to flow along the slope with signifi-
cant velocity, thus causing damage to protection
measures provided at road level. 8 REMEDIAL MEASURES ADOPTED

The following remedial measures have been adopted

7 STUDY OF AREA AS PER ROCK FALL so sorto control the damage at road level time to time.
HAZARD SYSTEM (RHRS) 1. Construction of retaining wall at road level
2. Construction of culvert at road level
The Highway and railway construction in mountain- 3. Diversion of drain at road level
ous regions present a special challenge to engineers 4. Construction of check dams across perennial drains
and geologists. The rock fall hazard rating system along slope.
‘‘RHRS’’ was developed by Pierson et al 1990. The 5. Concrete pavement
following analysis of Sonapur slide as has been made 6. Construction of chute along the slope
as per Rock Fall Hazard Rating System.
The RHRS system does not include recommenda- To keep the road free for vehicular movement
tion on actions to be taken for different rating this through-out the year, as a part long term planning
is because decisions on remedial action for specific the road maintenance Engineers are proposing con-
slope depends upon many factors such as budget allo- struction of twin landslide gallery, Alternatively con-
cation for highway work. However slopes with rating struction of approx. 500 m long tunnel by passing

235
slide—zone has also been recommended as a measure i. Construction catch pit at RL ± 200 m of (bigger
to the traffic smoothly by passing the slide zone. dimension 25 × 10 × 5) to arrest the movement of
water, boulders coming from up slopes
ii. Construction of cascading chute from catch pit to
9 CONCLUSION AND RECOMMENDATIONS road level along perennial drainage.
iii. Construction of gabion wall at road level.
The rock falls, are due to falling of blocks of sand iv. Construction of culvert at road level.
stone/shale from top i.e. near crown area of slide. v. Construction of chute from Invert level of culvert
Which are generated due to intersection of joints, with at road level to river bed for an inclined length of
in sand stone and shale. 50 m.
The pre-existing old landslide debris material lying
from river bed level to a height of around 200 meters
along slopes is mixed with clay and boulders. The ACKNOWLEDGEMENT
loose blocks resting on saturated slopes increases
the load on soil. Under these circumstances when Authors are thankful to Chairman and Managing direc-
the soil gets saturated with water, looses its cohesion tor ICT Shri. K.K. Kapila for encouraging the authors
and angle of internal friction added with increase of to write the manuscript of this paper.
overlying weight, starts flowing with velocity along
slopes. When such material with great momentum hits
out at remedial structures such as retaining wall, breast REFERENCES
wall, water chute concrete pavement and Geo-textile
causes complete damage of these structures. Hoek & Bray J.W. 1981. Rock slope engineering. Revised
The studies carried out earlier has suggested con- third edition. The institution of Mining and Metallurgy,
struction of ‘‘Twin Slide Shed’’ at the road level, allow London.
to pass the debris material over it, alternatively con- Pierson L.A., Davis S.A. & Van Vickle R. 1990. Rock fall
hazard Rating system—implementation manual. Federal
struction of approximately 500 meter long tunnel has highway administration (FHWA) Report FHWA OREG
been also recommended. 90-91. FHWA Department of transportation.
By studying the survey data and detailed section
of land slide area. The protection measures which is
appears to be more appropriate are

236
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

The viscous component in slow moving landslides: A practical case

D.A. González, A. Ledesma & J. Corominas

Department of Geotechnical Engineering & Geosciences, Tech. University of Catalonia (UPC),
Barcelona, Spain

ABSTRACT: The availability of continuous records of both velocities and groundwater table in monitored
landslides has increased the interest of the scientific community for the dynamics of the slow moving landslides. In
this paper we analyse the role of the viscous component in large landslides by using ten years record of monitoring
data from the Vallcebre translational landslide, located north of Barcelona (Spain). Previous research showed
that a viscosity term should be considered in order to reproduce the measured displacements in this landslide.
We discuss here the hysteretical behaviour of the landslide velocity records observed during some acceleration
events produced by the rise and withdraw of the groundwater table. The conceptual model we have used shows
that a constant viscous component of the movement is not able to explain the hysteretic behaviour and that
consequently, other mechanisms should be searched to explain such a behaviour.

1 INTRODUCTION

In many slow moving landslides a close relationship

between landslide velocity and position of the ground-
water level has been observed. Acceleration changes in
these landslides are usually controlled by water table.
Despite this evidence, research works describing in
detail the dynamics of such kind of relations are still
scarce. This is partly due to the difficulty in the past to
measure, in a continuous manner, both the velocities
and the position of the groundwater table.
Nakamura (1984) showed for a particular landslide
in Japan that the amount of landslide movement in
the rising limb of groundwater level is larger than that
observed in the lowering limb for the same ground-
water level. Figure 1 shows that points indicating the
groundwater level at the time of measurement line
up in a circle counterclockwise with time. A num-
ber at the side of each point indicates the month, and
‘‘e’’, ‘‘m’’ and ‘‘l’’ indicate early part, middle part and
the last part of the month, respectively. The author
compares the velocity of the landslide (amount of dis-
placement in 10 days) in different events for one year.
The landslide response is influenced by the different
Figure 1. Groundwater level in relation to amount of land-
soil conditions and particular characteristics of each slide movement for a landslide in Japan. After Nakamura
event. (1984).
Bertini et al (1984, 1986) and Picarelli (2004) per-
formed a similar analysis, based on measurements in
the Fosso San Martino landslide (Italy) suggesting dif- velocities are higher when water table is increasing
ferent velocities for the rising and lowering limb of the than when it is decreasing.
piezometric levels (Figure 2). In this paper we analyse the relationship between
Van Asch et al (2007) coincides with the latter groundwater level changes and the landslide displace-
authors saying that for a given groundwater level, ments and velocities using data from the Vallcebre

237
 been carried out every 20 minutes. Piezometric read-
ings have indicated that changes in groundwater lev-
els occur quickly. In-hole wire extensometers have
recorded sudden changes in displacement rates that
can be directly related to the fluctuations of the water
table which is governed by rainfall (Corominas et al
1999).
The wire extensometer measurements show that the
landslide has never stopped completely. It has been
moving since we started the continuous monitoring in
November 1996, although velocities slow down sig-
nificantly during dry periods (Corominas et al 2000).
On the other hand, the history of displacement of the
extensometers reflects that different parts of the land-
slide mass move synchronically but with a different
rate of displacement.
Figure 2. Displacement rate versus piezometric level for
Fosso San Martino landslide. After Bertini et al (1986).
2.3 Hydrological changes and landslide response

landslide. The analysis is restricted to acceleration The data show that groundwater reacts almost imme-
events occurring in the period between November diately to rainfall inputs, suggesting that water infil-
1996 and August 1998. To check whether these results tration is controlled by fissures and pipes rather than
are consistent, finally a basic conceptual model is by soil porosity. The role of the karstic network in the
presented as well. gypsum lenses is unclear but all the observed field
features are very shallow (up to 3 m depth), which is
well above of the normal groundwater level fluctua-
2 THE VALLCEBRE LANDSLIDE tion. Because of this, we have assumed that karstic
network (piping) play only a secondary role.
2.1 General setting A close relationship between the groundwater level
changes and landslide activity was observed at bore-
The Vallcebre landslide is a large, active slope failure
hole S2 (Figure 3). There exists a strong level of
located in the upper Llobregat river basin, in the East-
synchronism between the two records.
ern Pyrenees, 140 km north of Barcelona, Spain. The
Figure 4 shows an interesting relationship between
landslide is situated on the western slope of the Serra
observed velocities and the depth of water table at
de la Llacuna.
borehole S2 for the period considered. A cubic curve
The mobilised material consists of a set of shale,
may be fitted to the data. These data are going to
gypsum and claystone layers of continental origin
be analyzed thoroughly by considering the landslide
gliding over a thick limestone bed, all of which are
acceleration events separately. Each event is defined
of Upper Cretaceous—Lower Palaeocene age. The
by the rising and lowering groundwater level limb.
dimensions of the slide mass are 1200 m long and
Note in Figure 4, that when depth of groundwater
600 m wide. The entire landslide involves an area of
table is close to 6 m, velocities tend to be nil. That is,
0.8 km2 that shows superficial cracking and distinct
there is a level of water table below which landslide
ground displacements. stops.
The toe of the landslide extends to the Vallcebre In a previous analysis (Corominas et al 2005), we
torrent bed, and is pushing it towards the opposite have considered that beside frictional resisting forces,
bank. As a result of this, the Vallcebre torrent has been additional resisting forces (i.e. viscous forces) were
shifted to the west more than ten meters and the foot necessary to explain the rate of displacement of the
of the landslide has overridden the opposite slope to landslide.
form a back tilted surface. The torrent undermines the
landslide toe during floods, causing erosion and local
rotational failures which decrease the overall stability. 3 ANALYSIS OF THE DATA
A comprehensive description of the landslide is found
in Corominas et al (2005). 3.1 Field data observations
The data shown in Figure 4 correspond to different
2.2 Monitoring of the landslide
rainfall events occurred in the period considered with-
Since 1996, systematic recording of rainfall, ground- out discrimination whether the groundwater level is in
water level changes, and landslide displacements has a rising or lowering limb.

238
Figure 3. Velocities and water table depths for January, February and March 1997.

Figure 4. Velocities versus water table depths for November

1996 to August 1998 period. Data correspond to mean daily Figure 5. Velocities versus water table depths for December
values. 1996.

Figure 5 shows the relationship between ground- 1997–January 1998. Figure 6 shows that the landslide
water level changes and landslide velocities for one velocities of the rising limb and those of the lowering
of the biggest rainfall events of the study period, that limb do not coincide. Velocities of the rising limb are
of December 1996. This figure shows that velocities slower than the velocities in the lowering limb which is
in both limbs (rising and lowering) are very close. opposite to what has been found by authors mentioned
The difference between two points with the same previously (in particular, Nakamura, 1984, Bertini
groundwater level is negligible. et al., 1986).
However, other events of the same period of data This particular behaviour was also observed in other
show a completely different behaviour. Figures 6 and 7 events (Figure 8) mainly for small rainfall events where
show the changes in velocities and groundwater table the difference between the values of velocity for the
for the events of January–March 1997 and December same groundwater level is very small.

239
 At the beginning of the event, velocities increase
gradually. Close to the maximum level of water table,
velocities increase very quickly. Velocities are still
high when water table starts to withdraw but eventually
they decrease as well.
Once the rainfall has stopped the pore water pres-
sures start to decrease and also the velocity with a
decreasing rate which is different to the one during the
rising limb.
Therefore, a preliminary conclusion for this land-
slide is that the landslide-velocity response to the
rise and withdraws of water table depends on the ini-
tial hydrological conditions of the ground and on the
magnitude of the event.

Figure 6. Velocities versus water table depths for January, 3.2 Theoretical Analysis
February and March 1997. An attempt to simulate the landslide hydromechanical
behaviour was considered, and for that purpose, a basic
conceptual model based on the classical equilibrium
differential equation was developed:

 d2 u du
F=m· 2
+C· +k·u (1)
dt dt

where F = Equilibrium forces; m = mass; u =

displacement; t = time; C = damping coefficient
related to viscosity; and k = stiffness.
The analysis is based on the equilibrium of a infinite
slope, using the basic equations described in Coromi-
nas et al (2005). A representative soil block with unit
length and width and the initial parameters presented
in Table 1 were used for the analyses. The stiffness
was assumed nil and the soil parameters were obtained
Figure 7. Groundwater table depths versus time. Event for from laboratory experiments simulating residual con-
January, February and March 1997. ditions on ring shear equipments allowing for large
displacements.
Two different situations were considered: i) Model
including a viscous component; and ii) Model without
a viscous component (i.e., C = 0).
In Figure 9 the relationship between velocities and
depth of groundwater level taking into account a vis-
cous component is shown. Note that with this model,

Table 1. Initial model parameters from the Vallcebre

landslide.

Setting Value

Damping Coefficient (C) 2.46 × 1011 Ns/m

Stiffness (k) 0.0
Slope angle (α) 10◦
Friction angle (φ) 14◦
Figure 8. Velocities versus water table depths for December Cohesion (c ) 0.0
1997 and January 1998.

240
the response of velocities to the rising and lowering of velocity versus groundwater table. Therefore, a
limb of groundwater level is the same. hysteretic behaviour seems to be related either with
The results corresponding to the other case analyzed a non-constant viscosity or with another effect not
is shown in Figure 10. Now the viscosity term has been considered in the conceptual model. For instance, in
neglected. Note that in the rising limb the velocity is Vallcebre the toe of the landslide is eroded by the Vall-
increasing and in the lowering limb continues increas- cebre torrent in the case of heavy rains and this may
ing without deceleration of the mass movement when affect the equilibrium conditions. In such particular
the groundwater level is low. That is, the viscosity situations, one may expect a different behaviour of the
provides with a mechanism for dissipating energy in landslide in the rising limb or in the lowering limb of
the system, and eventually helps reducing velocity and the velocity curve.
stabilising the landslide.
This is the result of a simple analysis and a simple
model, but it may be useful to understand the dynamics 4 CONCLUSIONS
of the landslide and the effect of the viscosity term in
the movement. The recorded data available for Vallcebre landslide
According to this conceptual model, a constant has been used in this work to analyse the dynamics
viscosity will give a ‘‘reversible’’ response in terms of the movement during rising and lowering of the
ground water table. The velocity of the movement has
been considered as main variable to be analysed as
dependent of the ground water table level.
In some rainfall events the velocities in the rising
limb are very close or the same that in the lowering one
and comparing this case with the theoretical analysis
implies that the mechanism of slope movement has to
include a viscous component. However, other events
have shown a different behaviour with velocities in the
rising limb lesser than velocities in the lowering limb.
A simple conceptual model based on the infinite
slope analysis was considered to understand the rela-
tionship between velocities and groundwater table. It
was found that for a single event one should expect a
‘‘reversible’’ behaviour of velocities, and that has been
measured in many situations in Vallcebre.
However, sometimes a hysteretic effect has been
Figure 9. Results from the model: velocities, u versus water observed in single events, a situation that has been
table depths, H-d, including a viscous component. reported by other authors as well. That may be due to
a non constant value of the viscosity, or to changes in
the conditions acting on the landslide, i.e. the toe of
the moving mass has been eroded by the torrent and
therefore the conditions when rising or lowering the
water table can not be compared directly.

ACKNOWLEDGEMENTS

First author is on leave from University Centre-

occidental ‘‘Lisandro Alvarado’’ from Barquisimeto,
Venezuela and express his thanks for the financial
support. Moreover, grant provided by Gran Mariscal
de Ayacucho Foundation (FUNDAYACUCHO) is also
recognized.
This research work has been supported by the
Spanish Science and Education Ministry (MEC),
Figure 10. Results from the model: velocities, u versus project number CGL2005-05282 (MODEVALL) and
water table depths, H-d, without a viscous component. by the Institute of Geomodels (UPC-UB-CSIC).

241
REFERENCES Corominas, J., Moya, J., Lloret, A., Gili, J.A., Angeli, M.G.
& Pasuto, A. 2000. Measurement of landslide displace-
Bertini, T., Cugusi, F., D’Elia, B. & Rossi-Doria, M. 1984. ments using a wire extensometer. Engineering Geology,
Climatic conditions and slow movements of colluvial 55: 149–166.
covers in central Italy. Proceeding of the 4th Interna- Corominas, J., Moya, J., Ledesma, A., Lloret, A. & Gili, J.A.
tional Symposium on Landslides. Toronto, Canada. 1: 2005. Prediction of ground displacements and velocities
367–376. from groundwater level changes at the Vallcebre landslide
Bertini, T., Cugusi, F., D’Elia, B. & Rossi-Doria, M. (Eastern Pyrenees, Spain). Landslides 2: p. 83–96.
1986. Lenti movimenti di versante nell’Abruzzo Adri- Nakamura, H. 1984. Landslides in silts and sands mainly in
atico: Caratteri e criteri di stabilizzazione. Proc. XVI Japan. Proc. IV Int. Symp. On Landslides, Toronto 1984,
Convegno Nazionale di Geotecnica, Bologna maggio 1: 155–185.
1986: 91–100. Picarelli, L., Urciuoli, G., & Russo, C. 2004. Effect of
Corominas, J., Moya, J., Ledesma, A., Rius, J., Gili, J.A. groundwater regime on the behaviour of clayey slopes.
& Lloret, A.1999. Monitoring of the Vallcebre landslide, Canadian Geotechnical Journal, 41: 467–484.
Eastern Pyrenees, Spain. Proceedings Intern. Symp. on Van Asch, Th.J.W., Van Beek, L.P.H. & Bogaard, T.A.
Slope Stability Engineering: IS-Shikoku’99. Matsuyama. 2007. Problems in predicting the mobility of slow-moving
Japan, 2: 1239–1244. landslides. Engineering Geology 91: 46–55.

242
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

The systematic landslide investigation programme in Hong Kong

K.K.S. Ho & T.M.F. Lau

Geotechnical Engineering Office, Civil Engineering and Development Department, Hong Kong SAR, China

ABSTRACT: The Geotechnical Engineering Office (GEO) has been collecting data and conducting annual
reviews of rainfall and landslides since the 1980s. During this time, significant landslide incidents have been
selected for detailed study for the purposes of advancing the understanding of landslides. Under the GEO’s
systematic landslide investigation programme which was implemented in 1997, all reported landslides are
examined and significant landslide cases selected for study to document the failure, establish the probable
causes and identify the lessons to be learnt and the necessary follow-up actions. This paper provides an overview
of the systematic landslide investigation programme in Hong Kong.

1 THE GEOTECHNICAL ENGINEERING Kong (e.g. Hencher et al. 1984; Wong & Ho, 1995;
OFFICE Wong et al. 1998a, b).
As a result of the Kwun Lung Lau landslide on 23
Hong Kong is vulnerable to landslides due to its hilly July 1994, a new systematic landslide investigation
terrain with dense urban development, the presence (LI) programme was implemented by the GEO. This
of a large number of substandard man-made slopes fatal landslide occurred on a 100-year old masonry
mostly formed before the 1970s without adequate wall located within a public housing estate, resulting
geotechnical input and control, deep weathering pro- in 5 fatalities and 3 serious injuries, and temporary
files and high seasonal rainfall. In the aftermath of evacuation of more than 3,900 residents (Figure 1).
several serious landslides with multiple fatalities, the The GEO carried out a comprehensive investiga-
Geotechnical Control Office (renamed Geotechnical tion into the causes and mechanism of the landslide
Engineering Office (GEO) in 1991) was established (GEO, 1994) and an international geotechnical expert,
by the Hong Kong Government in 1977 to regu- Professor N R Morgenstern, was engaged by the Gov-
late the planning, investigation, design, construction, ernment to conduct an independent review of the
monitoring and maintenance of slopes in Hong Kong. technical investigation. The investigation established
Much of the enhanced slope engineering practice in that thin masonry walls are liable to fail in a brittle
recent years has originated from an improved under- manner without appreciable prior warning. The land-
standing of landslides in Hong Kong. In particular, slide also highlighted the adverse effects of leakage
the systematic landslide investigation programme of from buried water-carrying services on slope stability.
the GEO, which was implemented in 1997, has played
a key role in advancing the state of knowledge on slope
performance and facilitated a better understanding of
the causes and mechanisms of slope failures (Wong &
Ho, 2000a).

2 THE SYSTEMATIC LANDSLIDE

INVESTIGATION INITIATIVE

Between the 1980s and the early 1990s, significant

landslides were selected for detailed studies by GEO’s
in-house professionals as research and development
projects to enhance the understanding of causes and
mechanisms of landslides. The technical findings
from these selected landslide studies provided insights
for improvement to slope engineering practice in Hong Figure 1. 1994 Kwun Lung Lau landslide.

243
These findings led to the issue of guidelines to ratio- reliability of landslide preventive or slope remedial
nalize the assessment of the stability of masonry walls works;
and a Code of Practice on Inspection and Maintenance c. provision of data for reviewing the performance of
of Water-carrying Services Affecting Slopes (ETWB, the Government’s slope safety system and identify-
2006). ing areas for improvement;
In his independent review (Morgenstern, 1994), d. provision of evidence in forensic studies of serious
Professor Morgenstern concluded that ‘‘Practice in landslides that may involve coroner’s inquest, legal
Hong Kong with respect to evaluation of slope sta- action or financial dispute.
bility is excessively influenced in a restricted manner
by the slope catalogue and is not sufficiently respon-
sive to indications of potential problems on a project
or development scale’’. One of his recommendations 4 IMPLEMENTATION OF THE SYSTEMATIC
was for Government to introduce a more integrated LANDSLIDE INVESTIGATION
approach into the slope stability assessment process PROGRAMME
through review of landslides. In response to this, a
systematic LI programme was launched by the GEO In undertaking the investigation of significant land-
in 1997. slides, it is important to attend to the sites as soon as
Following a 3-year trial implementation to develop practicable in order to collect crucial field evidence
a new LI methodology for long-term use, the syste- that could otherwise be destroyed or removed as part
matic LI work has been integrated with the Landslip of the debris clearance operation or emergency repair
Preventive Measures (LPM) Programme since 2000. works. Since a large number of landslides may occur
The average annual cost of the systematic LI work is within a short period of time during severe rainstorms
about HK$25 million (about US$3 million). (e.g. over 250 landslides were reported to the GEO dur-
ing and immediately following the 19–22 August 2005
rainstorm), an adequate supply of standby resources
is essential for the prompt mobilization of a sufficient
3 OBJECTIVES OF SYSTEMATIC LANDSLIDE number of investigation teams. The GEO has been out-
INVESTIGATIONS sourcing the LI work to consultants under a standby
arrangement, which has worked well in meeting the
The main goals of the LI programme are illustrated in operational needs. The advantage of engaging con-
Figure 2 and described as follows: sultants to review the performance of Government’s
slope safety system through study of landslides is the
a. identification of slopes in need of early attention
impartiality of an independent party. This is especially
before the situation deteriorates to result in a serious
important for forensic investigations of fatal landslides
problem;
from a public accountability point of view. Also, over-
b. improvement in knowledge on the causes and
seas landslide experts can be mobilized as members of
mechanisms of landslides so as to formulate new
the landslide investigation teams on a need basis for
ideas for reducing landslide risk and enhancing the
serious landslides.

5 METHODOLOGY OF SYSTEMATIC
LANDSLIDE INVESTIGATIONS

Under the LI programme, all reported landslide inci-

dents are examined to collate data for analysis. The
landslides are screened by a panel of experienced
geotechnical professionals to identify cases that war-
rant follow-up inspections and detailed investigations.
On average, about 300 landslides are reported to the
GEO every year. About 20% of the cases would be
selected for inspection by the LI consultants (the vast
majority of the reported cases would be inspected by
the GEO under the emergency system in providing
advice to government departments). Typically, about
Figure 2. Main goals of the landslide investigation 10% of the landslides are found to deserve follow-up
programme. studies.

244
 The following are some of the relevant considera- 6 KEY FINDINGS FROM SYSTEMATIC
tions in screening the landslide incidents for inspec- LANDSLIDE INVESTIGATIONS
tions and follow-up studies:
6.1 Performance of engineered slopes
– large-scale failures; There was a perception in the past that the slope safety
– failures with serious consequences, e.g. casualties, problem in Hong Kong was dominated by failure of
major evacuation and significant social disruption; old substandard slopes formed before 1977 and that
– failures with technical interest, e.g. sites with the stability of engineered slopes built after 1977 to
special geological or hydrogeological features; a high safety standard should be of little concern. An
– failures involving slopes which were previously important development in slope engineering practice
designed and checked to the required safety in Hong Kong over the more recent years has been
standards; the realization of the fact that even engineered slopes
– failures of special engineered slopes, e.g. soil-nailed have quite a high failure rate and that there is a need
slopes; to further improve the practice in order to reduce the
– slopes with major signs of distress; rate of failure.
– slopes with landslide clustering or a history of From the systematic review of the landslide data
repeated failures. between 1997 and 2006, the annual average failure
rate for engineered slopes is about 0.015% for ‘major’
The following types of landslide studies are carried landslides (defined as landslide with a failure volume
out under the LI programme: ≥50 m3 ), and about 0.068% for minor landslides (viz.
<50 m3 ). The overall failure rate of engineered slopes
a. Landslide Examination—all the available informa- in terms of major failures is lower than that for the
tion on landslide incidents are examined shortly pre-1977 slopes (i.e. slopes formed before the estab-
after they are reported to collate data for analysis lishment of the GEO) by a factor of about 2 to 3
and the identify cases which deserve further studies. (Figure 3).
b. Landslide Review—these cover salient aspects of
selected landslide and focus on the most important
elements of the incident. This type of study is par- 6.2 Causes of failures of engineered slopes
ticularly relevant where the incident in itself does
not warrant a detailed landslide investigation, and Wong & Ho (2000a) have summarised the common
will enhance cost-effectiveness and ensure more mechanisms of failures of man-made slopes with fast-
effective use of resources. moving debris (which include fill slopes, soil cuts and
c. Landslide Study—these comprise in-depth studies rock cuts) and some technical observations on notable
of selected landslides examination of the history of generic factors contributing to landslides in Hong
the failed slope and identification of the causes and Kong, based on the findings of landslide studies. The
mechanisms of failure. key contributing factors to failure on engineered soil
d. Forensic investigation—these comprise detailed cuts have been discussed by Wong (2001). These fac-
investigations of fatal or serious landslides to the tors comprise adverse groundwater, weak geological
highest possible rigour of proof in order to prepare materials and inadequate slope maintenance.
a report that can be presented as evidence in legal A total of 56 landslides that occurred on engineered
proceedings. soil cut slopes between 1997 and 2006 have been
investigated in detail to diagnose the key contributory
In order to retain in-house expertise in landslide
investigations, a small number of landslides continue
to be studied by the GEO. In addition to the stud-
ies of individual landslides, a diagnostic review of
all the landslide data and findings from landslide
studies is carried out by the GEO every year to con-
solidate experience and make recommendations to
enhance slope engineering practice and landslide risk
management. Integrated thematic studies (e.g. review
of slope surface drainage with reference to landslide
studies, review of landslides at active construction
sites, review of landslides involving slopes affected
by water-carrying services, review of soil-nailed slope
failures, etc) are also conducted. Figure 3. Landslide data in Hong Kong (1997–2006).

245
causes of the failures. The findings are summarized in
Table 1 below.
About 80% of the landslides on engineered slopes
from 1997 to 2006 were minor failures. The main
problems with respect to minor failures of engineered
slopes are as follows:
– uncontrolled surface runoff,
– inadequate slope maintenance,
– poor detailing of slope surface cover and surface
drainage provisions (Figure 4), and
– local weaknesses in the ground mass, or a combi-
nation of the above factors.
With regard to major failures of engineered slopes,
the main contributing factors are related to:
a. inadequate consideration of ‘stability-critical’ fea-
tures in the geological models, such as relict dis-
continuities, sheeting joints, kaolin-rich seams in
the weathered profiles, and kaolin and manganese
oxide infill with discontinuities (Figure 5),

Table 1. Key contributory factors of engineered soil cut Figure 5. Selected landslides involving relict-jointed
failures. saprolite.

Minor Major b. adoption of inadequate hydrogeological models in

All failures failures slope design leading to under-prediction of ground-
failures (<50 m3 ) (≥50 m3 ) water pressures,
Contributory factor (56 nos.) (40 nos.) (16 nos.) c. progressive slope deterioration and displacement as
evident by signs of distress.
Adverse groundwater 30% 38% 75%
condition (17 nos.) (15 nos.) (12 nos.) The above diagnosis emphasizes the need to fur-
Weak geological 46% 40% 67% ther improve the design practice and detailing so as
materials (26 nos.) (16 nos.) (10 nos.) to further enhance the reliability of engineered slopes.
Inadequate slope 45% 55% 19%
Key areas that warrant attention are highlighted, which
maintenance (25 nos.) (22 nos.) (3 nos.)
Inadequate surface 5% 7.5% include inadequate consideration of local weaknesses
drainage provisions (3 nos.) (3 nos.) − in the groundmass, inadequate engineering geological
Uncontrolled, 2% 2.5% input during investigation, design and construction
concentrated (1 nos.) (1 nos.) − leading to the adoption of inadequate geological and
surface water flow hydrogeological models, insufficient attention to his-
tory of instability, inadequate consideration of overall
site setting in an integrated perspective, uncontrolled
surface runoff, poor detailing in slope drainage provi-
sions and inadequate slope maintenance.

7 BENEFITS OF THE SYSTEMATIC LI WORK

7.1 Landslide data

The comprehensive and good quality landslide data
collected from the systematic LI work have provided a
valuable source of information to facilitate the GEO to
make major technical advancements in the following
areas:
a. assessment of landslide debris mobility of man-
made slope and natural terrain failures (Wong
Figure 4. Surface runoff overflowing off road. et al. 1997; Wong, 2001)

246
b. the application of quantitative risk assessment • slopes with a history of failure despite having been
(QRA) to quantify the landslide risk (Wong & Ko, assessed or designed to the required geotechnical
2006), and standards,
c. continual refinement of the rainfall-landslide fre- • planar geological features (such as joints, faults,
quency correlation model for landslide risk man- weak seams, bedding, foliation, planar soil-rock
agement actions (e.g. criteria for issue of landslip interface), especially where they are dipping out
warnings) (Yu, 2004). of the slope, laterally persistent, show evidence of
previous movement, associated with weak materials
All of the above have contributed to the enhance- such as kaolin, and affect groundwater flow,
ment of the slope safety system in Hong Kong. • evidence of high groundwater, or seepage at high
levels, associated with drainage valleys, subsur-
7.2 Enhancement of technical knowledge face drainage concentrations (e.g. depression in
on landslides weathering front), dykes or persistent sub-vertical
The major advances made as a result of landslide discontinuities,
studies include: • complex groundwater conditions with a significant
response or delayed response to rainstorms,
a. improved understanding of the progressive nature • large cuttings in a deep weathering profile.
of some of the slope failures,
b. consideration of failure mechanisms and debris
7.3 Improvements to the slope safety system
movement mechanisms in the assessment of mobil-
and engineering practice
ity of landslide debris (Figure 6),
c. importance of subsurface water (e.g. perched Based on lessons learnt from landslide studies,
water) as well as surface runoff in triggering the following improvement measures have been
landslides, developed:
d. the significance of the range of adverse geological
features and the need to account for the potential for a. improved detailing for subsurface drainage provi-
local as well as large-scale failures in slope design, sions for recompacted soil fill slopes,
e. importance of robustness in slope design to com- b. improved detailing for surface drainage provisions
bat uncertainties in geological and groundwater in slope designs,
conditions. c. improved detailing for soil-nailed slopes,
d. technical guidelines for the enhancement of rock
The findings from landslide investigations high- slope engineering practice, and
lighted the following points as indicators of e. technical guidance to enhance the reliability and
potentially difficult sites that may have complex robustness of engineered soil cut slopes.
geological/hydrogeological conditions (Wong & Ho,
2000b):
8 OUTPUT OF THE SYSTEMATIC LANDSLIDE
• sites with relict massive failures, INVESTIGATION PROGRAMME
• evidence of progressive slope movement and dete-
rioration, Examples of work completed under the LI consultan-
cies to date include about 700 landslide inspections
(out of more than 3, 000 records examined), about 200
landslide studies, including six forensic investigations.
The findings of all landslides studies are pub-
lished in a series of reports, which are distributed
widely to the local profession. The key findings and
lessons learnt from landslide studies are also made
available on the Government’s slope safety website
http://hkss.cedd.gov.hk.

9 CONCLUSIONS

Based on the lessons learnt from landslide investi-

gations, improvements have been made to the slope
Figure 6. Data on debris mobility for different mechanisms engineering practice to mitigate small and large-scale
and scale of landslides in Hong Kong. failures of engineered slopes (Ho et al. 2003). The

247
systematic LI programme has resulted in improved Wong, H.N. 2001. Invited Paper—Recent Advances in slope
reliability of engineered slopes, improved slope detail- engineering in Hong Kong. Geotechnical Engineering,
ing and enhanced geotechnical and engineering geolo- Ho & Li (Eds), 2001, pp 641–659.
gical input for slopes. It has served as an important Wong, H.N. & Ho, K.K.S. 1995. General Report on Landslips
asset management tool and will continue to form on 5 November 1993 at Man-made Features in Lantau.
Geotechnical Engineering Office, Hong Kong, 78 p. &
an integral part of the Government’s long-term slope 1 drg, (GEO Report No. 44).
stabilisation programme. Wong, H.N. & Ho, K.K.S. 1996. Travel distance of land-
slide debris. Proceedings of the Seventh International
Symposium on Landslides, Trondheim, Norway, vol. 1,
ACKNOWLEDGEMENTS pp 417–422.
Wong, H.N. & Ho, K.K.S. 2000a. Learning from slope fail-
This paper is published with the permission of the ures in Hong Kong. Proceedings of the 8th International
Head of the Geotechnical Engineering Office and the Symposium on Landslides, Cardiff, Bromhead et al (Eds.),
Director of Civil Engineering and Development, Gov- Thomas Telford.
Wong, H.N. & Ho, K.K.S. 2000b. Review of 1997 and
ernment of the Hong Kong Special Administrative 1998 Landslides. Geotechnical Engineering Office, Hong
Region. Kong, 53 p (GEO Report No. 107).
Wong, H.N. & Ko, F.W.Y. 2006. Landslide Risk Assess-
ment—Application and Practice. Geotechnical Engineer-
REFERENCES ing Office, Hong Kong, 277 p (GEO Report No. 195).
Wong, H.N., Ho, K.K.S. & Chan, Y.C. 1997. Assessment
Environmental, Transport and Works Bureau. 2006. Code of consequence of landslides. Proceedings of the Inter-
of Practice on Monitoring and Maintenance of Water- national Workshop on Landslides Risk Assessment, edited
carrying Services Affecting Slopes. Environmental, Trans- by D.M. Cruden & R. Fell, Honolulu, February 1997,
port and Works Bureau. The Government of the Hong pp 111–149.
Kong Special Administrative Region. Second Edition. Wong, H.N., Ho, K.K.S., Pun, W.K. & Pang, P.L.R. 1998a.
November 2006, 93 p. Observations from some landslide studies in Hong Kong.
Geotechnical Engineering Office. 1994. Report on the Kwun Proceedings of the HKIE Geotechnical Division Semi-
Lung Lau landslide of 23 July 1994. Vol. 2—Findings nar on Slope Engineering in Hong Kong, 1998, edited by
of the Landslide Investigation. Geotechnical Engineering Li, K.S., Kay, J.N. & Ho, K.K.S., Hong Kong Institution
Office, Hong Kong, 379 p. of Engineers, pp 277–286.
Hencher, S.R., Massey, J.B. & Brand, E.W. 1984. Application Wong, H.N., Lam, K.C. & Ho, K.K.S. 1998b. Diagnostic
of back analysis to some Hong Kong landslides. Proceed- Report on the November 1993 Natural Terrain Landslides
ings of the Fourth International Symposium on Landslides, on Lantau Island. Geotechnical Engineering Office, Hong
Toronto, vol. 1, pp 631–638. Kong, 98 p. plus 1 drg. (GEO Report No. 69).
Ho, K.K.S., Sun, H.W. & Hui, T.H.H. 2003. Enhanc- Yu, Y.F. 2004. Correlations Between Rainfall, Landslide Fre-
ing the Reliability and Robustness of Engineered Slopes. quency and Slope Information for Registered Man-made
Geotechnical Engineering Office, Civil Engineering and Slopes. Geotechnical Engineering Office, Hong Kong,
Development Department, HKSAR Government, 63 p. 109 p. (GEO Report No. 144).
(GEO Report No. 139).
Morgenstern, N.R. 1994. Report on the Kwun Lung Lau land-
slide of 23 July 1994. Vol. 1—Causes of the Landslide and
Adequacy of Slope Safety Practice in Hong Kong. Report
prepared for the Hong Kong Government, 21 p.

248
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

General digital camera-based experiments for large-scale landslide

physical model measurement

X.W. Hu
Engineering Faculty, China University of Geosciences, Wuhan, China
School of Civil Engineering and Architecture, Wuhan University of Technology, Wuhan, China

H.M. Tang
Engineering Faculty, China University of Geosciences, Wuhan, China

J.S. Li
School of Remote Sensing and Information Engineering, Wuhan University, Wuhan, China

ABSTRACT: Physical model experiment is one of the most effective methods for studying deformation and
failure mechanisms of landslides. The traditional displacement sensors, which must be in contact with or be
buried in the landslide model, can only monitor limited points and limited displacement range, while digital
camera close-range photogrammetry can obtain 3D surface information by means of no-contact. In this paper,
on the basis of the Direct Linear Transformation theory, a large number of general digital camera close range
photogrammetric experiments on a 2D large-scale landslide model were conducted. The experiment results show
that (1) for the same camera with about 2 to 4 Mega-Pixels resolution, the measure accuracy hardly increase
with resolution of images; (2) the different resolutions of images also hardly affect measure accuracy when
the pixel size of images from different type cameras is between about 2 to 4 mega; (3) the precision is high
when convergent angle of two cameras is beyond 10 degree, which dose not change clearly with increasing of
the angle, but it is low when the angle is below 10 degree; (4) the increasing of station numbers and image
number per station does not improve the precision, but much more time and money to be spent; and (5) the
best two-dimensional measure accuracy ranges between 0.56 mm and 0.77 mm when resolutions of cameras are
between about 2 and 4 Mega-Pixels, but the accuracy is low when the resolutions are below 1.38 Mega-Pixels.

1 INTRODUCTION general digital cameras or images taken from at least

two different positions of one general digital camera
In the last decades, landslides in China are increasing at close range, which has several advantages over the
with the rapid development of engineering practices. conventional surveying methods. The first advantage
Physical model test is one of the most effective meth- is the non-contact measuring, and the tested land-
ods for studying deformations and failure mechanisms slide model is not affected by the measuring device
of landslides. Presently, more and more researchers in and the model do not have to be accessible. Sec-
China are using it to study landslides so as to well ondly, all surface information of the 2D physical model
understand and control landslides. Deformation and can be achieved almost immediately and conserved
failure process of a landslide model must be monitored, enduringly along with digital camera images. Data
which are commonly measured by conventional sen- processing can be implemented at any time there-
sors such as strain gauge and dial indicator (Hu, 2004). after. Thirdly, the measuring range can be adjusted
These sensors or some fittings must be positioned in to the measuring task. Finally, photogrammetry can
the model and can only measure the deformation or provide several kinds of products such as images, 2D
displacement of one direction of the measuring point, displacement and velocity, and displacement vector.
and likely influence model test results. Li (2003) investigated close-range photogrammet-
General digital camera close-range photogramme- ric network with variant interior elements of general
try can be used to get the surface information of a 2D digital camera. But his studies did not combine with a
landslide model (Hu, 2006). This photogrammetric large-scale physical model. On the condition of large-
technique measures the physical model by using two scale landslide physical model experiment, it is very

249
 model frame

landslide model

test table

Figure 1. 2-D landslide physical model. Figure 2. Layout of photogrammetry control and check
points with white number mark.

important to use general camera to achieve deforma-

tion characters of the landslide model. In this paper,
the goal of the studies is to investigate the opti-
mal parameters of general digital camera close-range
photogrammetry monitoring a large-scale landslide
physical model and verify the accuracy potential of
this technique.

2 LANDSLIDE PHYSICAL MODEL

A 2D landslide physical model (Fig.1) based on

the prototype of Zhaoshuling landslide in the Three Figure 3. Spatial reference frame and target points for
Gorges Reservoir in China, is about 118.75 cm in general digital camera close-range photogrammetry.
height, 300 cm in length and 20 cm thick. The model
was laid on a test table which could be raised. The
model frame was fixed on the test table.
The landslide model was made of mixtures. The SET-3B, were located on the model frame and the
sliding mass was comprised of barite, fine sand, man- sliding bedding side, respectively. Some of the high
ganese carbonate, gypsum, water and engine oil. The precise control and check points were made into
sliding bedding was comprised of barite, fine sand, the practical control points locating on the model
manganese carbonate and engine oil. Polyethylene frame and the model side to set up a 3D reference
membranes and polytetrafluoroethylene membranes frame, and the others were made into both check
were introduced as the sliding surface. points and target points so as to verify the precision
and the reliability of the results from the gen-
eral digital camera close-range photogrammetry.
3 CLOSE-RANGE PHOTOGRAMMETRIC Figure 3 shows that the coordinate plane XY and the
PHILOSOPHY coordinate Z are in parallel with and perpendicular
to the landslide model side, respectively.
General digital camera close-range photogrammetric 2. Convergent photography was taken to gain a series
method is different from conventional displacement of images with general digital cameras.
measure methods. Procedures of photogrammetry are 3. JPEG format of images is transformed into BMP
as follows: format, then target points in the images are iden-
tified and their 2-D image coordinates (x, y) are
1. Measuring points, pasted on the model frame and obtained.
the landslide model side (Fig. 2), were signalized 4. Based on the Direct Linear Transformation (DLT )
with white papers consisting of a black annular tar- analytical method, object spatial coordinates
get, white background and a small reticle target for (X, Y, Z) of the measuring points are calculated.
total station instrument reference measurements.
Enough control points and check points shown in The object spatial coordinates (X , Y , Z) of a target
Fig. 2, of which 3D coordinates (x, y, z) were sur- point can be calculated by the fundamental expression
veyed with a SOKKI∧ SET-2C and a SOKKI∧ (1) and (2) (Feng, 2002) below:

250
 Table 1. Photogrammetry with two stations of different convergent angles by a Kodak DC 4800 Zoom.

Mean errors of measuring points

Serial
(mm)
of ϕ ω k x0 y0 f XS YS ZS
images (◦ ) (◦ ) (◦ ) pixel pixel mm mm mm mm Mx My Mx,y

1 −18.81 −12.45 −4.98 905.04 −570.02 2837.38 2539.12 704.84 4244.26

28.55 9.43 1.33 928.16 −560.53 1712.65 −238.77 −575.44 2335.74 0.36 0.47 0.57
2 −19.27 0.15 −2.32 908.40 −563.87 2229.76 2305.42 −108.24 3307.48
3.92 3.96 −7.49 912.93 −562.06 1996.54 947.20 −483.26 3467.82 0.40 0.41 0.57
3 −19.24 6.85 −9.48 902.62 −564.98 2116.25 2422.98 −628.38 3336.21
−7.48 5.22 −10.32 910.64 −563.64 2117.87 1711.27 −521.84 3530.37 0.41 0.39 0.57
4 −18.80 −12.45 −4.98 905.04 −570.02 2837.38 2539.12 704.84 4244.26
−0.57 −13.25 −2.34 912.89 −568.03 2650.95 1245.91 722.95 4103.93 0.34 0.51 0.62
5 −0.57 −13.25 −2.34 912.89 −568.03 2650.95 1245.91 722.95 4103.93
3.92 3.96 −7.49 912.93 −562.06 1996.54 947.20 −483.26 3467.82 0.40 0.49 0.64
6 28.55 9.43 1.33 928.16 −560.53 1712.65 −238.77 −575.44 2335.74
−29.80 9.98 2.32 905.41 −563.31 1966.78 2677.64 −634.09 2749.34 0.42 0.49 0.65

251
7 −19.27 0.15 −2.32 908.40 −563.87 2229.76 2305.42 −108.24 3307.48
−0.57 −13.25 −2.34 912.89 −568.03 2650.95 1245.91 722.95 4103.93 0.41 0.51 0.65
8 −18.80 −12.45 −4.98 905.04 −570.02 2837.38 2539.12 704.84 4244.26
12.16 10.08 −0.03 921.16 −561.75 1509.74 532.41 −685.84 2568.83 0.43 0.52 0.68
9 0.81 10.76 0.25 914.63 −559.73 1470.66 1092.53 −699.20 2624.98
−0.57 −13.25 −2.34 912.89 −568.03 2650.95 1245.91 722.95 4103.93 0.44 0.53 0.68
10 28.55 9.43 1.33 928.16 −560.53 1712.65 −238.77 −575.44 2335.74
0.81 10.76 0.25 914.63 −559.73 1470.66 1092.53 −699.20 2624.98 0.52 0.48 0.71
11 12.16 10.08 −0.03 921.16 −561.75 1509.74 532.41 −685.84 2568.83
−29.80 9.98 2.32 905.41 −563.31 1966.78 2677.64 −634.09 2749.34 0.47 0.55 0.73
12 −19.27 0.15 −2.32 908.40 −563.87 2229.76 2305.42 −108.24 3307.48
0.81 10.76 0.25 914.63 −559.73 1470.66 1092.53 −699.20 2624.98 0.59 0.50 0.77
13 −19.27 0.15 −2.32 908.40 −563.87 2229.76 2305.42 −108.24 3307.48
−29.80 9.98 2.32 905.41 −563.31 1966.78 2677.64 −634.09 2749.34 1.27 0.63 1.41
14 3.92 3.96 −7.49 912.93 −562.06 1996.54 947.20 −483.26 3467.82
0.81 10.76 0.25 914.63 −559.73 1470.66 1092.53 −699.20 2624.98 1.59 0.51 1.67
 l1 X + l2 Y + l3 Z + l4
x+ =0 (1)
l9 X + l10 Y + l11 Z + 1 5

l 5 X + l6 Y + l7 Z + l8 4 Kodak DC4800 Zoom Digital Camera

y+ =0 (2) mean errors:
l9 X + l10 Y + l11 Z + 1 coordinate X

accuracy (mm)
coordinate Y
3
coordinate Z
where li (i = 1, 2, · · · , 11) are factors. During calcu- location of plain XY
lating, optical aberrance should be taken into account,
2
and the image coordinates (x, y) must be corrected
(Li, 2003).
1

4 CLOSE-RANGE PHOTOGRAMMETRIC 0
3/7 10/9 12/1 4/17 18/1 18/13 20/10 23/3 1/24 27/1 31/22 41/1 47/21 58/0
EXPERIMENTS AND RESULTS
/ ( )
4.1 Experiment preparation
Figure 4. Relationship between measuring accuracy and the
The model table and the landslide physical model were angle element difference ϕ(ω)of pairs of images from a
static. The general digital cameras used to test involved Kodak DC4800 Zoom camera.
a Canon Powershot45 Zoom camera with the images of
3.87, 1.92 and 0.78 M-Pixels, a Kodak DC290 Zoom
0.35 mm∼0.46 mm and 0.44 mm∼0.47 mm, respec-
camera with the resolutions of 3.36, 2.15 and 1.38
tively, and that of plane XY is 0.56 mm∼0.66 mm.
Mega-Pixels and a Kodak DC4800 Zoom camera with
While the angle element differences ϕ and ω both
2.16 Mega-Pixels resolution. At test, every camera was
are below or equal to 4 degrees, the accuracy value of
fixed on a tripod with the distance about 2 m∼4 m to
coordinate X is over 0.90 mm and that of plane XY is
the model.
over 1.00 mm.

4.2 Influence of the convergent angle of cameras

on the measuring accuracy 4.3 Influence of the image resolution from same
digital cameras on the measuring accuracy
A Kodak DC4800 Zoom camera with the image reso-
lution of 1800 × 1200 (2.16 Mpixels) and a Kodak A Canon Powershot45 Zoom camera and a Kodak
DC290 Zoom camera with the image resolution of DC290 Zoom camera were respectively used to make
2240 × 1500 (3.36 Mpixels) were adopted. Two pho- three team tests with two photographic stations, and
tographic stations with one image every station were one image every station was taken with convergent
applied to obtain 19 pairs of images with different photogrammetry.
convergent angles. Figure 6 shows that for two images with same pixel
The results shown in Table 1 and Figure 4 from of 1.92∼4.00 Mega, the accuracy value of coordinate
Kodak DC4800 Zoom camera indicate that when X is about 0.3 mm∼0.4 mm, that of coordinate Y is
one of the angle element difference ϕ and ω of about 0.5 mm∼0.6 mm and that of coordinate plane XY
every pair of images is excess of 10 degrees, the is about 0.65 mm. But, when two image resolutions are
accuracy values of coordinates X and Y both reach 0.78∼1.38 Mega-Pixels, the accuracy values of coor-
about 0.34 mm∼0.59 mm and that of plane XY is dinates X and Y is about 0.54 mm∼0.65 mm and about
0.57 mm∼0.77 mm. But the measuring accuracy of 0.82 mm∼2.04 mm respectively, and that of coordi-
coordinates X and Y is hardly improved with the nate plane XY is about 0.98 mm∼2.14 mm. The results
increase of the angle element difference ϕ and ω. illuminate that when the image pixel of general digital
While the angle element differences ϕ and ω are camera is about 1.92∼4.00 mega, the photogrammet-
below or equal to 10 degrees, the accuracy value of ric measuring accuracy value of coordinate plane XY
coordinate X is over 1.00 mm and that of plane XY is is about 0.65 mm and hardly influenced by the change
over 1.40 mm. of image resolution, but when the image pixel is under
The results shown in Table 2 and Figure 5 from about 1.5 mega, the measuring accuracy is low.
the combination of a Kodak DC290 Zoom camera
and a Kodak DC4800 Zoom camera indicate that
4.4 Influence of the image resolutions from different
when the angle element difference ϕ of a pair of
digital cameras on the measuring accuracy
images is between 13 and 28 degrees and the angle
element difference ω is between 3 and 5 degrees, The combination of a Canon Powershot45 Zoom cam-
the accuracy values of coordinates X and Y are about era and a Kodak DC290 Zoom camera was used to

252
 Mean errors of measuring points conduct five team tests with two photographic stations,
and one image every station was taken with convergent
Photogrammetry with two stations of different convergent angles by the combination of a Kodak DC 290 Zoom camera and a Kodak DC 4800 Zoom camera.

photogrammetry.
Mx,y

0.56

0.60

0.66

1.08

1.77
Figure 7 shows that when the place and the pose of
two cameras are constant and the two image pixels are
1.38∼3.87 mega, the accuracy values of coordinate
0.44

0.44

0.47

0.50

0.59
X , Y and plane XY are about 0.54 mm∼0.64 mm,
My
(mm)

0.35

0.42

0.46

0.93

1.66
Mx

6 Kodak DC290 Zoom Digital Camera

and Kodak DC4800 Zoom Digital Camera
5 mean errors:
coordinate X
3224.58
3311.47
3446.11
3287.34
3435.54
3682.97
3526.49
3745.83
3514.34
3435.12
coordinate Y
4 coordinate Z

accuracy (mm)
location of plain XY
mm
ZS

2
−424.64
−630.05
−462.24
−629.76
−462.08
−547.74
−503.40
−497.55
−256.38
−247.45

1
mm
YS

0
2/0 4/1 13/3 24/5 28/5
/ ( )
684.17
2431.52
932.45
2421.32

1874.13
1215.11
1536.39
1115.88
1368.97
932.8

Figure 5. Relationship between measuring accuracy and the

mm
XS

angle element difference ϕ(ω) of pairs of images from the

combination of a Kodak DC290 Zoom camera and a Kodak
DC4800 Zoom camera.
2455.87
2237.45
2463.44
2232.13
2455.14
2234.04
2454.64
2232.14
2461.77
1906.43
mm

2.4
coordinate X
f

2.2 coordinate Y
coordinate Z
2.0 location of plain (X,Y)
−758.04
−558.13
−755.68
−560.08
−759.71
−559.40
−746.60
−561.94
−755.79
−563.66

1.8
1.6
pixel

accuracy (mm)

1.4 Canon Powershot 45 Zoom camera

y0

1.2
1.0
1139.405
900.915

0.8
1139.90
903.35
1140.75
906.04
1134.00
909.51
1133.04
905.61

0.6
pixel

0.4
x0

0.2
50 100 150 200 250 300 350 400
resolution of images (M pixels)
−9.377
−7.346

−1.036
−7.32
−9.37
−7.35

−10.12
−7.64
−10.38
−1.86

1.8
(◦ )
k

1.6

1.4 coordinate X
coordinate Y
2.06
7.11
2.55
7.17
2.46
5.63
3.27
4.70
2.19
2.63

accuracy (mm)

coordinate Z
(◦ )

1.2
ω

location of plain (X,Y)

1.0
Kodak DC290 Zoom camera
−19.96

−20.32

−8.98

−3.97

−2.51
4.23

4.24

0.32

0.26
8.26

0.8
(◦ )

0.6
ϕ

0.4
100 150 200 250 300 350
Table 2.

resolution of images £¨(M pixels)£©

images
Serial
of

Figure 6. Relationship between measuring accuracy and

image resolution from a camera.

253
 2.2 2.0
2.0 1.8
1.8 1.6
mean errors:
1.6 coordinate X

accuracy (mm)
accuracy (mm)

coordinate Y 1.4
1.4 Kodak DC290 Zoom Digital Camera
coordinate Z
location of plain (XY) 1.2 resolution: 2240*1500
1.2 mean errors:
1.0 coordinate X
1.0 coordinate Y
0.8 coordinate Z
0.8 location of plain XY
0.6
0.6
0.4 0.4

down 2 stations (1) up 2 stations (1) 4 stations (1) 4 stations (2)

138-192 215-192 336-192 215-387 336-387
photographic station number (image number every station)
resolution of images (M pixels)

Figure 9. Relationship between measuring accuracy and the

Figure 7. Relationship between measuring accuracy and
number of station and image.
image resolution from two-station different digital cameras.

Figure 9 shows that when the place and the pose

camera of camera are nearly constant and the image pixels are
3.36 mega, the accuracy values of coordinate X , Y and
XY are about 0.35 mm∼0.40 mm, 0.39 mm, 0.44 mm
and 0.57 mm, respectively, and hardly influenced by
ct
obje photographic station number and image number every
station.

5 DISCUSSION AND CONCLUSION

table
From the above studies on general digital cam-
era close-range photogrammetry based on DLT and
convergent photography for a large-scale landslide
physical model, some conclusions drawn are as fol-
lows.
Figure 8. Sketch of 4 photographic stations. 1. When the resolutions of images from general dig-
ital camera range between 1.92 M-pixels and 4.00
M-pixels, the measure accuracy is high and hardly
influenced by the improving of image resolution.
0.33 mm∼0.41 mm and 0.65 mm∼0.74 mm, respec- However, image resolutions below 1.38 M-pixels
tively, and hardly influenced by the change of two evidently reduce the measuring accuracy. And the
image resolution. accuracy is independent of the type of general
digital camera.
2. Photogrammetric station numbers and image num-
4.5 Influence of station and image number
bers per station hardly influence the measuring
on the measuring accuracy
accuracy, which is different from the studied results
A Kodak DC290 Zoom camera was used to obtain by Dr. Li (2003).
images with the resolution of 2240 × 1500 (336 M 3. When the angle element differences ϕ and ω
pixels) and 1792 × 1200 (225 Mpixels). Four photo- of images are below or equal to 10 degrees, the
graphic stations (Fig. 8), two of which were set up 2-D measure accuracy is low. While one of angle
on ground and a table respectively, were taken with element differences is in excess of 10 degrees,
convergent photogrammetry. the 2-D measuring accuracy is high and its value

254
 reaches about 0.56 mm∼0.77 mm, but it does not Hu, X.W., Tang, H.M. & Li, J.S. 2006. Reliability of landslide
get higher because of the increase of angle element model monitoring based on digital camera. In proceedings
differences. of the Sixth International Conference on Physical Mod-
elling in Geotechnics (6th ICPMG ‘06), Hong Kong, 4–6
August 2006, Vol.1: 193–197.
Li, J.S. 2003. Research of digital close range photogrammet-
ric network with variant interior elements. PhD thesis,
REFERENCES Wuhan University.

Feng, W.H. 2002. Close-range photogrammetry: measura-

tion of shape and movement state of a object. Wuhan:
Wuhan University Publish Company.
Hu, X.W. 2004. Studies on similar material model experiment
for landslide stability of Zhoushuling in the Three Gorges
Reservoir. PhD thesis, China University of Geosciences.

255
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Shear strength of boundaries between soils and rocks in Korea

S.G. Lee, B.S. Kim & S.H. Jung

Department of Civil Engineering, The University of Seoul, Seoul, Korea

ABSTRACT: In Korea, heavy rainfall each summer triggers many landslides with consequent loss of life and
considerable damage to properties. Most of the landslides in mountain regions occur along the boundary between
an overlying soil layer of weathered rock and colluviums and the underlying less weathered rock, regardless of
rock type. This paper presents the results of a series of direct shear tests was performed to investigate the shear
strength of the boundary between soils and rocks as well as the intact soil above the boundary under natural
and saturated moisture conditions. Direct shear tests were conducted for a variety of igneous, sedimentary, and
metamorphic rocks and the results are compared to identify the difference related to various rock types. In
general it was found that the shear strength of boundaries between soil and rock are lower than the shear strength
of the intact soil for all rock types. Furthermore shear strength was generally lower for the saturated condition
compared to the natural moisture condition.

1 INTRODUCTION As the landslides occur frequently at the boundary

between soil and rock in Korea, irrespective of the
Seventy percent of Korea is mountainous. Numerous kind of rock, shear strength of contacts between soil
landslides occur during the rainy season from June to and rocks were studied using a variety of rock types.
September every summer resulting in on average 60
persons killed and hundreds of millions US dollars
of property damage and traffic disruption (Lee et al,
2 SELECTION OF SURVEY AREAS AND
2007). In the mountainous areas soil is typically thin
METHOD OF STUDY
around 1 m thickness, overlying rock. The majority of
landslides occur along this boundary between the thin
One hundred and seventy one landslides were surveyed
soils and rocks (Lee, 1987, 1988 & 1995; Lee et al.,
in various areas listed in Table 1, where numerous
2008).
landslides during rainfall associated with Typhoons
Numerous landslides occurred during the passage
Rusa in 2002 and Maemi in 2003 (Table 1). The gen-
of Typhoons Rusa in 2002 and Maemi in 2003 and,
eral geology of the studied areas are shown in Figure 1
as is usually the case, most of the landslides that were
(Um & Reedman, 1975). It can be seen that the studied
triggered by these storms occurred on the boundary
landslides occurred in terrain underlain by a variety of
between soil and rock (Lee et al. 2008).
igneous, sedimentary and metamorphic rocks.
The purpose of the study reported in this paper
was to investigate the shear strength of the boundaries
between soil and rocks which is an important ele-
Table 1. Areas and numbers of landslides surveyed.
ment for slope stability analysis and the geotechnical
engineering of landslides. Time of
In general, when analyzing slope stability, the shear Surveyed areas Numbers production
strength of intact soil is adopted in calculations on
the assumption that that shear strength is the mini- Igneous rock Gangreung 80 1 Sep 2002
mum. This is despite other workers suggesting that areas Donghae 40 1 Sep 2002
the shear strength may be lower at the soil-rock con- Taebaeg 6 30 Aug 2003
tact surface compared to the intact soil (Patton, 1968). Metamorphic Hamyang 26 1 Sep 2002
rock areas Samcheog 6 30 Aug 2003
Similarly Kanji (1974) experimentally demonstrated
Sedimentary
that the shear strength on the boundary between weath- rock areas Habcheon 6 30 Aug 2003
ered granitic soil and granitic rocks is lower than that Total 171
of the soil itself.

257
 Sheeting joint

Sliding Plane

Figure 3. Shear surface almost parallel to sheeting joint in

igneous rock.

Sliding Plane

Figure 1. Locations and geological characteristics of stud- Figure 4. Shear plane mismatched with the direction of
ied areas. sheeting joint developed in igneous rocks.

Bedding Plane

Sliding Plane

Figure 5. Landslides produced parallel with the bedding

plane of sedimentary rock area.

Bedding Plane

Figure 2. Survey of landslides.

Sliding Plane

3 PROPERTIES OF SHEAR STRENGTH

Figure 6. Landslide discordant with bedding discontinuities
OF THE AREAS IN LANDSLIDE AREAS in sedimentary rock area.
3.1 Landslide Studies
In the study areas, the original slope angles, the length, the study areas mostly consist of gneiss; in these areas
depth, slope, and width of landslides, location of land- the interface between the detached soil horizon and
slide, and strength characteristics of failure surfaces underlying rock is generally not associated with any
were recorded for more than 170 landslides (Figure 1) specific geological structure (Figure 7).
(Lee et al, 2008).
In granitic areas many landslides occurred parallel
3.2 Preparation of rock and soil samples
to unloading (sheeting) joints (Figure 3) but in some
cases the shear surface was mismatched with sheeting About 400 samples were collected for each rock type
joints (Figure 4). In sedimentary rocks, mainly com- using sampling rings to minimise disturbance. Shear
prising sandstone and shale, in some cases the shear tests were then conducted on the intact soil from
plane is parallel to bedding (Figure 5) but in other cases above the slip surface and on samples representing the
there is a mismatch (Figure 6) . Metamorphic rocks in boundary between soil and rocks at different normal

258
 Sliding Plane
(b)

Figure 7. In metamorphic rock areas slip surfaces generally

are not associated with pre-existing geological structures.

(a) (c)

(d) (e)
Figure 8. Collection of rock and soil samples.

stress levels, for natural moisture content (as collected)

and in a soaked condition. To perform the shear tests
on the boundary between soil and rocks, rock sam-
ples were collected as representative as possible of
small scale roughness of the areas where the landslides (f) (g)
occurred (Figure 8).
Figure 9. Details of direct shear strength test on boundary
between soil and rock (a) Components of shear box (b) Rock
3.3 Soil property test sample placed in the lower part of shear box (c) Intact soil
Soil samples collected in each area were wrapped to placed in the upper part (d) Porous plate placed above soil (e)
Application of normal load (f) LVDTs for measuring vertical
prevent changes in moisture content and to minimise and horizontal displacement (g) Shear test underway.
disturbance. The soil collected from each site was
tested by graded and various index tests conducted
including specific gravity, liquid limit, plastic limit,
and water content (Head, 1980). Mean values were
determined for each rock type. in case of igneous rock, 18% in case of sedimentary
rock, and 17% in case of metamorphic rock; the con-
tents were high. The range of specific gravity was
3.4 Shear strength test 2.56∼2.71 and soil classification through sieve anal-
Shear tests were carried out at moisture content as ysis showed well-graded result. As shown in Table 2,
recovered and in a soaked state. For soaked tests sam- as the result of direct shear strength tests, the c and 
ples were packed with filter papers and then immersed values of boundaries between soils and rocks were
in water for 5 days prior to testing. Shearing rate was smaller in all rock types compared to the values
set as 0.2 mm/min. Normal stress was given was varied obtained from soil and the c and  values were lower
over five steps from 30 kPa to 150 kPa and maximum when the tests were performed natural moisture con-
shear displacement was set as 9 mm, i.e. 15% of the tent condition compared to saturated condition. The
whole length of a sample. The tests on the boundary fact that the shear strength of boundaries is lower than
between soils and rocks were essentially the same as that of soil indicates the proof of the landslide collapse
those on intact soils but in these cases a rock specimen on the boundaries.
was placed inside the lower ring. Also it could be identified that, when the internal
friction angle of soil is assumed as 100% and when
the internal friction angle lowering of boundaries is
3.5 Result of shear property tests and shear
expressed in percent, the internal friction angles were
strength tests
lowered up to 71∼94% of natural moisture content
Based on the result of soil property tests (Table 2, and 82∼98% of saturated moisture content in case of
Figures 10, 11 & 12), unit weight of rock was igneous rocks, 74∼96% of natural moisture content
17.06∼18.04 kN/m3 , mean moisture content was 19% and 89∼99% of saturated moisture content in case of

259
 Table 2. Results of physical properties and shear strength tests on residual soils derived from various rock types.

γ w/t LL PI C 
Rock type (kN/m3 ) (%) Gs (%) (%) USCS (kPa) (%)

Igneous rock Breaking down by Soil dry 17.75 19.18 2.625 42.55 30.57 SP 18.5 24.3
sheeting joints wet 18.04 23.51 2.624 42.55 30.57 SP 17.3 20.5
Soil/Rock dry 17.95 21.56 2.606 42.55 30.57 SP 17.6 19.6
boundary wet 18.04 21.77 2.561 42.55 30.57 SP 19.6 18.8
Breaking with no Soil dry 17.26 17.82 2.607 31.88 11.56 SM 28.1 20.9
influence from wet 17.55 20.75 2.607 31.88 11.56 SM 27.1 20.5
sheeting joints Soil/Rock dry 17.46 16.18 2.646 31.88 11.56 SM 26.7 19.1
boundary wet 17.85 19.12 2.705 31.88 11.56 SM 28.1 18.8
Sedimentary rock Landslides produced Soil dry 17.65 18.82 2.654 35.81 11.25 SM 28.6 22.4
parallel with wet 17.65 19.04 2.614 35.81 11.25 SM 34.2 21.1

260
bedding plane Soil/Rock dry 17.65 19.87 2.647 35.81 11.25 SM 27.5 19.6
boundary wet 17.65 18.72 2.597 35.81 11.25 SM 29.7 18.4
Landslides produced Soil dry 17.75 18.94 2.603 45.54 13.24 SW 30.2 23.6
with quite wet 17.75 20.14 2.641 45.54 13.24 SW 27.4 22.1
mismatch with Soil/Rock dry 17.75 18.97 2.647 45.54 13.24 SW 29.7 20.5
bedding plane boundary wet 17.65 19.37 2.640 45.54 13.24 SW 26.4 19.2
Metamorphic rock Landslides produced Soil dry 17.65 18.68 2.598 40.30 19.76 SM 32.2 17.7
regardless joins wet 17.65 19.08 2.597 40.30 19.76 SM 28.8 15.8
or foliation Soil/Rock dry 17.55 14.45 2.592 40.30 19.76 SM 32.7 17.4
are the geological boundary wet 17.65 14.72 2.640 40.30 19.76 SM 34.9 16.2
structure of Soil dry 17.46 18.67 2.664 39.64 21.54 SW 31.6 21.5
metamorphic wet 17.65 18.98 2.564 39.64 21.54 SW 27.8 20.7
rocks Soil/Rock dry 17.06 18.28 2.578 39.64 21.54 SW 28.7 19.5
boundary wet 17.65 19.57 2.642 39.64 21.54 SW 27.4 18.3
Figure 10. Shear strength of Igneous rocks: (a) Landscapes having the collapse matching with sheeting joint; (b) Landscapes
having the collapse quite mismatching with sheeting joint orientation.

Figure 11. Shear strength of sedimentary rocks: (a) Landscapes having the collapse matching with bedding plane;
(b) Landscapes having the collapse quite mismatching with bedding plane.

Figure 12. Shear strength of metamorphic rocks: (a) Landslides in metamorphic rocks in Samcheog area of Gangwon-do;
(b) Landslides in metamorphic rocks in Geochang area of Gyeongnam.

metamorphic rocks, and 77∼94% of natural moisture 2. As the result of direct shear tests, it was found
content and 82∼98% of saturated moisture content in that, when the internal friction angles of soil was
case of sedimentary rocks. assumed as 100% and the internal friction angles
of boundary to the internal friction angles of soil
were expressed in percent, the internal friction
angles were lowered to 72%∼96% in natural mois-
4 CONCLUSION ture content condition and to 82∼94% in saturated
moisture content condition.
1. A survey of 117 landslides in Korea triggered by 3. The shear strength of boundaries was lower in all
Typhoons Rusa in 2002 and Maemi in 2003 showed rock types compared to the shear strength of soil
that most of the landslides involved sliding on the and, in both cases, lower shear strength was found
boundaries of soil and rocks. in saturated moisture condition compared to natural

261
 moisture condition. The fact that the shear strength Lee, S.G. 1987. Weathering and geotechnical characteriza-
of boundaries is lower than that of soil indicates the tion of Korean granites, PhD thesis, Imperial College,
proof of the landslide collapse on the boundaries. University of London.
Lee, S.G., 1988. A study on landslide in Korea, Researches
on geological hazards, Research report of Korea Institue
of Geoscience and Mineral Resources (KIGAM), KR-88-
ACKNOWLEDGEMENTS (B)-7, 145–148.
Lee, S.G., 1995. Natural hazard in Korea. Proc. Of the int.
This research was partially supported by a grant Forum on Natural hazard mapping, Geological Survey of
(NEMA-06-NH-05) from the Natural Hazard Mitiga- Japan Report, 281, 145–148.
tion Research Group, National Emergency Manage- Lee, S.G., Kim, M.S. & Park, D.C., 2008. A study on
ment Agency. the characteristics of landslides related to various rock
types in Korea. Proc. 10th Int. Symp. on Landslides and
Engineered Slopes, Xi’an, China (in press).
REFERENCES Patton, F.D. 1968. The determination of shear strength of
rock masses, Paper presented to the terrametric course
on measurement systems of control of construction and
Head, K.H. 1980. Manual of Soil Laboratory Testing. VI, Soil
mining, Denver: 37.
Classification and Compaction Tests. V2, Permeability,
Um, S.H. & Reedman, A.J., 1975. Geology of Korea. Korea
Shear Strength and Compressibility Tests. Publ, London.
Institue of Geoscience and Mineral Resources (KIGAM),
Pentach Press.
Seoul.
Kanji, M.A. 1974. Unconventional Laboratory Tests for
the Determination of the Shear Strength of Soil-Rock
Contacts. The 3rd Congress of ISRM. Denver 2, pp.
241–247.

262
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Cracks in saturated sand

X.B. Lu & S.Y. Wang

Institute of Mechanics, Chinese Academy of Sciences, Beijing, China

Peng Cui
Institute of Mountain Hazard and Environment, Chinese Academy of Sciences, Chengdu, China

ABSTRACT: The formation mechanism of water film (or crack) in saturated sand is analyzed numerically. It
is shown that there will be no stable ‘‘water film’’ in the saturated sand even if the strength of the skeleton is zero
and no positions are choked. The stable water films initiate and grow if the choking state keeps unchangeable
once the fluid velocities of one position decreases to zero in a liquefied sand column. A simplified method for
evaluating the thickness of water film is presented according to a solidification wave theory. The theoretical
results obtained by the simplified method are compared with the numerical results and the experimental results
of Kokusho.

1 INTRODUCTION to the velocity difference of grains and pore water

(Cheng, C.M. et al. 2000). The x axis is upward.
It is often occurred on the ground slope that sand    
deposit translates to lateral spreading or even landslide 1 ∂Q ∂Q λ u − us
or debris flow not only during, but also after earth- + us = −q
ρs ∂t ∂x T u∗
quakes. If the sand deposit on a slope are composed of
many sublayers, there will be a water film forms once Q Qc (x)
it liquefied (Kokush et al. 1998) which may serves if − ε(x, 0) ≤ ≤ (1)
ρs ρs
as a sliding surface for postliquefaction failure. As a
result, landslide or debris flow may happen on a slope  
1 ∂Q ∂Q
with very gentle slope-angle. Seed (1987) was the first + us ≤ 0 otherwise (2)
to suggest that the existence of water film (crack) in ρs ∂t ∂x
sand bed is the reason of slope failures in earthquakes.
Some researchers (Fiegel, G.L. & Kutter, B.L. 1994, in which Q is the sand mass eroded per unit volume of
Kokusho, T. 1999, Zhang Junfeng 1998) performed the sand/water mixture, ρs is the grain density, u and
some experiments to investigate the formation of crack us are the velocities of pore fluid and sand grains, q
in layered sand or in a sand containing a seam of non- is the volume fraction of sand carried in percolating
plastic silt. Nevertheless, the mechanism of the for- fluid, T and u∗ are physical parameters, λ is a small
mation of cracks or ‘‘water film’’ in a sand layer with dimensionless parameter, ε(x, t) is the porosity, Qc (x)
the porosity distributed continuously is not very clear. is the maximum Q that can be eroded at x.
In the viewpoints above, Firstly, we present a
pseudo-three-phase model describing the moving of
liquefied sand and numerically simulates. Secondly, 3 MODEL OF THE PROBLEM
we present a simplified method to analyze the evolu-
tion of the water film. The mass conservation equations are as follows:
⎧
⎪ ∂(ε − q)ρ ∂(ε − q)ρu
⎪
⎪ + =0
⎪
⎪ ∂t ∂x
2 FORMULATION OF THE PROBLEM ⎪
⎪
⎨
∂qρs ∂qρs u ∂Q ∂Q
It is considered here a horizontal sand layer, which is + = + us (3)
⎪
⎪ ∂t ∂x ∂t ∂x
water saturated and the porosity changes only verti- ⎪
⎪
⎪
⎪
cally. The fine grains may be eroded from the skeleton ⎩ ∂(1 − ε)ρs + ∂(1 − ε)ρs us = − ∂Q − us
⎪
and the eroding relation is assumed to be proportional ∂t ∂x ∂t

263
in which ρ is the density of water. A general equation For Tg/ut >> 1, the inertia terms are negligible,
may be obtained by eq. (3) the last equation of eq. (5) becomes
 2
εu + (1 − ε)us = U (t) (4) 1−ε k0 ρs g(1 − ρ/ρs )
ū = (ε − q)f (q, ε)
ε ut
 2
in which U (t) is the total mass of fluid and grains at a 1−ε
transect. The momentum equations are: = (ε − q)f (q, ε) (10)
ε
⎧   ut = k0 ρs g(1 − ρ/ρs ) (11)
⎪ ∂u ∂u
⎪
⎪ [(ε − q)ρ + qρs ] +u
⎪
⎪ ∂t ∂x
⎪
⎪ The initial conditions are:
⎪
⎪
⎪
⎪ ∂p ε (u − us )
2
⎪
⎪ = −ε − − [(ε − q)ρ + qρs ]g ε(ξ , 0) = ε0 (ξ ), q(ξ , 0) = 0. (12)
⎪
⎪ ∂x k(ε, q)
⎪
⎪  
⎪
⎪
⎪
⎪ ∂u ∂u
⎪
⎨ ×[(ε − q)ρ + qρ s ] + u
∂t ∂x 4 NUMERICAL RESULTS
⎪   (5)
⎪
⎪ ∂u s ∂u s
⎪
⎪ +(1 − ε)ρs + us Numerical simulation is carried out based on eq. (8).
⎪
⎪ ∂t ∂x
⎪
⎪ The parameters adopted in simulation are as follows:
⎪
⎪ ∂p ∂σ t = 9 × 10−4 , ζ = 5 × 10−3 , β = 46∼56, κ =
⎪
⎪ e
⎪ =−
⎪ − − [(ε − q)ρ + qρs ]g 50.0, a = 0.08, ρs = 2400 kg/m3 , ρw =
⎪
⎪ ∂x ∂x
⎪
⎪   1000 kg/m3 , u∗ = 0.04, k0 = 4 × 10−6 m/s, α = 1.
⎪
⎪ ∂Q ∂Q
⎪
⎩ The boundary conditions:
−(1 − ε)ρs g − + us (u − us )
∂t ∂x
1. The initial porosity distribution is ε0 (x) = ε̄0 (1−a
tanh((x − 0.5L)/2) · κ), in which ε̄0 = 0.4,
in which p is the pore pressure. L = 1, 0 ≤ x ≤ 1, L is the length of sand column.
Here k is assumed as following There is an assumption that u keeps zero once it
drops to zero.
k(ε, q) = k0 f (q, ε) = k0 (−αq + βε) (6) 2. The distribution of initial porosity is the same as
that in condition 1, there is no assumption.

in which α, β are parameters and 1 < β << α, which Figure 1 shows that if we assume that once the sand
indicates that changes in q overweighs that of ε. column at some point is jammed, they keep this state
The mass conservation equation (4) yield assuming forever, then the sand above the jammed position will
both u and us are zero at x = 0. be prevented to drop cross the jammed point and so
the porosity becomes smaller and smaller, while the
sand below the point will settle gradually and makes
εu + (1 − ε)us = U (t) = 0 (7)

Taking T as characteristic time. ut the characteristic

velocity and L the characteristic length of the problem. 1.0

We make eq. (1) non-dimensional. Letting

0.8 1 initial time
2 30000 dt
3 1400000 dt
u t x 4 1700000 dt

ū = , τ= , ξ=
d t is the time step length
(8) 0.6
ut T Tut
e

0.4 1
Instituting equ. (1), (2), (4), (7) (8) into eqs. (3), we 2
3
may obtain
0.2 4

∂ε ∂εū ut
+ = ū ∗ −q 0.0

∂τ ∂ξ u (1 − ε)
(9) 0 20 40 60 80 100

∂q ∂qū
x
ut
+ = ū ∗ −q
∂τ ∂ξ u (1 − ε) Figure 1. The evolution of cracks in the condition 1.

264
 0.7 ρw n0 − n1 h
t= (14)
ρ 1 − n1 k
0.6
1 intial time
2 900000 dt
3 1400000 dt in which h is the height of any point in the sand column.
0.5 dt is the time step length
1 Assumption The final settlement of the top surface of the sand
once v=0, there is

0.4
no sediment and erosion layer is
n0 − n1
e

2
0.3 L = H (15)
3 1 − n0
0.2
in which H is the maximum height of sand layer.
0.1 The rate of settlement is
0.0 γ k
ṡa = (16)
0 20 40 h 60 80 100 γw

The settlement at any time is

Figure 2. The evolution of cracks in condition 2.
γ k
sa = t (17)
γw
the crack extends gradually. But if we do not adopt the The settlement velocity ve of the elements above
assumption as in Figure 1, the crack will form first and the water film is determined by the combined perme-
then disappear gradually (Figure 2). The results show ability kes of the middle layer and the upper layer as
that the forming conditions of stable water film are: (Kokusho, 2002)
(1) the porosity of the upper part of the sand column m
must be smaller than that of the lower. (2) The keeping m 

of the jamming state to prevent the free dropping of kes = Li Li ki (18)

the grain or the skin friction. 1 1

The upward seepage flow and the settlement of

5 A SIMPLIFIED EVALUATION METHOD grains have the following Velocities:
nv
A simplified method is presented here for analyze the v = kes ie ; u=− (19)
evolution of cracks fast and practically. Florin and 1−n
Ivanov (1961) pointed out that when the settling par- ie is the average hydraulic gradient, n is the porosity.
ticles reach solid material, such as the unliquefied The deform of the skeleton by the geostatic stress
underlying sand, or the container base in a experi- in the solidification zone after the solidification may
ment, they accumulate to form a solidified zone which be expressed as[5]
increases in thickness with time. A solidification front
therefore moves upward until it reaches the surface, or 1 ρg 2
overlying unliquefied material. Scott et al. (1986) had s2 = z (t) (20)
2 ms
analyzed the development of the solidification.
Assuming that the whole mass reaches its terminal in which ms is the compressible modulus. The total
velocity, k, which is the permeability, instantaneously deformation is:
at the end of liquefaction, Florin have given an expres-
sion for the constant velocity ż, of the solidification s = L + s2 (21)
front:
Instituting eqs. (15) and (20) into eq. (21), consi-
ρ 1 − n1 dering that the initial height is equal to maximum
ż = k (13)
ρw n0 − n1 height of the liquefied zone and the solidified zone
and the water layer above the sand surface:
in which ρ  = ρs − ρw is the buoyant unit weight of
the liquefied sand n0 is the porosity of the liquefied n0 − n2 ρg
sand, n1 is the porosity of the solidified sand. s = (z + s) + zz (22)
1 − n2 ms
From eq. (13), we can obtained the duration of liq-
uefaction and subsequent excess pore pressure decline because h = h0 + L + s, which yields h =
for any point in the sand column. (L + s).

265
 Eq. (16) may be written as

s ρ
=k (23)
t ρw

According to eqs. (22) and (23), the increase

velocity of the solidification thickness:
 
z kρ  n 0 − n1 1 − n1 ρ  g
= + z (24)
t ρw 1 − n1 1 − n0 ms

Then we can obtain the duration of any location that

the solidification front arrives at as follows:
 
ρw n0 − n1 1 1 − n1 ρ  g 2 Figure 3. The comparison of the results computed by
t= z + z (25) simplified method with the experiment of Kokusho.
kρ  1 − n1 2 1 − n0 ms

Side friction may be expressed if it should be ACKNOWLEDGEMENTS

considered
This paper is supported by Chinese Academy of Sci-
σs = μK0 σz (26) ences and The Key Program of the Academy of
Sciences (KZCX2-YW-302-02), Key Laboratory of
The effect of the changes of porosity on the perme- Mountain Hazards and Surface Process.
ability k is considered as a linear relation:

k = k0 [1 − α(n0 − n)] (27) REFERENCES

in which k0 is the initial porosity, α is a parameter, n0 Cheng, C.M., Tan, Q.M. & Peng, F.J. 2000. On the mecha-
is the initial porosity. nism of the formation of horizontal cracks in a vertical col-
By considering eqs. (26) and (27) in the pore pres- umn of saturated sand. ACTA Mechanica Sinica (English
sure gradient and considering the consolidation of the Serials) 17(1): 1–9.
Fiegel, G.L. & Kutter, B.L. 1994. Liquefaction mecha-
solidification zone, we can compute the development nism for layered sands. ASCE J Geotech Engrg. 120(4):
of cracks at these conditions. 737–755.
Florin, V.A. & Ivanov, P.L. 1961. Liquefaction of saturated
sandy soils. In British National Society (ed.), Proc. 5th
6 COMPARISON WITH EXPERIMENTAL int. Conf. On Soil Mech. Found. Engrg., 17–22 July, 1961.
RESULTS Paris, DUNOD, 1: 107–111.
Kokusho, T. 1999. Water film in liquefied sand and its effect
on lateral spread. J Geotech and Geoenviron Engrg. 10:
It is shown that the results computed by numerical
817–826.
method and the simplified method are close to the Kokusho, T., Watanabe, K. & Sawano, T. 1998. Effect of
experimental results (Figure 3, parameters used in water film on lateral flow failure of liquefied sand. Proc.
computing is the same given in literature 7). The sim- 11th European Conf. Earthquake Engrg., Paris, CD
plified method presented in this paper may be used to publication, ECEE/T2/kokeow.pdf.
compute the evolution of the water film. Scott, R.F. 1986. Solidification and consolidation of a
liquefied sand column. Soils and Foundations 26(4):
23–31.
Seed, H.B. 1987. Design problems in sand liquefaction.
7 CONCLUSIONS ASCE J Geotech Engrg. 113(8): 827–845.
Zhang Junfeng 1998. Experimental study on the strength-
Numerical simulations show that there are stable water ening of percolation and the damage of structure under
films only in the conditions that: (1) the porosity of the impact loading. dissertation for Ph.D, Institute of mechan-
upper part of the sand column must be smaller than that ics, Chinese Academy of Sciences.
of the lower. (2) The keeping of the jamming state.
A simplified method for evaluating the thickness of
water film is presented. It is shown that the simplified
method are agree with the experimental results.

266
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Some geomorphological techniques used in constraining the likelihood

of landsliding – Selected Australian examples

A.S. Miner & P. Flentje

Faculty of Engineering, University of Wollongong, Wollongong, NSW, Australia

C. Mazengarb & J.M. Selkirk-Bell

Mineral Resources Tasmania, Department of Energy, Infrastructure and Resources, Hobart, Australia

P.G. Dahlhaus
Department of Geology, University of Ballarat, Ballarat, Victoria, Australia

ABSTRACT: Techniques for landslide risk management in Australia have evolved considerably since the
publication of the first formal process in 1985. The Australian Geomechanics Society recently published the next
generation of updated landslide risk documents in 2007. The estimation of landslide likelihood is fundamental to
the outcome of the landslide risk management process. However, experienced practitioners still regard this com-
ponent as one of the most difficult and challenging aspects of the assessment as it requires information about the
age of landslides, an understanding of landscape processes and the rate of slope evolution. Such information is dif-
ficult to obtain and is often not a core competency among practitioners undertaking landslide risk assessment. In
order to provide insight into the methods of estimating and constraining landslide likelihood, a number of different
geomorphological approaches are herewith reviewed through a series of selected Australian cases studies. Whilst
the case studies highlight inherent limitations and uncertainties they also demonstrate how geomorphological
studies can provide validation and constraints to a quantification of likelihood and ultimately risk.

1 INTRODUCTION 2 LANDSLIDE FREQUENCY

Techniques for landslide risk management in Aus- The AGS 2007 lists various techniques for assessing
tralia have evolved considerably since the publication landslide frequency. These include:
of the first formal process in 1985 (Walker et al
• Landslide inventory compilation by gathering local
1985). In 2007 the Australian Geomechanics Society
historical records.
released a significant set of updated guidelines that
• Correlation between landslide events and rainfall.
will help shape the nature of slope stability investiga-
• Relationships to geomorphology and geology.
tions in Australia including the need for quantitative
• Simulation modeling and probabilistic analyses.
risk assessment (AGS 2007a & c). The adoption of
• Knowledge based expert judgment.
these guidelines as a requirement for development
approval by regulators, pose considerable challenges The first two techniques involve periods of record
for stakeholders. There are several reasons for this, rarely spanning longer than 140 years. This typically
including a paucity of suitable published geological allows estimation of annual frequencies of the order of
information for the urban environment and a skills 0.01 (10−2 ) for a single event in a particular location.
shortage for suitably qualified practitioners. As such, these techniques have greatest application in
However, one of the most significant issues imped- the first two categories of the AGS likelihood clas-
ing the estimation of risk is an underinvestment of sification system as shown in Table 1 (AGS 2007c,
hill country geomorphological research; hence land- Appendix C).
scape age, process rates and likelihood are often Whilst the latter two methods are commonly used
unknown or poorly constrained. Without an adequate to provide validation to inventory and observational
understanding of likelihood, a risk assessment may techniques, they are more importantly used to pro-
require a highly conservative approach or worse be vide estimates of longer frequency occurrences which
unreliable. are evident in the landscape. To further highlight this

267
Table 1. Qualitative and quantitative measures of likelihood (AGS 2007c).

Approximate annual
probability

Indicative National Implied indicative landslide

value boundary recurrence interval Description Descriptor Level

10−1 5 × 10−2 10 years 20 years The event is expected to occur over ALMOST A
the design life. CERTAIN
10−2 5 × 10−3 100 years 200 years The event will probably occur under adverse LIKELY B
conditions over the design life.
10−3 5 × 10−4 1000 years 2,000 years The event could occur underadverse POSSIBLE C
conditions over the design life.
10−4 5 × 10−5 10,000 years 20,000 years The event might occur under very UNLIKELY D
adverse circumstances over the design life.
10−5 5 × 10−6 100,000 years 200,000 years The event is conceivable but only under ex- RARE E
ceptional circumstances over the design life.
10−6 1,000,000 years The event is inconceivable of fanciful BARELY F
over the design life. CREDIBLE

• This table should be used from left to right; use approximate annual probability or description to assign descriptor, not vice
versa.

Figure 1. Areas of human knowledge and historical interpretation displayed with indicative time ranges.

fact the application of various evidence based meth- operating in the past are the same as those that can be
ods, including indicative time ranges over which these observed operating in the present. Unfortunately, uni-
evidence sources may extend, is shown in Figure 1. formitarianism only tells part of the tale in that we can
The case studies provided in this paper specifically now see that the rates of geomorphological processes
focus on the use of geomorphological techniques to have changed over geological time due to such factors
assist in the classification, validation and constraint of as plate tectonics and changing sea levels and climate.
landslide frequency. With good research and site based observations, we
are increasingly able to place some constraints on these
issues and factors. This is well demonstrated in the fol-
3 GEOMORPHOLOGICAL PROCESSES lowing three case studies. The location of the three case
study sites are shown in Figure 2.
Geomorphology is the study of landforms, includ-
ing their origin and evolution, and the processes that
shape them. Geomorphology focuses on the analysis 4 SOUTH WESTERN VICTORIA
of interconnected physical processes which shape or
have shaped the landscape. Integral to the study of On March 28 1990, two high school students were trag-
geomorphology is Hutton’s 18th century principle of ically killed by a rockfall at the Lal Lal Falls Reserve.
uniformitarianism that assumes the natural processes The accident was the subject of a coroner’s inquest

268
 using a digital terrain model. It has been estimated
(Dahlhaus and Miner, 2005) that the Lal Lal Falls
have retreated a distance of approximately 1650 m in
the past 3 million years which averages 0.55 mm/year.
By comparison, the adjacent Moorabool Falls have
retreated about 1400 m over the same time period—
an average of 0.46 mm/year. Assuming a constant
erosion rate, the valley slopes have receded away from
the streams at 0.015 mm/m2 /year. Based on these cal-
culations, a volume of 0.04 m3 /year has eroded from
the upper columnar basalt cliff of the entire south-
ern slope over the past 3 million years. Based on
the observed average column diameter of 0.8 m, this
equates to an average column height of 80 mm per
year or more realistically a 0.8 m diameter column of
0.8 m length eroding every decade (an ARI of 1 in
10 years and a classification of Almost Certain as per
AGS 2007). Despite the unavoidable assumption of a
uniform rate of landscape evolution the calculated rate
serves as a useful estimate.

4.2 Rock fall frequency assessment

4.2.1 Hazard types and historical inventory data
The entire cliff face surrounding the falls is prone to
both block toppling and rockfall. Open joints, large
Figure 2. South Eastern Australia showing the location of
three case study sites discussed in this paper. columns leaning out from the cliff face and undercut
columns are evident at the site.
Information on previous accidents at the Lal Lal
Falls Reserve as sourced from local newspaper and his-
which has resulted in detailed geological and geomor-
torical records is poor. However historic photographs,
phological investigations at the site (Cooney 1990;
taken on previous geology student excursions, proved
Dahlhaus and Miner 2000).
to be a useful record of rockfall events. Comparison
The site is a small recreation and scenic reserve
of photographs from 1975 to 1990 shows few changes
surrounding a waterfall, about 18 km southeast of the
in the cliff faces, and major rockfalls are not evident.
City of Ballarat. The waterfall, approximately 35 m
However the actual site of the 1990 fatality was pho-
high, is located at the head of a gorge bordered by
tographed in 1972, with both the column involved
steep cliffs along the Lal Lal Creek (Figure 2).
in the accident and a 0.8 m length that had fallen
Dense, olivine-rich basalt of Pliocene - Pleistocene
sometime prior, still in place. Both columns represent
age is exposed in the gorge. The basalt comprises
approximately 0.2 m3 volume.
two separate flows, both exhibiting columnar jointing,
Another rockfall event at a different site on the
with the columns in the upper flow being narrower
upper cliff was also observed in May 1992. This fall
(1/2 − 1 m) than in the bottom flow (1 1/2 − 2 m)
involved a toppling failure of a column 1.5 m high by
(Dahlhaus and Miner, 2000).
0.5 m diameter (approx 0.3 m3 ).

4.1 Geomorphic evolution 4.2.2 Calculation of rockfall frequency

Based on the calculated average erosion rates, the
The falls have formed by headward erosion of the Lal observed volumes of the rockfalls at the fatality site
Lal Creek over the past 3 million years. Below the falls each represent five year’s average erosion volume, sug-
the creek follows a deeply incised valley that reflects gesting an average frequency of 1 in 5 years. The
the original meandering course of the stream. Above 0.3 m3 rockfall observed in 1992 is 7 1/2 year’s average
the falls the valley of the meandering Lal Lal Creek is erosion volume, or a 1 in 7 1/2 year event. All three
very shallow. events are of a similar size.
A total of 0.7 m3 is known to have fallen from the
4.1.1 Calculation of erosion rates southern cliff face in three separate events between
The volume of material eroded from the landscape 1972 and 1992 indicating an observed annual rate of
since the emplacement of the basalt was calculated (0.035 m3 /y). The photographic record and historical

269
records suggest that no other significant falls occurred
during this time. The observed annual rate compares
well to the calculated erosion rate (0.04 m3 /y) with
the additional volume probably made up from smaller
falls and erosion.

4.3 Likelihood estimates

Calculation of rockfall likelihood for the Lal Lal site
was achieved through the use of the calculation of the
erosion rate in combination with site observations and
the establishment of a site inventory through anecdotal
and historical evidence.
Rockfalls within the study area with an estimated
frequency as described fall into the Almost Certain
category as per the AGS classification system. It is
however very important to note that the techniques
only allow for such a likelihood classification if it is
applied to the entire study area as average rates of
erosion have been used. Hence a better statement about
rockfall likelihood from this study is that rockfalls of
Figure 3. Proposed landslide size frequency curves for
approximately 0.2 to 0.3 m3 could be expected to occur
slides and flows normalised for 1 km of escarpment slopes
somewhere within the 1.65 km length of the study area between the suburbs of Thirroul and Clifton. It is not
every 5 to 7.5 years. appropriate to use these curves for slopes outside this area.

5 ILLAWARRA REGION, NEW SOUTH WALES

from the actual mapped slides and flows have been
used as a guide in the development of the interpretive
Landsliding is a common occurrence along the
size-frequency curves shown in Figure 3 below (as
Illawarra Escarpment. The University of Wollon-
discussed in Moon et al 2005).
gong Landslide Inventory (LI) identifies almost 600
The area under each curve (the landslide process
landslide sites on the escarpment. Most of these land-
rate) represents the annual average volume ‘mobilised’
slides have occurred since 1950. Susceptibility zoning
by slide and flow landsliding respectively per one km
(Flentje et al, 2007) identifies 12% of the land as either
of escarpment in this area. This approximate total vol-
high or medium susceptibility whilst containing 92%
ume is 29,000 m3 . Of course, the material mobilised
of the known landslides.
does not mean this amount of material is directly
removed from the escarpment slopes. Not withstand-
5.1 Landslide process rates for the thirroul ing the inherent uncertainties involved, if an ‘average’
to Clifton escarpment rate of movement of 1 cm per year is adopted for all
the landslides combined, the volume removed annu-
The Illawarra escarpment is a variable geomorpholog-
ally from the escarpment reduces to approximately
ical feature. In the southern half of the study area it is
300 m3 .
quite a mature feature with a wide coastal plain and
The annual number of landslides per volume class
well developed and incised drainage lines.
per 1 km of escarpment slope can then be used to
North of the suburb of Thirroul, to Clifton the
assign a broad range of landslide likelihoods as per
escarpment coastline comprises a series of small bays
the AGS classification system. However, each volume
and headlands which merge to the north into a cliff
class must be considered independently as the rate of
line up to 100 m in height. This 10 km section of
displacement and associated consequences for each
escarpment is currently being undercut at or near sea
can greatly affect the risk.
level by marine erosion and is relatively over steep-
ened when compared to the escarpment further to the
5.2 Comparison of process rates with estimates
south. It is not surprising then to note that it contains
derived from rates of sea-floor spreading
a significantly higher density of landslides.
This area includes 155 landslide sites, of which 123 The Illawarra Escarpment is an erosional escarp-
are slide category landslides, 16 are debris flows and a ment that has evolved on a passive continental mar-
further 16 are rock falls (from various different source gin. This margin developed following the rifting and
areas). The annual size frequency distributions derived continental breakup between Australia and the Lord

270
Howe Rise/Dampier Ridge approximately 80 million landslide risk, particularly in terms of the new AGS
years ago (Brown et al, 2003). guidelines and documents.
The outer edge of the continental shelf is approx- The overall approach being used by MRT includes
imately 30 km offshore north of Wollongong. If it is geological and geomorphological mapping and the
assumed that the outer edge of the continental shelf compilation of a landslide inventory. When combined
represents the point from which the escarpment has within a GIS framework these information layers are
retreated since the onset of sea floor spreading, then then used to produce landslide susceptibility maps.
an overall rate of escarpment retreat can be estimated. While this form of a landslide assessment is useful,
These limitations suggest 30 km of escarpment retreat the future production of hazard maps would be a sig-
in 80 Million years (Ma), or approximately 0.375 m nificant advance allowing, for example, tolerable risk
per 1000 years, on average during the geological life levels to be determined.
of the escarpment. However it must be acknowledged Whilst the MRT landslide mapping project is yet
that this does not consider changes in sea level during to be completed the following example provides an
this period, other tectonic influences and climatic vari- insight into landslide likelihood as gained from geo-
ations which will also impact upon rates of escarpment morphic evaluation of the landscape.
retreat. Such influences will have had marked impacts
on this ‘average’ rate of retreat, at times causing it to
be higher and at others forcing it lower.
6.1 Devonport-Burnie, NW Tasmania
This rate indicates an average annual volume of
approximately 600 m3 of material is removed from The Devonport-Burnie area is located on the north-
a 1 km width of the escarpment (given an escarpment west coast of Tasmania, overlooking Bass Strait which
slope length of 1700 m, such as north of Thirroul). separates the island from mainland Australia to the
A percentage of this can occur as landslides in one north.
form or another (such as slides, flows or falls) whilst The dominant features of the coastal strip are a
some will also be lost by alluvial processes. hinterland of deeply dissected sub horizontal Tertiary
Hence this volume as determined from a broad basalt, an escarpment (fossil sea cliff) and associated
geomorphological approach is of the same order of coastal plain as well as contemporary coastal cliffs on
magnitude (albeit double the value) as the figure headlands. Fluvial terraces occur in major waterways.
determined above on the basis of data from the LI. Based on dating by Murray-Wallace & Goede (1995)
the escarpment is believed to have formed during the
last interglacial sea level high at about 125,000 years
5.3 Discussion (125 Ka) BP, and tectonically uplifted by about 22 m.
The uplift is not only reflected in the elevated coastal
Clearly, the LI does not identify all the landslides and
plain (marine terrace), but also in incised catchments
the reported volumes within the LI are not precise. In
and the presence of fluvial aggradational terraces at
addition the rate of average movement is uncertain.
similar elevations above the channels adjacent to the
However, despite these ‘averages’ it is encouraging
elevated coastal plain.
to see that the estimates from the two approaches to
The fossil sea cliff and incised valley walls have
process rate determination are of comparable magni-
collapsed in numerous places and in two major styles
tude. This compatibility helps validate the modeling
(deep seated landslides and earthflows) after sea
assumptions and adds substantial credibility to the out-
levels retreated. Landslides that runout over the
comes, despite the uncertainty which is implicit. Such
marine terrace and matching aggradational fluvial ter-
assessments, founded in sound site based observations
races are therefore stratigraphically less than 125 Ka
and a rational geomorphological understanding of the
old (Figures 4 and 5). Such features are relatively
processes active in slope formation, are fundamental in
common.
achieving realistic assessment of landslide frequency
The age constraints provided brings the understand-
required under the AGS classification system.
ing of landslide age from millions of years (basalts
ages are over 10 Ma old) down almost three mag-
nitudes (<125,000 years). This would initially tend
6 NORTH WEST TASMANIA to suggest a conservative estimate of Rare likelihood
under the AGS system of classification if age is simply
Landslides are a widespread feature in the Tasmanian equated to annual probability in the absence of known
landscape and there is much pressure to develop land- triggering events.
slide affected terrain. Mineral Resources Tasmania However whilst rainfall and elevated groundwater
(MRT) is progressively undertaking regional land- are likely to be key triggering mechanisms, if climate
slide mapping of urban areas and their surrounds to was considered to be cool and dry during the glacials
assist stakeholders such as local councils in addressing (these last in the order of 100,000 years), it may imply

271
 that they would not occur in the present day. However
as the discussion indicates, likelihoods may well
be at least an order of magnitude higher. Without
considered opinion of geomorphic processes initial
unsupported assumptions may be dangerous as they
could significantly underestimate landslide hazard.

7 DISCUSSION AND CONCLUSIONS

Figure 4. Orthophotograph of relic (prehistoric) landslide
that has runout nearly 200 m over the 125 Ka bench. Other The case studies provide insight into how an under-
terrain features are also depicted to provide indication of map- standing of long term geomorphological processes
ping approach. It is believed that a significant component of can validate inventory and observational techniques
the colluvial footslope is composed of earth flow deposits and allow estimates of low frequency landslide events.
derived from the basaltic soils in the escarpment. Orthophoto
from Department of Primary Industries and Water, Hobart.
Techniques such as scarp retreat rate and regional sur-
face dating can provide opportunities in constraining
landslide likelihood as defined in the AGS system of
classification. The case studies also clearly highlight
the limitations and uncertainties with such estimations
and emphasize the importance of using all avail-
able techniques when assessing landslide likelihood
as required under the AGS system of landslide risk
management.

REFERENCES

AGS, 2007a. Australian Geomechanics Society Working

Figure 5. Conceptual model of how landscape chronology Group. Guideline for landslide susceptibility, hazard and
can be used to constrain the age of landslides in an incised risk zoning for land use planning. Australian Geomechan-
valley situation. ics Journal, Volume 42, No. 1 March, pp 13–37.
AGS, 2007c. Australian Geomechanics Society Working
Group. Practice note guidelines for landslide risk man-
that landslide activity was greater during the inter- agement. Australian Geomechanics Journal, Volume 42,
glacials. Given that we are currently in an interglacial No. 1 March, pp 63–114.
period, likelihood may not be simply calculated as an Brown B.J. Muller, R.D. Gaina, C. Struckmeyer, H.M. Stagg,
H.M.J. & Symonds, P.A., 2003. Formation and evolution
average over 125 Ka but could be higher by as much of Australian passive margins: implications for locating
as 2 orders of magnitude. Hence the current likeli- the boundary between continental and oceanic crust. In
hood might more reasonably be assumed to be Unlikely Hillis and Muller (eds.) Evolution and Dynamics of the
to Possible under the AGS system when considering Australian Plate. Geol Soc of Aust Special Pub. 22.
individual landslides. Cooney, A.M. 1990. Geologocaical report on the rock fall
at Lal Lal. Report to the Deputy Coroner. Unpublished
report 1990/13 geological survey of Victoria.
6.2 Discussion of likelihood estimate Dahlhaus, P.G. & Miner A.S. 2000. Estimating the occur-
rence of Rockfalls in columnar Basalt. GeoEng 2000.
While there are many landslides in the Tasmanian (Reprinted in Australian Geomechnaics Vol 36, No. 2 June
landscape, almost all of these features are undated 2001).
which makes it difficult for estimating likelihood of Flentje, P., Stirling, D. & Chowdhury, R.N., 2007. Land-
future landslides and understanding risk at a regional slide Susceptibility and Hazard derived from a Landslide
or local level with any confidence. The popular opin- Inventory using Data Mining—An Australian Case Study.
ion by the geotechnical community, as expressed in Proceedings of the First North American Landslide Con-
geotechnical reports supporting development applica- ference, Landslides and Society: Integrated Science,
Engineering, Management, and Mitigation. Vail, Col-
tions, is that the deep seated landslides such as the one orado June 3–8, 2007. CD, Paper number 17823–024,
described at Burnie are ‘‘fossil’’ features (more cor- 10 pages.
rectly referred to as relict landslides) and formed under Moon, A.T., Wilson, R.A. & Flentje, P.N., 2005. Developing
differing climatic conditions. This implies AGS clas- and using landslide size frequency models. Joint Technical
sifications of Rare and Barely Credible and suggests Committee on Landslides and Engineered Slopes, JTC-1,

272
 in association with Vancouver Geotechnical Society. Pro- Walker, B.F., Dale, M., Fell. R., Jeffrey, R., Leventhal,
ceedings of the International Conference on Landslide A., McMahon, M., Mostyn, G. & Phillips, A. 1985.
Risk Management/18th Annual Vancouver Geotechni- Geotechnical Risk Association with Hillside Develop-
cal Society Symposium, Vancouver. May 31 to June 4, ment. Australian Geomechanics News, No10, pp 29–35.
pp 589–598.
Murray-Wallace, C.V. & Goede, A. 1995. Aminostratigraphy
and electron spin resonance dating of Quaternary coastal
neotectonism in Tasmania and the Bass Strait islands.
Australian Journal of Earth Sciences, 42: 51–67.

273
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Rock failures in karst

Mario Parise
National Research Council, IRPI, Bari, Italy

ABSTRACT: Rock failures in karst environments present peculiar features related to the typical setting of karst,
characterized by presence of caves created by chemical solution of soluble rocks, direct connection between the
surface and the subsurface, and an overall high fragility of the environment. In addition to landslides s.s. and
sinkholes, breakdown processes within caves are extremely common in karst, and may represent a geohazard even
to the built-up environment, due to possibility of void migration toward the surface. They deserve, therefore, great
attention by scientists, and should be carefully examined through an integrated approach which has necessarily
to include direct surveying in natural karst caves.

1 INTRODUCTION Rock failures in caves represent a danger to cavers

and karst scientists, and, at those sites where show
Karst processes affect soluble rocks, creating pecu- cave exploitation is carried out, even to tourists. Fur-
liar landforms at the surface, and caves and drainage ther, the instabilities that occur underground may have
systems underground, through chemical solution of indirect effects at the surface, producing likely dam-
the bedrock. The main rock type interested by karst age to the built-up environment, and creating a very
development is carbonate. Nevertheless, evaporites subtle geohazard to engineers.
are greatly involved too, and other rock types such as
sandstones and quartzites, even though very resistant,
may be also affected by development of karst and/or 2 TYPES OF FAILURES
pseudokarst caves.
An important, often underestimated, process in Karst environments present peculiar features that cre-
karst is represented by rock failures, with the produc- ate the possibility of different types of rock failures:
tion of huge amount of breakdown deposits in caves
1. landslides sensu strictu, in those karst areas charac-
(White & White 1969). Slope instability in karst plays
terized by slope relief and geological settings prone
actually a twofold role: it occurs upon it, at the sur-
to slope movements;
face, as in many other settings where steep to vertical
2. sinkholes related to the presence of subterranean
rock walls crop out. In addition, rock failures in karst
cavities;
significantly shape the underground systems. Some
3. rock failures in the underground environment.
of the largest rooms in caves all over the world are
situated at the sites of intersections of faults and/or In the following, I will briefly deal with categories
fractures, that is at the main weakness zones in the 1 and 2 above, mostly focusing the attention on the
rock mass. Caves, as we see them today, have been features within natural karst caves.
produced by the combined action of dissolution (karst
sensu strictu) and breakdown processes. The latter
2.1 Landslides
become particularly significant once the cave pas-
sages are abandoned by the flowing water, which The phenomena belonging to category 1 are studied
level is lowered below the passage horizon (White and analyzed following the usual approaches defined
1988). Lack of the support exerted by water on the and well established for slope movements, in particular
cave walls and roof, together with the degree of frac- for those affecting rock slopes (De Freitas & Watters
turing in the rock mass, determines the continuous 1973; Hoek & Bray 1981; Cruden & Varnes 1996).
detachment of blocks and slabs of rock, and their later Even in flat karst regions, where slope relief is gen-
fragmentation into chips, thus enlarging the cave size. erally low, occurrence of rock failures may be common
Upward, the process may proceed until reaching the along the walls of the main valleys, as shown by the
ground surface, to produce collapses of variable size topples in figure 1. In addition to the usual factors that
and depth. generally influence or predispose to slope movements,

275
 Figure 2. Rockslide that occurred in January, 2002 in the
Amalfi coast (Campania, southern Italy).

Figure 1. Topple failures long the walls of a deep karst

valley (Laterza, southern Italy). Figure 3. Collapse sinkhole in carbonate rocks of the
Gargano Promontory (Apulia, southern Italy).

the presence of karst voids and conduits, and the deriv- Sinkholes may be classified into six main types,
ing modality of water circulation may further favour distinguished by their genetic process and morpho-
in karst the occurrence of failures, as illustrated by logy (Waltham et al. 2005). These include: solution
Santo and co-workers (2007) in their study of carbon- sinkhole, created by the slow process of dissolutional
ate rock failures in the Campania region of southern lowering of the surface; collapse sinkhole, caused by
Italy (Figure 2). rock roof failure into an underlying cave; caprock sink-
hole, involving failure of insoluble rock into cave in
soluble rock below; dropout sinkhole, when the soil
collapses into soil void formed over bedrock fissure;
suffosion sinkhole, consisting of down-washing of soil
2.2 Sinkholes into fissures in bedrock; and, eventually, buried sink-
Sinkholes are rock failures typical of karst envi- hole, when an old sinkhole in rock becomes soil-filled
ronments, being related to the presence of subter- after environmental changes have occurred.
ranean cavities (Figure 3), both natural and man-made
(Waltham et al. 1986; Culshaw & Waltham 1987). In
the last decades they have been object of many studies 3 ROCK FAILURES IN KARST
in different parts of the world, and their research has
become an important and specific field of engineer- Caves are a notable karst geohazard, due to the poten-
ing geology in many countries (Delle Rose et al. 2004, tial for gravitational collapse of rock and/or soil into
Gutierrez et al. 2004, Dogann 2005). them, either naturally or under induced load (Waltham
Sinkholes are a subtle type of failure, since they & Lu 2007).
may occur even on very gentle slopes, and potentially Karst caves formed in phreatic environment ide-
affect inhabited areas and man-made infrastructures. ally present a cylindrical shape, produced through

276
slow development in geological times, and may have
reached a stable configuration with equilibrium roof
arches. However, reduction in the thickness of the roof
might determine further instability processes. In frac-
tured rock masses, the setting is more complicated
depending upon frequency and orientation of the main
discontinuity systems. In any case, instability phe-
nomena may become the main evolution process in
caves initiated by classical karst evolution. Through
natural stoping and upward cavity migration, the pro-
gressive rock failures can go on until reaching the
ground surface.
Rock failures in underground environments are rep-
resented by falls, topples, and slides that affect the
walls and roofs of karst caves. Since inspection of
the underground world is generally restricted to expe-
rienced cavers, it becomes clear that these types of
failures have to be studied by scientists-cavers, or alter-
natively by a team at least partly composed by cavers.
Besides the inherent difficulties of the hosting environ-
ment, rock failures within caves often differs in some
Figure 4. The dotted line marks the ‘‘tension dome’’ or zone
ways from the analogue types observed at the earth’s of maximum shear stress induced by the presence of a cavity
surface, due to greater importance of weathering pro- (modified after White 1988).
cesses (Fookes & Hawkins 1988, Anon 1995, Zupan
Hajna 2003, Forti & Parise 2008), and groundwater
chemistry and circulation (Harmon & Wicks 2006,
Ford & Williams 2007). fractured, stability of the arched roof profiles is much
Cave roof failure is the type of great worrying, since less guaranteed.
its evolution, and upward stoping, may eventually lead Among the different approaches to analyze such
to ultimate surface collapse. Mechanisms and rates of an evolution, numerical modelling has been advanced
cave roof failure is essentially dependant upon geo- by defining fractured rock masses in terms of their
logical features of the rock mass, namely attitude and ‘‘rock mass ratings’’ (RMR). The geomechanic sys-
bedding (Waltham & Lu 2007). tem derives RMR by summing rating values ascribed
Even though the matter is extremely complicated, on the basis of RQD (rock quality designation, based
and a number of other factors might play crucial roles, on fracture intersections in borehole core), mean frac-
a commonly accepted rule of thumb is that an indi- ture spacing, fracture conditions, fracture orientation,
vidual bed of strong limestone is going to fail when unconfined compressive strength of the intact rock and
the cave’s unsupported span exceeds 10–20 times the groundwater state (Bieniawski 1973; Hoek & Brown
bed thickness. Besides the aforementioned attitude 1980).
and bedding of the rock mass, however, other factors RMR values range from more than 80 for very good
must be considered, including presence and typology rock of rock mass class I to less than 20 for very poor
of discontinuities, weathering, and water flow. rock of rock mass class V; they may be correlated
Discontinuities weaken the rock mass, breaking its with Q values derived from the alternative Norwegian
continuity, and constitute preferential ways to water classification scheme (Barton et al. 1974).
flow. In karst, the latter means essentially a combina- RMR for typical cavernous karst in strong lime-
tion of mechanical erosion and chemical solution that stones is taken conservatively as between 30 and 40,
may result in creating voids in the rock. whilst in chalk and other weak or thinly bedded lime-
Over a void, a zone of ground compression in stones it may be estimated as nearer 20 (Waltham &
the form of an arch naturally develops (Figure 4). Lu 2007).
In a cave, the arch is produced through progressive However, direct observations and surveys in natural
detachment of beds from the tension zone beneath the caves show a great variety in the range of values of the
compression arch. In this way, caves may reach a stable stability parameters which are strongly a function of
configuration, with arched roof profiles approach- the rock mass quality (Figure 5).
ing the stability of a voussoir arch in uncemented The instability process within stratified hard rock
masonry (an arch formed of shaped blocks designed masses has been widely discussed by Diederichs &
to be stable in compression). When jointing becomes Kaiser (1999), according to the traditional Voussoir
extensive, and the rock mass is heavily to extremely beam theory (Evans 1941), suitably revisited by them.

277
 Andrejchuk (2002) in their study on the gypsum
caves in Western Ukraine (a region where the five
largest known gypsum caves in the world are present),
the determination of the degree of propagation of the
breakdown structures toward the surface through the
cover, allows to assess the related geohazard with a
precision and certainty unavailable by the approaches
of conventional engineering geology (Klimchouk &
Andrejchuk 2002).
In order to best evaluate the gravity-related pro-
cesses in the cave, there is the necessity to map any
outlet features that may indicate considerable break-
out in the vault: from domes and cupolas, to domepits
in cave passages. In addition, mapping of rockfall
deposits, together with determination of their size, and
possibly of the origin of detachment (vault, wall, etc.),
and of taluses deposits is also important (Parise &
Figure 5. Cave stability related to cave width and rock mass Trisciuzzi 2007). Stage and activity of the breakdown
quality (after Waltham, 2002). Q value is after Barton et al. may be ascertained through recording several features
1974. in the cave such as maximum height of the cupolas,
presence of open cracks and fissures at its margin,
water seepage or flow in the breakdown taluses, damp-
In particular, they suggest that timing for roof collapse ness of breakdown sediments, freshness of fallen
is often controlled by residual tensile strength available blocks.
along the existing rock bridges (Diederichs & Kaise, Klimchouk & Andrejchuk (2002) observe, in their
1999). According to the Authors, the tensile strength study on the breakdown structures in the Zoloushka
degradation due to humidity and chemically assisted cave, that the largest density of breakdown structures
stress corrosion can be responsible for failure of spans was observed in regions of not very large passages;
that remained safe for very long term. Therefore, the this finding strongly contrasts with one of the most
transition from systems of stratified continuous beams established views, which suggests that breakdown for-
to systems of stratified voussoir beams due to time- mation is controlled primarily by passage size. The
dependant tensile strength degradation as an effect of Authors above identify two different mechanisms of
weathering processes, and the following collapse of breakdown formation, with mechanism A leading only
the stratified blocky roof, is thought to be among the to gradual subsidence and not collapse, which, on
main processes acting in underground karst systems. the other hand, is produced through mechanism B
Such a process is locally favoured by water infiltra- (Fig. 6).
tion from the ground surface, which may produce
significant weathering effects.

3.1 The importance of sub-surface investigations

High variability is probably the main feature of karst
landforms and features. At a same site, part of a cave
may be stable and present sound roofs, whilst another
may be characterized by weak zones with high percent-
age of joints and/or broken rocks. This translates into
the need to pursue as much as possible direct observa-
tions (including those from caving activities) in order
to collect useful databases to implement numerical
modelling aimed at assessing the stability conditions
in underground karst environments.
Investigating extensive subsurface cave systems,
and mapping the main morphological features related
to breakdown processes, including those not yet mani-
fested at the surface, may provide a significant contri- Figure 6. The two mechanisms of breakdown formation
bution to the understanding of the different stages of hypothesized by Klimchouk & Andrejchuk (2002) in their
breakdown development. As shown by Klimchouk & study of gypsum caves of Western Ukraine.

278
 Even the presence of smooth dolines at the surface Parise 2003) may work to strongly change the natural
may be related to past events of collapse, likely within setting, with likely severe consequences and damage
still unexplored caves. The example of Dan’kivsky for the anthropogenic environment.
shaft (a 22 m deep shaft that opened on January 11, Studying rock failures in karst is therefore a chal-
1998 in Ukraine) illustrates how the shape of a surface lenging but highly stimulating issue. In many cases,
form is not necessarily indication of its sudden or grad- however, traditional approaches are not sufficient to
ual origin (in other words, of a collapse or subsidence fully appreciate the phenomena. Lacking detailed data
formation). The original deep shaft, in fact, within few on discontinuity systems in the rock mass, and/or
years transformed into a gently sloping doline, due to direct observations in caves, an estimate of the overall
the presence of soft sediments within the overburden conditions of a site can be derived from the engineering
(Klimchouk & Andrejchuk 2002). classification of the karst, assessed from broad visual
In stratified soluble rocks, especially in evaporites inspection (Waltham & Fookes 2003). However, such
showing a cover beds (Figure 7), most of the break- estimates should be taken into account only as a first
down features are generally originated through more approximation, whilst collection of further data (pos-
or less prolonged multi-stage development rather than sibly through direct inspection and observations) is
through a single massive collapse of the roof (Iovine strongly encouraged.
et al. 2007). This is also indicated by connection An interesting point is that availability of sur-
through the different storeys by large pits, which rep- face and subsurface investigation may allow a double
resented hydraulic communication at different levels geomechanical survey, which has to be carried out
during the cave formation. according to the internationally accepted and codi-
fied standards (ISRM 1978). Through this approach,
it will be possible to identify the main discontinuity
sets in the rock mass, their continuity within the cave
4 DISCUSSION AND CONCLUSIONS
systems, and the control they have on the occurrence
of slope failures. The data so collected can be used
Due to their intrinsic characteristics, karst environ-
to identify the main processes that trigger breakdown
ments are extremely susceptible to a number of geo-
in caves, to assess the related hazard to man and the
hazards (White, 1988; Parise & Pascali 2003; Parise
anthropogenic structures, and to implement numerical
& Gunn 2007), including rock failures. These may
models dedicated to the comprehension of the moda-
involve the landscape both at the surface and in the
lity of evolution of the rock mass (Lollino et al. 2004;
subterranean world, which in karst are strongly and
Waltham & Swift 2004).
intimately linked. Changes induced by man in the
surface karst environment may produce in fact rapid
events underground, contributing to accelerate the
generally slow process of karst solution. For example,
external loading due to road construction and traffic REFERENCES
(Kambesis & Brucker 2005), or high water flow deriv-
ing from diversion of the natural surface flow, clogging Anon, 1995. The description and classification of weathered
of original swallets, and/or impermeabilization of nat- rocks for engineering purposes: Geological Society Engi-
neering Group Working Party Report. Quart. J. Eng. Geol.
ural surface drainage by man (White & White 1984; 28: 207–242.
Barton, N., Lien, R. & Lunde, J. 1974. Engineering classifi-
cation of rock masses for tunnel design. Rock Mechanics
6: 189–236.
Bieniawski, Z.T. 1973. Engineering classification of jointed
rock masses. Trans. South Afr. Inst. Civil Eng. 15:
335–343.
Cruden, D.M. & Varnes, D.J. 1996. Landslide types and pro-
cesses. In A.K. Turner & R.L. Schuster (eds) Landslides.
Investigation and Mitigation. Transp. Res. Board, sp. rep.
247: 36–75.
Culshaw, M.G. & Waltham, A.C. 1987. Natural and artificial
cavities as ground engineering hazards. Quart. J. Eng.
Geol. 20: 139–150.
De Freitas, M.H. & Watters, R.J. 1973. Some field examples
of toppling failure. Geotechnique 23 (4): 495–514.
Delle Rose, M., Federico, A. & Parise, M. 2004. Sinkhole
genesis and evolution in Apulia, and their interrelations
Figure 7. Rockfall deposits at the entrance of a gypsum cave with the anthropogenic environment. Natural Hazards and
in Calabria, southern Italy. Earth System Sciences 4: 747–755.

279
Diederichs, M.S. & Kaiser, P.K. 1999. Tensile strength and Italy. In H. Konietzky (ed), Proc. 1st Int. UDEC Symp.
abutment relaxation as failure control mechanisms in ‘‘Numerical modeling of discrete materials’’, Bochum
underground excavations. Int. J. Rock Mech. Min. Sc. 36: (Germany), 29 september—1 october 2004: 49–55.
69–96. Parise, M. 2003. Flood history in the karst environment
Dogann, U. 2005. Land subsidence and caprock dolines of Castellana-Grotte (Apulia, southern Italy). Natural
caused by subsurface gypsum dissolution and the effect of Hazard and Earth System Sciences 3: 593–604.
subsidence on the fluvial system in the Upper Tigris basin Parise, M. & Pascali, V. 2003. Surface and subsurface envi-
(between Bismil-Batman, Turkey). Geomorphology 71: ronmental degradation in the karst of Apulia (southern
389–401. Italy). Environ. Geol. 44: 247–256.
Evans, W.H. 1941. The strength of undermined strata. Trans. Parise, M. & Gunn, J. (eds) Natural and Anthropogenic
Inst. Min. Metall. 50: 475–500. Hazards in Karst Areas: Recognition, Analysis and
Fookes, P.G. & Hawkins, A.B. 1988. Limestone weathering: Mitigation. Geol. Soc. London, sp. publ. 279, 202 pp.
its engineering significance and a proposed classification Parise, M. & Trisciuzzi, M.A. 2007. Geomechanical charac-
scheme. Quart. J. Eng. Geol. 21: 7–31. terization of carbonate rock masses in underground karst
Ford, D. & Williams, P. 2007. Karst hydrogeology and systems. Kras i speleologia.
geomorphology. John Wiley & Sons, 562 pp. Santo, A., Del Prete, S., Di Crescenzo, G. & Rotella, M.
Forti, P. & Parise, M. 2008. The role of weathering in favour- 2007. Karst processes and slope instability: some investi-
ing instability processes in natural karst caves. In D. gations in the carbonate Apennine of Campania (southern
Calcaterra, D. Campbell & M. Parise (eds) Weathering Italy). In M. Parise & J. Gunn (eds) Natural and Anthro-
as predisposing factor to slope movements. Geol. Soc. pogenic Hazards in Karst Areas: Recognition, Analysis
London, Engineering Series sp. publ. and Mitigation. Geol. Soc. London, sp. publ. 279: 59–72.
Gutierrez, F., Guerrero, J. & Lucha, P. 2004. Paleosubsidence Waltham, A.C. 2002. The engineering classification of karst
and active subsidence due to evaporite dissolution in the with respect to the role and influence of caves. Int. J.
Zaragoza area (Huerva river valley, NE Spain): processes, Speleol. 31 (1/4): 19–35.
spatial distribution and protection measures for transport Waltham, A.C. & Fookes, P.G. 2003. Engineering classifica-
routes. Engineering Geology 72: 309–329. tion of karst ground conditions. Quart. J. Eng. Geol. and
Harmon, R.S. & Wicks, C.M. (eds) 2006. Perspectives on Hydrog. 36: 101–118.
karst geomorphology, hydrology, and geochemistry. Geol. Waltham, T. & Lu, Z. 2007. Natural and anthropogenic rock
Soc. Am., sp. paper 404: 366 pp. collapse over open caves. In M. Parise & J. Gunn (eds)
Hoek, E. & Bray, J.W. 1981. Rock slope engineering. Natural and anthropogenic hazards in karst areas: recog-
Institution of Mining and Metallurgy, London, 358 pp. nition, analysis and mitigation. Geol. Soc. London, sp.
Hoek, E. & Brown, T. 1980. Underground excavations in publ. 279: 13–21.
rock. Institution of Mining and Metallurgy. Waltham, A.C., Vandenver, G. & Ek, C.M. 1986. Site inves-
Iovine, G., Parise, M. & Trocino, A. 2007. Looking at tigations on cavernous limestone for the Remouchamps
instability phenomena from different perspectives: an Viaduct, Belgium. Ground Engineering 19 (8): 16–18.
experience from the Verzino area (Calabria, southern Waltham, T., Bell, F. & Culshaw, M. 2005. Sinkholes and
Italy). Epitome 2: 209. subsidence: karst and cavernous rocks in engineering and
ISRM 1978. Suggested methods for the quantitative descrip- construction. Springer, Berlin.
tion of discontinuities. Int. J. Rock Mech. Min. Sc. 15: White, W.B. 1988. Geomorphology and hydrology of karst
319–368. terrains. Oxford Univ. Press, 464 pp.
Kambesis, P. & Brucker, R. 2005. Collapse sinkhole at Dish- White, E. & White, W. 1969. Processes of cavern breakdown.
man Lane, Kentucky. In T. Waltham, F. Bell & M. Culshaw Bull. Natl. Speleol. Soc. 31 (4): 83–96.
Sinkholes and subsidence: karst and cavernous rocks in White, E.L. & White, W. 1984. Flood hazards in karst
engineering and construction. Springer: 277–282. terrains: lessons from the Hurricane Agnes storm. In
Klimchouk, A. & Andrejchuk, V. 2002 Karst breakdown A. Burger & L. Dubertret (eds) Hydrogeology of karst
mechanisms from observations in the gypsum caves of terrains 1: 261–264.
the Western Ukraine: implications for subsidence hazard Zupan Hajna, N. 2003. Incomplete solution: weathering of
assessment. Int. J. Speleol. 31 (1/4): 55–88. cave walls and the production, transport and deposition of
Lollino, P., Parise, M. & Reina, A. 2004. Numerical anal- carbonate fines. Carsologica, Postojna-Ljubljana, 167 pp.
ysis of the behavior of a karst cave at Castellana-Grotte,

280
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Geotechnical study at Sirwani landslide site, India

V.K. Singh
Scientist, Slope Stability Division, Central Mining Research Institute, Dhanbad, Jharkhand, India

ABSTRACT: The paper deals with the landslide study at Sirwani, Sikkim. It was also aimed to know the
influence of water on the safety factor by sensitivity analysis. The rock discontinuities were mapped along
the landslide and road cutting exposures as per the norms of International Society of Rock Mechanics. After
identifying kinematically possible failure modes, detailed stability analysis is carried out by limit equilibrium.
Landslides occur frequently in the Himalayas. This is due to the high intensity of rainfall that contributes to rapid
erosion and weathering of the rock mass, which leads to reduction in the stability of natural slope. Remedial
measures were suggested for better stability in the landslide prone zone.

1 INTRODUCTION The Top most lithology in the area is soil, which is

underlain by shales and phyllites.
The research work was aimed for the landslide study of
Sirwani landslide, Sikkim. It was also aimed to know
the influence of water on the safety factor by sensitivity 3 PHYSICAL CHARACTERISATION
analysis, which tells the importance of the parameter OF DISCONTINUITIES
in the slope. A more justified and suitable remedial
measure can be planned for any critical slope after The geotechnical mapping was done along the expo-
sensitivity analysis. sures of road cuttings. The mapping was done as per
The Sirwani landslide is located at a distance of the norms of ISRM (1978). The general structural pat-
about 28 km from Gangtok and at about 1.7 km from tern in the rock mass of the phyllite is presented in
Singtam on the road of Singtam to Dikchu. The yearly Table 1. It will be appropriate to mention here that the
precipitation in the capital city of Gangtok is relatively characteristics of discontinuities are varying in differ-
high with an annual average of 3500 mm. ent parts of the landslide zone because of the local
The stability of slopes depends on the geologi- folding and the area is geologically complex.
cal structures, geomechanical properties of the slope
materials, groundwater/rainwater condition. The rock
discontinuities were mapped along the exposures
of landslide and road cuttings as per the norms 4 GEOMECHANICAL PROPERTIES
of International Society of Rock Mechanics (ISRM
1978). Geotechnical mapping was undertaken to deter- The samples of the slope mass in which the failure had
mine the critical orientation of structural discontinu- taken place, were tested for various geo-mechanical
ities. After identifying kinematically possible failure properties at CMRI geo-mechanical testing labora-
modes, detailed slope stability analysis was carried out tory. The mineralogical analysis was done with help
by limit equilibrium method.
Table 1. Orientation of major joint sets in phyllite.

2 GEOLOGY OF THE AREA Mean Orientation of Joint Sets

The area near landslide consists of metamorphic are- Joint Dip direction Dip amount Spacing Persis-
naceous and argillaceous rocks intruded by basic sills sets (◦ ) (◦ ) (cm) tence (m)
that have been metamorphosed to epidiorite and tal-
J1 N251 44 20 to 40 0.5 to 10
cose phyllites. A layer of weathered rock material
J2 N186 48 30 to 50 0.2 to 0.4
consisting of sand and silt covers most of these phyllite J3 N320 72 200 to 400 0.4 to 0.7
formations. The area is characterized by steep slope.

281
Table 2. Geo-mechanical properties of the slope material
mass at Sirwani landslide site, Sikkim.

Swell angle
Pressure
(degree) Bulk Density Cohesion Friction
(kg/cm2 ) (kN/m3 ) (kPa)

0.02 2.39 (DC) 0.92 (DC)

35 (DC)
2.64 (UC) 0.32 (UC)
26(UC)

Abbreviations: DC—Drained condition, UC—Undrained

condition.
Major minerals in slope material: Quartz, Illite, Muscovite,
Clinochlore, Phlogopite.
Slake durability test on shale: First cycle: 90.00 %, second
cycle: 58.00 % (Medium durability).

Figure 1. Kinematic analysis for types of failure.

of DB Advance X-ray Diffraction system (from M/s
Bruker AXS, Germany). The relevant geotechnical
parameters of slope material have been presented in
Table 2. the upper weathered-highly fractured rock mass and
the same was observed in the field also. The failure in
this type of slope is always characterised by circular
type of failure. The earthquake effect is simulated in
5 SLOPE STABILITY ANALYSIS
the GALENA software by assigning a value for basic
horizontal seismic coefficient for the concerned area.
Initially the kinematic analysis is done to identify the
The presence of rain/surface water may decrease the
types of failure. The detailed stability analysis was
shear strength of the clays, so sensitivity analysis was
done with the help of GALENA software, which is
done separately to determine the effect of water on fac-
based on Limit Equilibrium Method.
tor of safety. The drained slope mass condition means
that the slope mass has been provided proper drainage
5.1 Kinematic Analysis in and around the landslide site. The undrained con-
dition means that either the drains are not provided or
The average orientations of the discontinuity sets the provided drains are not effectively maintained.
(Table 1) determined from the geologic structural The contour plan (Fig. 2) has been prepared based
mapping were analysed to assess kinematically possi- on the data obtained with the help of sophisticated
ble failure modes involving structural discontinuities & precise Electronic Distance Meter (WILD DI4L,
(Figure 1). The critical discontinuity must lie within Swiss make) during the fieldwork of the landslide
20 degree of the slope face for plane failure to occur. area. The typical section of the landslide area shows
The kinematic analysis was done to determine the the existing steepest slope face direction. The stability
types of failure in the rock mass (Figure 1). The joint analysis was done along this section in drained (Fig. 3)
sets (J1, J2 and J3) present in the area are dipping and undrained (Fig. 4) conditions.
obliquely to the slope face. So a large-scale plane fail- The basic horizontal seismic coefficient for
ure is unlikely in the rock mass. The failure along Gangtok is 0.05 (IS: 1893–1984. 2000). This value
the wedges (W1 and W3) is unlikely because these was used during stability analyses along with the other
wedges are dipping obliquely to the slope face. Small- geotechnical parameters of slope mass (Table 2). The
scale failures are possible along W2 wedge. But it is a results of stability analyses are presented in Table 3.
general phenomenon in the hilly terrain. The stability analysis shows a factor of safety of
1.5 in drained condition, so the slope is likely to be
stable in drained condition. But the factor of safety
5.2 Stability analysis by limit equilibrium method
under undrained condition reduces to less than the
The kinematic analysis shows that large scale plane cut-off value of factor of safety (1.5). So, every attempt
or wedge failures are unlikely in the rock mass. So, should be made to keep the slope in drained condition
the failure was limited to the topsoil mass along with for the slope stability.

282
 Table 3. Slope stability analysis of Sirwani landslide site.

Sr. No. Slope mass condition Factor of safety

1. Drained slope mass 1.48

2. Undrained slope 0.90

maintained. Every attempt should be made to mini-

mize the entry of surface/rainwater to the landslide
area.
There are main two reasons of slides in this area.
First reason is the formation of deep gullies along the
flow channel of the failed slope. In rainy season, a
huge quantity of surface water flows down through
these nala with a great speed. The current of water
Figure 2. Contour plan.
flow used to be very high due to steep gradient of the
topography. The flowing of water with great current
undercuts the existing/overlying slope mass. When the
undercutting is deep enough then the undercut overly-
ing slope mass losses equilibrium and fails down the
slope. The failed mass is again washed away down the
slope after subsequent rains. The failed mass used to
be top debris and under cut highly weathered phyllitic
rock mass. The slide area is being provided by a gar-
land drain to check the entry of rainwater in to the
slide zone. The slide zone is being terraced by provid-
ing stepped concrete steps to check the momentum of
the running water during rainstorm.
The second reason of sliding is the under cutting
at the toe position of the hill slope below road by the
high current of Tista river. Due to this undercutting,
Figure 3. Stability analysis in drained condition.
the overlying slope mass becomes unstable. The heavy
gabions are being provided near the undercut to check
the undercutting by the speedily flowing river water
during the rainy season.
Further, it is likely that topsoil and highly
weathered-fractured phyllitic formation became sat-
urated and lost shear strength due to the intense
infiltration of water during the rainstorm. The water
could not percolate to the underlying unweathered
phyllite. It resulted in high water pressure in the top
fragile formation. Pore pressure reduces the available
shear strength and the weight of the water adds to the
forces that induce sliding.
The slope movements in the landslides involve a
combination of earth slide and debris flow. These land-
slides are characterised as sliding of earth followed
Figure 4. Stability analysis in undrained condition. by flow of the displaced material. The convergence
of surface run-off of all the catchments area in a
small area provides water for an initial failure. If a
6 DISCUSSION landslide occurs during a run-off producing storm,
then surface run-off water will flow down the catch-
Sensitivity analysis shows that the influence of water ments area and underground water flowing from slope
is alarming. The slope may become unstable if the face will spill into the initial failure. Deformation
drainage in and around the failed zone is not effectively accompanying an initial failure may allow further

283
incorporation of water emanating from bedrock this undercutting, the overlying slope mass becomes
springs and surface runoff into the failed material, thus unstable. Proper terracing of the slide area and heavy
increasing debris mobility. gabions are being provided to check the slides trig-
The amount of water available for mixing with land- gered by these two factors.
slide debris and the gradient of the down slope channel The topsoil and highly weathered-fractured phyl-
contribute to the transition of an initial landslide into litic formation becomes saturated and loose strength
a mobile debris flow. Incorporation of excessive vol- due to the intense infiltration of water during the rain-
umes of water may dilute landslide debris and increase storm. The water could not percolate to the underlying
its mobility. unweathered phyllite resulting in high water pressure
in the top fragile formation. So every attempt is now
being made to keep the slope in drained condition and
7 CONCLUSIONS to minimize the entry of surface/rainwater to the land-
slide area by providing suitable drainage in and around
An assessment of the engineering and structural geol- the slide area.
ogy, strength properties and related geotechnical con-
trols show that the existing slope is likely to be stable in
drained condition but the same slope becomes unstable
in undrained condition. REFERENCES
Sensitivity analysis shows that the influence of
water is alarming. The slope may become unstable if IS:1893–1984. 2000. Indian Standard Criteria for Earth-
the drainage in and around the failed zone is not effec- quake Resistant Design of Structures. Bureau of Indian
tively maintained. The landslide is occurring mainly Standards, pp. 58–59.
ISRM. 1981. Suggested methods for the quantitative descrip-
due to the formation of deep gullies and subsequent tion of discontinuities in rock masses. Rock Characteriza-
undercuts formed by the currents of uncontrolled flow- tion, Testing and Monitoring, ISRM Suggested Methods:
ing rainwater down the hill slope. The second reason 3–45. E.T. Brown (ed.), Published for the Commis-
of sliding is the under cutting of the hill slope at the sion on Testing Methods, International Society of Rock
toe position by the flowing water of Tista river. Due to Mechanics, Pergamon Press.

284
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Inferences from morphological differences in deposits of similar

large rockslides

A.L. Strom
Institute of Geospheres Dynamics, Russian Academy of Sciences, Moscow, Russia

ABSTRACT: Case-by-case comparisons of otherwise similar rockslides and rock avalanches that differ in
morphology can provide reliable and important data for better understanding rockslide motion. Several groups
of rockslide examples with more or less similar failure conditions but different deposit shapes or runouts are
described. A common theme of differences in mode of momentum transfer is offered in explanation of the
observations.

1 INTRODUCTION between the components is defined by a concave ‘sec-

ondary scar’ on the downstream slope of the compact
A number of rockslides having similar failure condi- body (Figure 1-B).
tions, but differing deposit morphologies are described The Northern Karakungey and the Kashkasu rock-
briefly. Events are compared that have similar vol- slides jumped from benches and struck their val-
umes, occur on slopes of similar height and steepness, leys’ bottoms nearly at a right angle. These also
are composed of similar lithologies, and moved over formed compact bodies accompanied by avalanche-
more or less similar terrain, but which formed deposits like tongues. But here, the compact portions have
with differing morphologies and/or have different convex slopes, while their mobile debris portions
runout lengths. Such comparisons help to reveal the gradually thin towards the distal margins (Figure 1-C).
factors influencing rockslide morphology and, thus, It was proposed (Strom 1996, 2006) that these
provide better understanding of their motion. variations in rockslide debris distribution and mor-
phology depends on the shape of the collapsing
slope foot where the initial acceleration ceases and
2 COMPARATIVE CASE STUDIES motion continues due to the momentum of the rock-
slide mass.
2.1 Differing debris distributions along
rock avalanche paths 2.2 Assymmetry of up- and down-valley flow
Significant difference in debris distribution along Rockslide-deposit shape differs also in the up- and
rock avalanche paths can be illustrated with examples down-valley direction when rockslides spread both
from the Kokomeren River valley in the Central Tien up- and down-valley—transverse to the initial debris
Shan. Several rockslides in granite and gneiss, 10 to motion vector. This variability in spreading can be
20 Mm3 in volume had fell from slopes of similar observed at rockslides falling from one valley side and
height and steepness. Nevertheless, they formed rock striking the opposite slope at a nearly right angle to
avalanches with widely different debris distributions form a high natural dam. Some examples are shown
in their transition-deposition zones, forming different in Figure 2.
shaped debris accumulations. It is hypothesized that the extent of the spreading
One of these examples—the Seit rock avalanche— depends on the presence or absence of a tributary val-
entailed the entire rock mass in the avalanche-like ley at the collision zone to act as a ‘‘trap’’ for the rapidly
motion (Figure 1-A). moving rockslide mass. Its presence at a rockslide
In contrast, the Southern Karakungey and the axis prevents transverse debris spreading and leads
Chongsu rock avalanches both formed a compact body to a symmetric dam with steep up- and downstream
at the foot of their source slopes and a secondary slopes (Figure 2-A). If such ‘‘trap’’ is positioned up- or
rock avalanche extending beyond it. Both compact and downstream from the rockslide axis, spreading along
long runout parts include significant portions of the the main valley is much more pronounced just on
collapsed rock mass. In each example, the boundary the side of the blockage farther from the tributary

285
Figure 1. Rock avalanches featuring different debris distri-
bution along their path. A—the Seit rock avalanche ∼15×106
m3 in volume, which debris accumulated at the distal part of
the 3-km long rock avalanche path (R.A.). B—The Chongsu
rock avalanche ∼6 × 106 m3 in volume that formed com-
pact body with well-expressed secondary scar (S.S.) and
1.5-km long rock avalanche (R.A.) that involves about 65% of
debris. C—Aerial and hand-made photographs of the North-
ern Karakungey rock avalanche ∼10×106 m3 in volume that
jumped from a bench and formed a compact blockage with
convex slopes; its portion turned right and traveled ∼0.5 km
downstream (R.A.).

mouth, irrespective of whether is it located on the up-

(Figure 2-B) or downstream side (Figure 2-C).
Figure 2. Large rockslide dams (maps and along-valley
2.3 Runout anomalies cross-sections). A—the symmetric Aksu rockslide dam,
Tien Shan; B—the Usoi rockslide dam, Pamirs, that
Another interesting phenomenon is variation in runout demonstrates distinct downstream spreading while its
distance for rock avalanches originating from the upstream slope is very steep; C—the Djashilkul rockslide
slopes of nearly similar height and steepness. This dam, Tien Shan, with more pronounced upstream spreading
contradicts the ‘‘standard’’ H/L vs. rockslide rather than downstream one.

286
Figure 3. The Chaartash rock avalanches, Tien Shan. A—the Chaartash-2; B—the Chaartash-3, C—the Chaartash-1.
White circle marks narrowing of the valley through which the Chaartash-2 rock avalanche passed.

volume relationship. The latter implies that increas- the slightly inclined intermontane depression bottom
ing rockslide volume results in decreasing H/L ratio (Strom1998). One—the Chaartash-2 rock avalanche
(Scheidegger 1973, Hsu 1975, Shaller 1991). of about 120 × 106 m3 in volume—initiated from the
Such anomaly can be exemplified by two rock 600-m high slopes of a deep gully, so that collapsing
avalanches that descended from the slopes of the masses collided each other at the gully bottom. This
Western Akshiyriak range in the Central Tien Shan. collision formed a secondary rock avalanche, which
Both are composed of granite and moved over passed beyond the narrow gully mouth to travel 7 km

287
Figure 4. Long runout Snake-head rock avalanche. The initial massive rock slope failure at A was accompanied by rock
avalanche that traveled along the narrow gorge about 2.2 kilometers up to B. C—part of the transitional zone where debris of
additional landslides ‘‘fell into’’ the main rock avalanche. Circle marks the funnel-shape narrowing shown on Figure 5.

288
Figure 5. Funnel-shape narrowing after which the initial rock slope failure of the Snake-head rockslide converted into long
runout rock avalanche (see Figure 4).

(!) across the depression on a mean slope of 6◦ –7◦ steep slope onto a valley bottom leads to the formation
(Figure 3-A). At the end of its path it crossed the of a ‘‘primary’’ rock avalanche where most of debris
Kugart River valley and stopped at the opposite bank. mass retains its mobility (see Figure 1-A).
Its total runout is 9 km and H/L ratio is 0.18. The In contrast, a sharp slope-to-foot transition or sig-
second example—the Chaartsh-3 rock avalanche of nificant narrowing of the debris path results in an
about 200 × 106 m3 in volume—descended a 800 m abrupt deceleration of part of the debris so that it
high slope rising just above the unconfined plain, and forms a compact body and, as I assume, the momentum
traveled 4 km across the depression on a mean slope is transferred to the frontal or downvalley portion of
of 8◦ –9◦ (Figure 3-B). Here, the total runout is 4.96 debris, which retains possibility of further motion.
km and the H/L ratio is 0.27. This part moves as a ‘‘secondary’’ rock avalanche
Seemingly, the Chaartash-3 conditions should be (see Figure 1-B). The same phenomenon may explain
more favorable for long runout—there is no confine- the asymmetric along-valley profiles of the large
ment, whereas the Chaartash-2 had to pass through rockslide dams (see Figure 2-B, C). It seems that
a ‘‘bottleneck’’ of a confining valley. Moreover, the friction consumes the momentum of rapidly mov-
Chaartash-3 event had the greater volume and ini- ing debris entering the tributary valley, while after
tial fall height, however the Chaartash-2 rockslide direct impact on an opposite slope, some momentum
travelled farther. transfer occurs, producing transverse spreading and,
in the extreme case, formation of a secondary rock
avalanche.
3 DISCUSSION Momentum transfer from abruptly decelerating
debris to a still moving portion may also explain
The qualitative model was proposed to explain the the abnormally high mobility of the secondary rock
phenomena described above (Strom, 1996, 2006). It avalanches that passed through the funnel-shape nar-
was hypothesize that debris distribution and, corre- rowing in comparison with other types of rock
spondingly, its motion mechanism is governed by the avalanches. Such a ‘‘funnel effect’’ likely took place
shape of the collapsing slope foot, where the ini- at the Chaartash-2 secondary rock avalanche (see
tial acceleration ceases and motion continues due to Figure 3-A) and at the much smaller Snake-head
rockslide-mass inertia. A smooth transition from the rock avalanche (Figures 4 & 5) in the Kokomeren

289
River basin (Central Tien Shan) (Strom et al., 2006). Terrestrial and, probably, Martial (Shaller 1991)
Considering the rather small volume of latter rock- rock avalanches that had moved along the similar
slide (5 × 106 − 7 × 106 m3 only) its 2815 m runout unconfined terrain. The Chaartash-2 event should be
is very long, and its H/L ratio is an abnormally low excluded from this sampling, since it occurred in a
0.12. Since only ∼10% of the collapsed rock mass was significantly different geomorphic situation.
involved in the avalanche-like motion the mobility of Many more rockslide case studies of this type will
this particular part is extraordinary. provide more representative sampling and allow more
Transformation to rock avalanches in rockslides strict selection criteria, thus providing more reliable
that fall from benches and strike the valley bot- relationships between basic parameters.
tom nearly at a right angle could be governed by
another mechanism supporting debris mobility. We
should consider the linear dimensions of the catas- ACKNOWLEDGEMENTS
trophically collapsing slopes in question; these are
large—hundreds of meters, up to kilometer or so—and I want to express my gratitude to anonymous reviewers
their complete descend should last from dozens of for critical analysis of the manuscript and to Mauri
seconds to several minutes. Therefore the first part McSaveney for his kind assistance in improving the
of the moving debris to reach the valley bottom text and useful comments.
could be over-ridden by subsequent debris. This
may result in some type of fluidization of the over-
ridden portion of debris which than spreads from
under the main body as a viscous liquid. Perhaps REFERENCES
the process is analogous to concrete released from a
Eppler, D.B., Fink, J. & Fletcher, R. 1987. Rheologic
dumper. properties and kinematics of emplacement of the Chaos
Jumbles rockfall avalanche, Lassen Volcaic National Park,
California, Journal of Geophysycal Research, B95:
4 CONCLUSIONS 3623–3633.
Hermanns, R.L., Niedermann, S., Garsia, A.V., Gomes J.S. &
Case-by-case comparisons of otherwise similar rock Strecker M.R. 2001. Neotectonics and catasrophic failure
avalanches that differ in morphology can provide of mountain fronts in the southern intra-Andean Plateau,
reliable and important data for better understanding Argentina, Geology 29: 619–623.
rockslide motion. Such qualitative comparisons pro- Hsü, K.J. 1975. Catastrophic debris streams (sturzstroms)
generated by rockfalls. Geological Society of America
vide a basis for predicting future styles of debris Bulletin 86:129–140.
motion and runout from potential large-scale rock- Johnson, B. 1978. Blackhawk landslide, California, U.S.A,
slides, if we can compare their situations with studied In B. Voight (ed), Rockslides and Avalanches, 1, Natural
events of similar geological and, above all, geomorphic Phenomena, 481–504. Amsterdam: Elsevier.
conditions. Shaller, P.J. 1991. Analysis and implications of large Mar-
Besides clarification of some aspects of rockslide- tian and Terrestrial landslides. Ph.D. Thesis. California
motion mechanisms, case-by-case comparative stud- Institute of Technology.
ies allow more strict selection of case studies that Sheidegger, A.E. 1973. On the prediction of the reach and
can be used to derive relationships between various velocity of catastrophic landslides. Rock Mechanics 5:
231–236.
parameters characterizing rockslide size and mobil- Shreve, R.L. 1968. The Blackhawk landslide, Geological
ity. Only those events should be compared quan- Society of America Special Paper 108.
titatively that have occurred under similar ‘basic’ Strom, A.L. 1996. Some morphological types of long-runout
conditions. For example, the Chaartash-3 unconfined rockslides: effect of the relief on their mechanism and on
rock avalanche can be compared with the Blackhawk the rockslide deposits distribution, In: K. Senneset (ed)
(Shreve 1968, Johnson 1978) and the Chaos Jumbles Landslides; Proc. of the Seventh International Symposium
(Eppler, et al. 1987) rock avalanches in California, on Landslides, 1996, Trondheim, Norway: 1977–1982.
the Bayan-Nur rock avalanche in Western Mongolia Rotterdam, Balkema.
(Strom 2005) and 2 other old rock avalanches in Strom, A.L. 1998. Giant ancient rockslides and rock
avalanches in the Tien Shan Mountains, Kyrgyzstan,
the same region at 47◦ 42 36.83 N, 92◦ 35 40.90 E Landslide News. 11: 20–23.
and 47◦ 45 22.64 N, 92◦ 33 33.41 E, with the 1st Strom, A.L. 2005. Gigantic rockslides and rock avalanches
rock avalanche of the Sierra Laguna Blanka in the Central Asian region, In: Senneset, K., Flaate,
complex (Argentina)—(Hermanns et al. 2001), K. & Larsen, J.O. (eds.), Landslides and Avalanches,
with rock avalanche in Chinese Tien Shan at ICFL 2005 Norway : 343–348. Taylor & Francis Group,
42◦ 16 48.95 N, 87◦ 18 41.59 E, and with some other London.

290
Strom, A.L. 2006. Morphology and internal structure of Strom, A.L., Djumabaeva, A.B., Dyikanalieva, J.K.
rockslides and rock avalanches: grounds and constraints & Ormukov, Ch.A. 2006. The Snake-head rock
for their modelling, In S.G. Evans, G. Scarascia Mug- avalanche: rock slope failure caused by horizontal seis-
nozza, A. Strom, and R.L. Hermanns (eds.), Land- mic acceleration, In: Quantitative Geology from Mul-
slides from Massive Rock Slope Failure, NATO Science tiple Sources, Liège, Belgium, September 3–8, 2006.
Series: IV: Earth and Environmental Sciences, 49: CD-ROM. S08–12.
321–346. Springer, Dordrecht.

291
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Movements of a large urban slope in the town of Santa Cruz

do Sul (RGS), Brazil

L.A. Bressani
PPGEC/Federal University of RGS, Porto Alegre, RS, Brazil

R.J.B. Pinheiro
PPGEC/Federal University of Santa Maria, Santa Maria, RS, Brazil

A.V. D. Bica
PPGEC/Federal University of RGS, Porto Alegre, RS, Brazil

C.N. Eisenberger & J.M.D. Soares

PPGEC/Federal University of Santa Maria, Santa Maria, RS, Brazil

ABSTRACT: This paper presents a study of an unstable urban slope situated in the city of Santa Cruz do
Sul, RGS, Brazil. The slope is located in the northern part of the city, where several houses have been affected
by slope movements. Its overall size is around 8 hectares. The slope consists of a layer of colluvium and man
made fill. Instruments for field monitoring have been installed and they showed a relationship between slope
displacements and peaks of piezometric level. Slope stability analyses were carried out using shear strength
parameters from direct shear tests. Factors of safety obtained in the analysis were typically close to one,
presenting some variability. An analysis of building pathologies caused by slope movements was also car-
ried out, using data obtained with site inspection and interviews, which indicated the lateral extension of the
movement.

1 INTRODUCTION 2 GEOLOGY AND SITE INVESTIGATION

The town of Santa Cruz do Sul is located in the central The geology of the North and East parts of the town
part of Rio Grande do Sul State, with 107.000 inhab- is formed by siltstones of the Santa Maria Forma-
itants. The town is surrounded by colluvium slopes, tion at the base, followed by Botucatu sandstone
especially to the North and East. Its average altitude is and basaltic layers of the Serra Geral Formation. In
about 122 m. The region has a sub-tropical climate some places the Botucatu is not present, as in the
with humid and hot summers. The rains have low case described here. The Santa Maria siltstones are
intensity and long durations during winter and high quite red in colour which makes its field identifica-
intensity but short durations during summer. In the tion very easy. The basaltic layers present on the top
period of 1988 to 2003 the accumulated precipitation of the sequence are heterogeneous and with varying
measured in the town during the year varied between thickness. These rocks present an intense vertical and
1419 mm and 2100 mm. horizontal fracturing (Grehs 1976, Wenzel 1996).
The town is known to have problems of slope The Santa Maria Formation is formed basically by
instability since the 70’s (Grehs 1976). This paper clayey-siltstones with a massive aspect, reddish in
describes the study carried out in a region with some colour and with smectite type clays. This Formation
houses damaged at various degrees. By chance, the is widespread in the town on altitudes of 30–100 m.
work was carried out in a period with large rainfall Most of the central portion of the town is located
and the soil mass movements were clearly seen (and on this material. Its permeability is much lower than
measured). the permeability of the basalts and this has important

293
 140 25 kPa

120 50 kPa

100 kPa
100

Shear stress (kPa)

150 kPa
80
200 kPa
60

40

20

0
0 5 10 15
Horizontal displacement (mm)

Figure 2. Results of direct shear tests on colluvium.

significant on grading, plasticity limits and physical

indexes of samples taken from 12 borings.
Direct shear test on colluvium and red siltstone
specimens soaked in water were carried out (5×5 cm e
Figure 1. Typical geotechnical profile (some presence of 10 × 10 cm sizes). The colluvium samples were taken
cobbles at the materials contact, 6–8 m). from 2 positions on the slope and the void ratio varied
between 1.2 and 1.5. Drained triaxial (CID) tests in
the colluvium soil were also carried out on specimens
consequences. Grehs (1976) has indicated that regions of 5 cm in diameter. The results of direct shear tests
where saprolitic soil of this Formation are present have have not shown peak strength on any of the specimens.
slope problems. The colluvium specimens generally maintain a con-
stant level of shear strength after some deformation.
2.1 Site investigations The volumetric variation only presented compressive
behaviour.
A number of SPT borings and manual auger borings The resulting envelopes obtained have shown some
have been made by Pinheiro et al. (2002) to obtain variation but they seem to pass through the origin. The
the geotechnical profile of the region. Some sam- internal friction angle varied from 23◦ to 29◦ , with
ples for laboratory testing were also collected. From c = zero. Figure 2 presents the curves of a set of
these results the geotechnical profile was described direct shear tests carried out on colluvium.
as a deposit of colluvium soil and fill with varying
thickness (3–10 m). The soil can be described as a
sandy silt-clay mixture having boulders of basalt and
sandstone and brown to reddish colours. Underneath
this layer, the red siltstones of Santa Maria Formation 3 FIELD INSTRUMENTATION
were present. Figure 1 shows a typical sounding pro-
file of the region. The change in materials was clearly The instrumentation for monitoring of this slope
seen in the field by the difference in colour and blow was first installed in 2001 and during 2002 other
counts. Presence of cobbles was sometimes noticed at instruments were installed. The monitoring included
the contact. Casagrande piezometers, inclinometer tubes, an auto-
matic pluviometer and physical markers to measure
opening of fissures in buildings and walls. The instal-
2.2 Laboratory tests
lation procedures generally follow recommendations
Pinheiro et al. (2002) and Eisenberger (2003) carried of Dunniclif (1988). Most of the data presented here
out a series of characterization tests with the materials is base on a time scale of 707 days, which cover the
taken from the borings. The scatter in the results was period March 30, 2001 to March 08, 2003.

294
3.1 Automatic pluviometer and piezometers
Four Casagrande type piezometers were installed on
auger borings of 100 mm diameter previously pre-
pared. The piezometers were made of plastic tubes
with sand filter of 60 cm and a bentonite seal on top of
it. One of the piezometers (P4) had an electric trans-
ducer connected to a data acquisition system installed
inside of the tube. This piezometer recorded the pore
water pressure every hour. The other piezometers (P1,
P2 and P3) were read manually by conventional means
(a probe inserted inside of the plastic tube).
An automatic pluviometer was installed in the same
place as one of the inclinometers and close to the
piezometers. The main purpose was to obtain the
rain intensity as real as possible. The system reg-
istered the rains with 1h intervals during the period
of June 24, 2002 to February 15, 2003. The records
collected during this period showed monthly average
values larger than the average of the last 16 years and in
some cases being even larger than the maximum previ-
ously recorded. During October 2002 the precipitation
reached 369 mm and that was the maximum of the his- (b)
torical series (Figure 3). The rains during 2002 have
shown larger variation than normal and the town had a
number of street floodings as a result. The coincidence
of this natural phenomenon with the instrumentation
timing was an unexpected opportunity for the work.
The measurements obtained from the manual
piezometers from May 2001 indicated a similar trend
of pore water pressure in all of them. The piezometers
have shown variation of 2 to 3 m in the water level. No
artesian pressures were observed.

3.2 Inclinometers
The inclinometers tubes installed were positioned
close to the position were most of the damages in
buildings were first observed. The results obtained
Figure 4. Results obtained from inclinometers; (a) incli-
are presented in Figure 4 for the two inclinometers.
nometer I1; (b) inclinometer I2.

The results presented in Figure 4(a) shows a very clear

slip surface at 6 m depth, at the contact colluvium-
siltstone. The accumulated displacement of the soil
mass was of 35 mm during 22 months.
The second tube (I2) was situated above the first
inclinometer and crosses a smaller colluvium deposit.
The results presented in Figure 4(b) have shown a slip
surface at 3 m depth and so well defined as the previ-
ous one. The results show a gradient of displacements
reducing with depth, but the maximum accumulated
displacement measured during the 22 months was
63 mm.
Figure 3. Precipitation records during the period compared The displacements measured with the inclinome-
to historical averages. ters have shown clear periods of acceleration which

295
 the slope studied presented the most severe damages.
Using evidences taken from fissures in a garden paved
floor it was possible to infer movements of 40 cm in
the horizontal direction from 1997 to 2002. The dam-
ages reduce in severity as the residences are situated
further from the center line of the main movement, but
as the damages are a bit scattered in these regions the
limits could not be defined accurately. The total area
with movements has around 7 to 9 hectares.

4 DATA ANALYSIS

The analysis of data obtained gave rise to some

Figure 5. Displacement at the top of the inclinometer tubes interesting findings. The measurements of automatic
versus time. piezometer P4 sometimes gave responses to precipi-
tation which were quite fast in relation to the pre-
cipitation. This peaks were generally correlated with
coincide with periods of intense rains. The larger movements of the slope measured by the inclinometers
displacements were after periods of intense rains, espe- and displacement pins, especially during September
cially on Oct 30, 2001 and Oct 12, 2002. Figure 5 2002.
shows the horizontal displacements measured at the Figure 6 shows the water level (depth below sur-
top of the tubes versus time of monitoring (days mea- face) during the critical period of rain. Comparing it
sured from the beginning of monitoring). It is evident with Figure 5 it can be seen that the peaks of water pres-
that the soil mass have periods of no-displacements sure correlate well with the inclinometers movement
and periods of acceleration. This correlates well with acceleration. Those peaks have not been shown by the
the observed rains. manually read piezometers, as the velocity of increase
and decreased of water level is quite fast (some hours
3.3 Displacement pins in most cases).
Figure 7(a) shows the results of precipitation during
In order to monitor the opening of existing fissures the month of October 2002, one of the wettest in the
caused by the slope movements in buildings, three sets 16 year period and Figure 7(b) shows the measure-
of pins have being installed on damaged structures in ments of water level obtained from piezometer P4. As
the area. The arrangement of pins formed a vertical can be seen, the first precipitation of 45 mm (day 3)
triangle which allowed the measurement of horizontal caused a pronounced peak of 44 cm on it, but the larger
and vertical components of movement across the fis- amount of subsequent rain did not appear to show any
sure. The resolution of readings was of the order of influence, in particular the precipitation of 60 mm on
1 mm. The sections called T1 and T2 were established day 7. So, sometimes it responds quite well to heavy
in one house which showed considerable damages. rains, other times showed almost no response. This
The total horizontal displacement was of 2.4 cm and may be a consequence of the movement of the slope
the vertical displacement of 0,6 cm. The other set, as this may change the internal mass permeability.
12 m away, showed 2.7 cm and 3.9 cm, respectively, in In general, the horizontal displacements measured
the same period of around 6 months (Sept 2002 until at the surface through the pins positioned at the
March 2003).

3.4 Buildings damage

In order to have a better mapping of the movements,
a series of inspections were made on buildings in the
region and interviews with the owners. This allowed
the construction of a house damage map. The damages
were classified by type, origin and severity, following
Alexander (1986) and Chiocchio et al. (1997).
The residences inspected generally follow the same
standard: the structures were made of reinforced con-
crete supported by concrete or metal piles. There was Figure 6. The water pressure measured by piezometer P4
generally some earth moving in the plots. The center of during September and October 2002.

296
 surfaces reproduce quite well the movements
observed in place and also shown by the monitored
displacements.

6 CONCLUSIONS

The area studied was known to have some damages

in the houses but little more was known. The area is
formed by a gentle slope composed by a colluvium
and man-made fill of siltstone with 3 to 10 m depth.
A comprehensive damage survey in the slope buildings
revealed that the damages varied between negligible to
severe (following Alexander, 1986) depending on their
position on the slope (there is a central region were the
worse cases are found).
Borings were opened through the area and a number
of soil samples were taken. The thickness of the soil
mass varied between 3 and 10 m. Tests showed that the
soil has some variability and this affect the shearing
parameters as well.
The monitoring of the slope was carried out
Figure 7. (a) Precipitation during October, 2002; (b) water with piezometers, inclinometers and fissure measure-
depth measured by Piezometer 4 on the same days. ments. The use of an automatic piezometer and plu-
viometers (1 hour interval readings) showed that the
pore-water pressure has a fast response to the rains
after a threshold value. The automatic piezometer
buildings showed the same trend but much larger dis-
showed peaks of up to 1 m (10 kPa) for intense daily
placements than the inclinometers. The reason is that
rains and some response for accumulated rains over
they were installed in the fissures where most of the
3 days periods. The manual readings of the other
movements are concentrated. The inclinometers and
piezometers showed no such peaks (as they were not
the pins showed the larger movements around the day
read at the proper times). The rains which caused peaks
550. It is interesting to note that the peak of water pres-
of pore-water pressures were those above a 50 mm/day
sure on day 625 had effects on inclinometer I2 and
during wet periods.
pin T3 but no significant influence on inclinometer
The monitoring covered a period in which the
I1 and pins T1 and T2. This data confirms other evi-
monthly rains were more intense than the average of
dences that the movement of the slope is a sequence of
the last 16 years. This caused large displacements
small retrogressive movements of individual blocks.
of the mass and the possibility to observe clearly
these movements. The critical period, with larger
inclinometer readings and surface displacements, was
5 STABILITY ANALYSES September and October of 2002 coincident with larger
precipitations. Two inclinometer readings showed
Two representative cross-sections of the slope were (a) a failure well defined in the contact of materials and
analyzed through the SLOPE/W program. The (b) a distortion zone up to 3 m depth. Both instruments
sections were coincident with the areas where scars showed an acceleration of movements in that period.
are visible at the surface. Various simulations were The peaks of pore-water pressures have a good
carried out using the observed pore-water pressures correlation with the movements of the slope as mea-
and measured internal friction angles. In general sured by inclinometers and fissure measurements. The
the factors of safety obtained were close to 1.0 but movements seem to happen in retrogressive manner
the results were quite sensitive to small values of following a temporal sequence.
cohesion. The most critical surfaces were found at Stability analysis carried out on typical sections
the contact between the siltstone and the colluvium showed factors of safety around 1.0 in a number
soil. The critical area pointed out by the software of critical surfaces which were limited in extension.
agreed quite well with the main scar found in the It is interesting to observe that the composition of
region but it was interesting to observe that the anal- these individual surfaces appears to represent well the
ysis showed not one critical surface covering the observed movement of the overall slope. In general
whole length, but a series of smaller surfaces. These the movements are slow (40 cm in 5 years is a good

297
estimate) but with acceleration during the periods of Chiocchio, C., Iovine, G. & Parise, M. 1997. A proposal for
pore-water pressure peaks, as was the case during surveying and classifying landslide damage to buildings in
monitoring. urban areas. Engineering Geology and the Environment,
Further work still in progress has shown that other Rotterdam, pp. 553–558.
areas of the town are also affected by movements of Dunniclif, J. 1988. Geotechnical Instrumentation for Moni-
toring Field Performance. John Wiley & Sons Inc., New
similar magnitudes. York, 577p.
Eisenberger, C.N. 2003. Study of an urban colluvium slope
in Santa Cruz do Sul-RS. Porto Alegre: UFRGS, 2003.
ACKNOWLEDGMENTS 111p. Master dissertation. Escola de Engenharia, UFRGS
(in Portuguese).
L.A. Bressani and A.V.D. Bica have been supported Grehs, S.A. 1976. Geological Mapping of Santa Cruz do
by CNPq. The MSc dissertation of C. Eisenberger was Sul—Basic information for Integrated Planning, ACTA
also supported by CNPq. The municipal local govern- GEOLÓGICA LEOPOLDENSIA, v.1, Universidade Fed-
ment gave support for the field work. CAPES provided eral do Vale do Rio dos Sinos, p. 121–152 (in Portuguese).
Pinheiro, R.J.B., Soares, J.M.D., Bica, A.V.D. Bressani, L.A.
funds for some of the instrumentation. & Eisenberg, C.N. 2002. Geotechnical investigation of an
urban slope in Santa Cruz do Sul—RS. Proc. XII Brazilian
Cong. Soil Mech. Geotech. Engng, ABMS, Sã Paulo,
REFERENCES Vol.,2, 1247–1257 (in Portuguese).
Wenzel, J.A. 1996. Structura, geological mapping of the town
Alexander, D. 1986. Landslide damage to buildings. Environ. of Sant Cruz d Su fo urba planning u U NISC Sant Cruz d
Geol. Water Sci., vol. 8, no. 3, pp. 147–151. Sul. (in Portuguese).

298
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Geotechnical analysis of a complex slope movement in sedimentary

successions of the southern Apennines (Molise, Italy)

D. Calcaterra, D. Di Martire, M. Ramondini & F. Calò

Department of Geotechnical Engineering, Federico II University of Naples, Naples, Italy

M. Parise
National Research Council, IRPI, Bari, Italy

ABSTRACT: The results of a geological and geotechnical study on a complex landslide occurred in Molise,
Italy, are described. The slope movement, a roto-translational slide evolving in an earth flow, was firstly re-
activated in January 2003 by intense rainstorms, causing serious damage to man-made structures and the
evacuation of 15 families from their houses. After a thorough archival research, a geological field survey
was performed, integrated by interpretation of multi-temporal air-photos and by two site investigation and moni-
toring campaigns. Inclinometer measurements revealed some slip surfaces at depths between 5 and 30 metres.
The geotechnical laboratory tests, aimed at characterizing the upper, weathered portion of the sequence and at
comparing it with the unsheared materials, were used for a preliminary back-analysis, to verify the shear strength
mobilized under the stability limit conditions, in view of developing a more suitable kinematic model for the
landslide.

1 INTRODUCTION province. Known as the Colle Lapponi—Piano Ovetta

landslide (CL-PO landslide in the following), after the
The southern Italian Apennines are well known as localities where it develop, the phenomenon is part
areas highly susceptible to landslides (Cotecchia & of an historic landslide, which already affected in the
Melidoro 1974; Pellegrino et al. 2003). The complex past the right slope of the Verrino Torrent. The present
geological setting, where a great variety of differ- study analyzes the CL-PO landslide, illustrating the
ent lithologies are present within the framework of a results of a geological, geomorphologic and geotech-
recently-built mountain chain, makes these territories nical characterization, aimed at better comprehending
among the most affected by slope movements in Italy. the landslide kinematics.
Predominance of clays and marls in the Miocene and
Pliocene formations cause flow-type landslides to be
the most diffuse typology. Usually induced by rainfall,
2 SLOPE MOVEMENTS AT AGNONE
earth flows can even be triggered by human activities,
whilst the role played as triggering factor by earth-
2.1 Historical landslides
quakes has also to be noted, in particular as regards
the reactivation of large phenomena. Analysis of historical data on landslides, as well as
Notwithstanding the limited extension, Molise, the on any other type of natural hazards, has repeatedly
smallest region of central-southern Italy, is severely proved to be a necessary phase in the process of
affected by landslides and erosional processes. In the the hazard assessment. In particular as regards slope
last decade, several events brought to the attention of movements, availability of information on past events,
the public and the scientific community some cases of including location, date of occurrence, and triggering
slope movements, triggered either by seismic shocks factor, is extremely important for a proper compre-
(Bozzano et al. 2004) or by heavy and/or prolonged hension of the recent geomorphologic evolution of the
rainfall events (Corbi et al. 1999; Picarelli & Napoli slopes. Regions where a long history of documenta-
2003). tion is available, as in southern Italy, are particularly
In January 2003, a complex landslide was reac- suitable for this type of research, as demonstrated in
tivated as a consequence of an intense rainstorm in many cases (e.g. Calcaterra & Parise 2001; Calcaterra
the territory of Agnone, a town situated in the Isernia et al. 2003).

299
Figure 1. Plan view of the CL-PO landslide and its evolution in the time-span June 2004–June 2006.

As for many other districts in Molise, landslides in abutments of a viaduct along the State Road no. 86
the municipality of Agnone have been known for long ‘‘Istonia’’.
times. The oldest report of gravity-related phenomena,
in fact, dates back to March 1905, when a slope move-
2.2 The Colle Lapponi—Piano Ovetta landslide
ment was registered in the Vallone S. Nicola, due to
snow-melting and a very intense rainfall period. The The CL-PO landslide (Fig. 1) is a complex slope move-
landslide damaged the bridge of one of the main roads ment, consisting of a roto-translational slide which
entering the historic centre of Agnone (Almagià 1910). evolves in an earth flow. The landslide took place in
It is interesting to note that the locality involved in the the catchment of Vallone S. Nicola, a tributary to the
1905 event is the same dealt with here. Verrino Torrent, which runs west of Agnone. It was
More recently, the Agnone area has been affected by the consequence of the intense pluviometric event that
several landslides, including some large phenomena occurred on January 23–27, 2003, in southern Italy.
covering many hectares. The historical and archival The event was characterized by a maximum rainfall
research on landslides carried out within the frame- height of about 200 mm over 72 hours, with less than
work of a national project (AVI, an acronym which 50 mm registered at Agnone. The landslide caused
stands for Areas Vulnerated in Italy; Guzzetti et al. serious damage to rural buildings and the local road
1994) has shown the occurrence of at least 60 land- network. The main social consequence was the cau-
slides at Agnone and the surrounding territory in the tional evacuation of 15 families from their houses,
time-span 1970–1988. Among these, it is worth to located in the areas directly affected by the slope
remind of the landslide which occurred at Colle Lap- movement or in its immediate surroundings.
poni in February 1994, and which caused a country The geological setting of the landslide area is
road to be interrupted. A few hundreds of metres east characterized by the presence of the Agnone Fly-
of the site where the 2003 event took place, another sch, a structurally complex formation dated to Upper
landslide was triggered on February 1984 (Guadagno Miocene (Vezzani et al. 2004). In the study area, the
et al. 1987). Known as the Fonte Griciatta landslide, lower member of the Agnone Formation is present,
this rotational slide evolving to earth flow affected two made up of marly clays, clayey marls, silty-sandy clays

300
and subordinate sandy levels, with intercalated car-
bonate beds (thickness between 10 cm and 2 m). The
clayey-marly terms show features typical of a weak
rock and a scaly structure; along the scales, weath-
ering evidences can be found, represented by reddish
coatings. The formation is overlain by a weathered
mantle, constituted by brownish-yellowish clays, silty
clays, often iron-oxidated, rich in organic matter and
including heterometric carbonate fragments.
A recent site investigation allowed an in-depth anal- Figure 3. Geological cross-section through the CL-PO
ysis of the lithostratigraphical features of the terrains, landslide.
explored down to 40 m from the ground surface. In
view of the eventual geotechnical slope stability anal-
ysis, four homogenous horizons have been recognized,
dips toward the east and the north-east) the slope move-
of which their characters are as follows (Fig. 2):
ment evolves from rotational to translational slide and
successively into an earth flow.
– Complex A: chaotic, plastic, remoulded greyish to As evident from Figure 1, the landslide has shown
brown clay deposits, directly involved in the CL-PO both a retrogressive and an advancing tendency, which
reactivations. Thickness: 6.70−11.20 m; resulted in an overall increase in the unstable area,
– Complex B: light- to dark grey clays, silty clays, evaluated in about 80,000 m2 from 2004 to 2006,
sandy clays, silty sands. Thickness: 4−12 m; reaching a global value of about 240,000 m2 . From
– Complex C: calcareous levels. Thickness variable September 2005 to June 2006 the foot has progressed
from dm to m; of about 70 m; consequently, the landslide has reached
– Complex D: dark grey, scaly marly clays, clayey a total length of about 1500 m.
marls and marls, which minimum depth is at The CL-PO has completely altered the local hydro-
between 15.30 and 20 m from ground surface. graphic network. In fact, during the main phases of
reactivations, the San Nicola torrent has been repeat-
As regards the relative position of the various com- edly obstructed; in addition, a number of ponds were
plexes, A, B and D are usually found in such sequence noted, partly related to groundwater circulation hosted
from the ground surface downward, while C can be by the uppermost weathered horizons and often coin-
found interbedded with A, B and D. Complex C is ciding, during the rainy periods, with the ground
represented by lithoid, highly permeable beds, and surface. To this regard, it is important to note that the
assumes a fundamental role with respect to groundwa- presence of carbonate rocks in contact with the struc-
ter circulation. Following the above litostratigraphical turally complex formation involved in the landslide
scheme, a cross section has been reconstructed (Fig. 3) provide a huge amount of water from the above. Fur-
where layers belonging to complex C are not visible ther evidence of the active slope morphodynamics are
due to the scale adopted. given by several minor scarps, located upslope of the
The rotational character is shown in the source area main crown of the CL-PO body, and by a number of
by the steeply inclined main scarp and by the overall small flows triggered by the side-slopes undercutting
spoon-like shape. Due to strata attitude (which, even of the San Nicola torrent.
though with several folds and irregularities, mostly A source of major concern is represented by the
toe zone. In fact, a possible further remobilization
of the landslide could rapidly reach the intersection
with the Verrino torrent, which at present is about
350 m apart. In such case, a landslide dam could build
up resulting in adverse consequences to downriver
man-made structures, among which the above cited
viaduct, already damaged by the 1984 Fonte Griciatta
landslide.

3 THE MONITORING CAMPAIGNS

With the aim of identifying the viable remedial mea-

sures, the CL-PO landslide has been under observation
Figure 2. Lithological complexes recognized in the bore- since 2003, through three site and laboratory investi-
holes. gation campaigns (2003, 2004, 2006). The 2004 and

301
2006 campaigns consisted of 26 boreholes, 11 incli- From Figure 5, it is clear that the groundwater lev-
nometers, 14 Casagrande and 1 open-pipe piezome- els have shown variations in a depth range between
ters, optical levelling (58 benchmarks) and 36 samples −19 m and the ground surface. Different reasons can
for geotechnical laboratory analyses. be invoked to explain this situation, such as a marked
Measurements on a network of 58 benchmarks have morphological disorder of the slope, caused by the
shown that horizontal movements at the surface of the CL-PO landslide and hydrogeological heterogeneities
landslide ranged from 20 to 76 cm between April 2006 among the lithological complexes. In particular the
and April 2007, hence showing a displacement rate gauges located in the C complex or near to it show a
between 1.5 and 5.8 cm/month. Such rate is lower than behavior independent by the rainfalls measured at the
that empirically evaluated during the March 14–15, local raingauge, probably depending on the upslope
2004 reactivation, when the bridge over the San Nicola connection of these levels to the carbonate aquifer.
torrent was overtopped; on that occasion velocities
between 0.5 and 1 m/hr were calculated. Compar-
ing two series of air-photos, the landslide toe showed 4 GEOTECHNICAL DATA
a total displacement of about 280 m over the June
2004–September 2005 period, corresponding to an The geotechnical characterization of the CL-PO ter-
average rate of 18.6 m/month. Field surveys in the rains derives from the re-interpretation of the 2004
toe zone allowed to recognize a further downslope and 2006 laboratory tests, which, in turn, can be use-
extension of the landslide of about 70 m between fully compared with the literature data available for
September 2005 and June 2006 (Fig. 1), corresponding the Agnone Fmn. (Cotecchia et al. 1977; Guadagno
to 7.7 m/month. et al. 1987). In particular, Guadagno et al. (1987),
The borehole inclinometers of the 2006 campaign studying the Fonte Griciatta landslide, geotechnically
(Figure 1) have recorded, from January 2006 to Jan- characterized the clayey terms of the formation, dis-
uary 2007, similar rates of movement, with maximum tinguishing intact clays from the mudslide weathered
displacements of about 30 cm. The depths of the major clays.
active shear surfaces range from 7 m to about 27 m. Following the already recalled subdivision, quite
Accordingly, a longitudinal profile of the landslide has different parameters can be referred to the A, B, and D
been reconstructed (Fig. 4). From these data, it is esti- complexes (Table 1 and Fig. 6). In the following, com-
mated that the active landslide has a volume of about plex D parameters are derived from Cotecchia et al.
3.5 million m3 . (1977) and Guadagno et al. (1987).
Complex A is composed by lapideous fragments
and shales plunged into a softened, fine grained
matrix, consisting of a mix of clay and thin hard lumps.
The clay fraction varies from 48% to 63%; dry density
ranges between 14.0 and 16.5 kN/m3 , with a saturation
ratio always high, in agreement with the water table

Figure 4. Slip surface by inclinometer measurements. Table 1. Mean values of the geotechnical parameters of the
complexes A, B, D.

Parameter Complex A Complex B Complex D

29/10/06

28/11/06

28/12/06
31/7/06

30/8/06

29/9/06

27/1/07

26/2/07

28/3/07
1/2/06

3/3/06

2/4/06

2/5/06

1/6/06

1/7/06

3 60
γ (kN/m3 ) 19.4 19.7 22.5
γd (kN/m3 ) 15.2 16.0 21.2
γsat (kN/m3 )
0
50 19.6 20.1 22.9
-3
S1 γs (kN/m3 ) 27.2 27.4 27.0
40
S2 w (-) 0.27 0.23 0.10
Daily rainfall (mm)

-6
S3 (9.0 m)
e (-) 0.79 0.71 0.28
Depth (m)

-9 30
S3 (19.0 m)

S4
n (-) 0.44 0.41 0.22
-12
20 Sr (-) 0.93 0.89 0.99
-15 wl (-) 0.59 0.62 0.44
-18
10 wp (-) 0.26 0.26 0.20
lp (-) 0.33 0.35 0.24
-21 0
φ  (◦ ) 19 23 22
c (kPa) 20 28 59
 (◦ )
φres 17 22 9
Figure 5. Groundwater levels compared with rainfall mea-  (kPa)
cres 0 0 0
surements.

302
 d (kN/m )
3
w (%) average equal to 22%, showing a slightly higher value
0
12 14 16 18 20 22 24
0
0 5 10 15 20 25 30 35 40
towards the top of the horizon. Low porosity also
5 5
explains the high Sr values, even though the water
10 10
content is quite low (Table 1). While plasticity and
activity values are comparable to those belonging to
Depth (m)

Depth (m)
15 15
complex B, the shear strength results higher: peak fric-
tion angle is in fact near 22◦ with a cohesion around
20 20

Soil A
25 25
Soil B
Soil D 60 kPa.
30 30
The significant difference between the peak stren-
n (%) Sr (%) gth values in the three complexes disappears in the

residual strength; in fact the friction angles φres
10 20 30 40 50 60 50 60 70 80 90 100
0 0 are
5 5 near 17◦ , 22◦ and 9◦ referring to complex A, B and D
10 10 respectively. These results are probably due to better
Depth (m)
Depth (m)

15 15
mechanic characteristics of the complex B if compared
20 20
to complex A. The lower value measured for complex
25 25
D is to correlate with its scaly structure, characterized
by low friction strength between each scale after the
30 30
rupture of the cohesive bond.
Figure 6. Physical properties of the investigated complexes.

5 SLOPE STABILITY BACK-ANALYSIS

position, near the ground surface. The porosity val-
ues vary between 40% near the bottom of the complex Aimed at defining a suitable kinematic model, the
and 47% in its uppermost portion. As regards plastic- best parameters governing the CL-PO landslide have
ity, complex A falls in the field of high plasticity clays, been sought. Hence, a preliminary stability back-
characterized by low activity. These values are based analysis has been performed, adopting the classical
on the measurements of the ‘‘overall’’ water content; infinite slope model, that can be usefully applied to the
however, these results depend on the low water content case-study, considering the section of the longitudinal
of the lithorelicts included in the matrix. In fact, when profile (depth/length 1–Fig. 4).
the lithic inclusions are separated from the remaining In the back-analysis the original pre-landslide pro-
part of the matrix, a higher value of the water content file has been considered, subdividing the slope into
and of the liquid index can be obtained (Picarelli 1993). segments with constant slope angle; accordingly, the
The shear strength, detected by means of direct shear safety factor (SF) has been calculated for each seg-
tests, is characterized by a value of the peak friction ment. By doing so, slope sectors characterized by a
angle near to 19◦ and a cohesion around 20 kPa. SF lower than 1 have been defined, for different posi-
Complex B has a grain distribution as clays, silty tions of the water table, considered with a flow parallel
clays, sandy clays, silty sands. with a clay fraction to the soil profile. Three hypotheses have been made
varying between 50% and 55%, and a silt fraction as concerns the surface of rupture, placed at 5, 10 and
slightly higher than complex A. Physical properties 20 m below the ground surface respectively. Both peak
show a limited variability: dry density is between 15.0 and residual shear strength have been considered.
and 16.5 kN/m3 while Sr varies from 85% to 95%. Figure 7 shows that the upper portion of the slope is
Such values, apparently contrasting with the bound- under limit equilibrium conditions, while downslope
ary hydraulic conditions, depend on the structure of SF is higher than 1. It can be also noted that, to explain
the complex and its sandy fraction. Porosity is lower the observed displacements, residual strength should
than in complex A, with n values from 40% to 44%. have been available along some slope segments, since
In terms of plasticity, complex B can be referred to for ϕ = ϕp FS is always higher that 1. As regards the
medium plasticity clays, again showing a low activity. groundwater table, the critical conditions for the slope
The shear strength offers a value of the peak friction are reached when Hw is close to the ground surface
angle near 23◦ and a cohesion value around 28 kPa. for ϕ = ϕp . On the other hand, when ϕres
is taken into
Complex D has a grain distribution as marly clays account, SF is close to 1 when Hw is between 2 and
and clayey marls, with a clay fraction varying between 8 m below the ground surface.
25% and 35%, and a silty fraction varying between These results, referred to the measured Hw piezo-
50% and 60% (Cotecchia et al. 1977; Guadagno et al. metric data (Fig. 5), show some interesting behaviors
1987). Physical properties are significantly differ- (Fig. 8). In the upper zone of the slope, in fact, the SF
ent from the upper complexes: dry density is in fact values are lower than or close to 1 even to peak friction,
definitely higher (20.0–22.0 kN/m3 ), porosity n is on at variance with the actual displacement regime; such

303
 In fact, following the Cruden & Varnes (1996)
velocity scale it can be stated that the CL-PO land-
slide has reached values of displacements typical of
a rapid movement (10−4 m/s) during the March 2004
crisis, slowing down to moderate velocities (10−6 m/s)
in the June 2004–June 2006 period. In the last months
the optical levelling indicates a further reduction of the
displacements with a mean rate between 10−8 and
10−9 m/s.
As for the geotechnical characterization, the CL-PO
data are in good agreement with the literature, which
indicates the weathering processes as responsible for a
number of mudslides (Chandler 1972; Taylor & Cripps
1987; Cafaro & Cotecchia 2001; Picarelli et al. 2005).
In the CL-PO case-study, the presence of three overly-
ing soil complexes provided with different properties
supports the importance of the weathering processes
as a key factor in predisposing the local instability.
To this respect, Guadagno et al. (1987), studying the
Figure 7. SF variations along the slope as a function of the Fonte Griciatta landslide, recognized two main com-
groundwater table position (surface of rupture = 20 m). a = plexes, distinguishing intact clays from weathered
peak strength; b = residual strength. clays, based on significant differences between the two
complexes.
The preliminary slope stability back-analysis high-
lighted the complex behavior of the CL-PO slope,
which can be defined as a ‘‘long-extension’’ slope
(Pellegrino et al. 2003; Picarelli & Russo 2004), char-
acterized by a length in the order of several hundred
metres and, consequently, by a high length-thickness
ratio. In these kinds of slopes it is very difficult to
hypothesize the occurrence of the same geotechnical
characteristics and the same behavior along the whole
slip surface. It will be therefore necessary to use more
Figure 8. SF variations along the slope (surface of rupture = complex geotechnical models (i.e. constitutive models
20 m) in relation to the piezometer data. a = peak strength;
b = residual strength.
including creep behavior) to perform a best-fit of the
slope evolution for variations of the rain-water level
regime.

evidence has probably to be correlated to some recon-

solidation phenomena developed in this zone after the
landslide occurrence. On the other hand, in the low- REFERENCES
ermost segments of the slope, the SF values are quite
Almagià, R. 1910. Studi geografici sulle frane in Italia.
high with residual strength too: it can therefore be
L’Appennino centrale e meridionale. Conclusioni gener-
hypothesized that in this zone, during its occurrence, ali. Mem. Soc. Geogr. It. 14 (2): 1–431.
the landslide has changed its rheological features due Bozzano, F., Martino, S., Naso, G., Prestininzi, A., Romeo,
to its velocity, evolving from a slide to a flow. R.W. & Scarascia Mugnozza, G. 2004. The large Salcito
landslide triggered by the 2002 Molise, Italy earthquake.
Earthquake Spectra 20 (2): 1–11.
Cafaro, F. & Cotecchia, F. 2001. Structure degradation and
6 DISCUSSION AND CONCLUSIONS changes in the mechanical behaviour of a stiff clay due to
weathering. Géotechnique 51 (5): 441–453.
Evidence from archival sources, field survey, air- Calcaterra, D. & Parise, M. 2001. The contribution of his-
torical information in the assessment of landslide hazard.
photo interpretation and monitoring campaigns clearly In T. Glade, A. Albini & F. Frances (eds), The use of
demonstrate that the 2003–2007 CL-PO landslide is historical data in natural hazard assessment: 201–217.
the reactivation of a pre-existing mass movement, Dordrecht: Kluwer Acad. Publ.
known since 1905 at least, showing a behavior which Calcaterra, D., Parise, M. & Palma, B. 2003. Combining
can be ascribed to an intermittent kinematics. historical and geological data for the assessment of the

304
 landslide hazard: a case study from Campania, Italy. Nat. Pellegrino, A., Picarelli, L. & Urciuoli, G. 2003. Experiences
Hazards and Earth System Sc. 3 (1–2): 3–16. of mudslides in Italy. Proc. Int. Works. on Occurrence and
Chandler, R.J. 1972. Lias clay: weathering processes and their Mechanisms of Flow-like Landslides in Natural Slopes
effects on shear strength. Géotechnique 22: 403–431. and Earthfills, Sorrento: 191–206. Bologna: Pàtron
Corbi, I., De Vita, P., Guida, D., Guida, M., Lanzara, Editore.
R. & Vallario, A. 1999. Evoluzione geomorfologica a Picarelli, L. 1993. Structure and properties of clay shales in-
medio termine del vallone in località Covatta (Bacino del volved in earthflows. Proc. Intern. Symp. on The Geotech-
Fiume Biferno, Molise). Geografia Fisica e Dinamica nical Engineering of Hard Soils-Soft Rocks, Athens:
Quaternaria 22: 115–128. 2009–2019. Rotterdam: Balkema.
Cotecchia, V. & Melidoro, G. 1974. Some principal geolog- Picarelli, L. & Napoli, V. 2003. Some features of two large
ical aspects of the landslides of southern Italy. Bull. Int. earthflows in intensely fissured tectonized clay shales
Ass. Eng. Geology 9: 23–32. and criteria for risk mitigation. Proc. Int. Conf. on Fast
Cotecchia, V., Monterisi, L., Salvemini, A., Spilotro, G. & Slope Movements. Prediction and Prevention for Risk
Trisorio Liuzzi, G. 1977. Geolithological, structural and Mitigation, Naples: 431–438. Bologna: Pàtron Editore.
geotechnical aspects of some arenaceous-marly forma- Picarelli, L. & Russo, C. 2004. Remarks on the mechanics
tions cropping out in Central-Southern Apennines. Proc. of slow active landslides and the interaction with man-
Int. Symp. on Structurally Complex Formations, Capri: made works. Proc. 9th Intern. Symp. on Landslides, Rio
71–78. de Janeiro, vol. 2: 1141–1176. Rotterdam: Balkema.
Cruden, D.M. & Varnes, D.J. 1996. Landslide types and pro- Picarelli, L., Urciuoli, G., Ramondini, M. & Comegna, L.
cesses. In: A.K. Turner & R.L. Schuster (eds), Landslides. 2005. Main features of mudslides in tectonised highly
Investigation and mitigation. Transp. Res. Board, spec. fissured clay shales. Landslides 2: 15–30.
rep. 247: 129–177. Nat. Acad. Press, Washington, D.C. Taylor, R.K. & Cripps, J.C. 1987. Weathering effects: slopes
Guadagno, F.M., Palmieri, M., Siviero, V. & Vallario, A. in mudrocks and overconsolidated clays. In M.G. Ander-
1987. La frana del febbraio 1984 in località Fonte Griciatta son & K.S. Richards (eds), Slope Stability: Geotechnical
nel comune di Agnone (Isernia). Mem. Soc. Geol. It. 37: Engineering and Geomorphology: 405–445. New York:
127–134. John Wiley & Sons.
Guzzetti, F., Cardinali, M. & Reichenbach, P. 1994. The AVI Vezzani, L., Ghisetti, F. & Festa, A. 2004. Carta geologica
Project: a bibliographical and archive inventory of land- del Molise. Firenze: S.EL.CA.
slides and floods in Italy. Environ. Management 18 (4):
623–633.

305
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Application of surface wave and micro-tremor survey in landslide

investigation in the Three Gorges reservoir area

Ailan Che & Xianqi Luo

China Three Gorges University Key Laboratory of Geological Hazards on Three Gorges Reservoir Area,
Ministry of Education, China

Shaokong Feng
Chuo Kaihatsu Corporation, Tokyo, Japan

Oda Yoshiya
Xi’an Jiaotong University, Xi’an, China

ABSTRACT: The surface wave and micro-tremor survey are new methods of the engineering geophysical
investigation. They have been widely used in recent years because they are of low cost and easy operation and
they give the dynamic property of the ground, i.e. the shear wave velocity. This paper discusses the application
of micro-tremor and surface wave in the Qianjiangping and Shuping landslides of the Three Gorges reservoir
area. For surface wave survey, we analyzed the dispersion of the recorded data and then estimated the shear wave
velocity structure by phase velocity inversion. For micro-tremor survey, we observed the ground micro-tremor
signals, and analyzed them by fast Fourier transform method (FFT). The soil response to frequency of ground
micro-tremor is revealed and the function with frequency-dependence and frequency-selection of micro-tremor
for different foundation strata are investigated. The vertical to horizontal spectra ratio (H/V, Nakamura technique)
of micro-tremor observed at the surface was used to evaluate the site’s predominant period. All of these have
provided a more dependable basis for the landslide material classification and for analysis of these hazardous
phenomena. The experimental foundation and the deduction process of the method were described in detail.
Comparison of the geophysical investigation results with the geological data shows their good agreement, thus
proving the applicability of the methods in landslide study.

1 INTRODUCTION Gorges reservoir, and this has prompted geotechnical

researchers to pay special attention to the problem.
The Three Gorges dam on the Yangtze River in China It is well known that the ground structure is great
is the largest hydroelectric project in the world. After important for geotechnical engineering. However, the
the first impoundment in June 2003, many landslides study by boreholes is too expensive, especially in com-
in her upstream area occurred or were reactivated. For plex geological conditions, which is the typical case of
example, in the early morning at 00:20 July 13, 2003, the Three Gorges area landslides. The Surface wave
the Qianjiangping landslide occurred on the left bank survey and micro tremor observation are one of the
of the Qinggan River, a tributary of the Yangtze River, most convenient methods to investigate the dynamic
three kilometers away from its confluence with the characteristics of the surface ground (Yamanaka, et al.,
Yangtze mainstream (Figure 1). 14 people were killed 1996, Bard, 1998). They were applied for investigation
and another 10 listed as missing (Wang et al. 2004). of Qianjiangping— rockslide and Shuping landslide.
Direct economic losses caused by the landslide were
about 7 million USD, and the asset value of Shazhenxi
2 GEOLOGICAL CONDITIONS
town was reduced by 40%. The Shuping landslide that
IN RESEARCH AREAS
is 3.5 km from the previous one is one of the most
active landslides at the south bank of Yangtse River
2.1 The Qianjiangping landslide
near the Shazhenxi town (see Figure 1) (Gan et al.
2004). It is anticipated that the frequency of landslides The Qianjiangping rockslide is located about 50 km
will increase due to the impoundment of the Three upstream from the Three Gorges Dam, China

307
Figure 1. Location of the Qianjiangping landslide and Shuping landslide in Three Gorges area.

(Figure 2) (Wang et al. 2004). The landslide is a

tongue-shaped in plan, 1200 m long, and 1000 m wide.
It moved about 250 m in the main sliding direction
S45◦ E. The average thickness of the sliding mass was
about 20 m, thinner in the upper part and thicker at
the lower part. It was a typical rockslide in weathered
shale and sandstone, a block of which slipped down
along the bedding plane. The exposed sliding surface
at the upper part was very straight, sub-parallel to the
bedrock strata.
The site is composed of quartzo-feldspathic sand-
stone, fine sandstone with carbonaceous siltstone,
siltstone with mudstone, and silty mudstone of the
Shazhenxi group of Late Triassic age. Three units can
be roughly found in the cross-section — the slide mass, Figure 2. The Qianjiangping landslide.
sliding zone and the bedrock, respectively, as shown
in Table 1 (Luo, et al., 2001).
Table 1. Physico-mechanical parameters of main units.

Natural
2.2 The Shuping landslide Specific water Natural Void
Name gravity content density ratio
The Shuping reactivated landslide occurred in shale
Serial of soil
and sandstone of Tertiary period. As soon as the first
number -layers (Gs) (%) (g/cm3 ) (eo )
impoundment of the Three Gorges reservoir was con-
ducted, serious deformation occurred in the Shuping 1 Slip mass 2.63 19.89 1.63 0.93
landslide. It formed a large dangerous factor for the 2 Slip band 2.70 15.76 2.02 0.55
local residents and the shipment in the main stream of
the Yangtze River (Gan et al. 2004).
The Shuping landslide is an old landslide composed
of two blocks. The sliding mass consists of red muddy
debris of old landslide, and of sandy mudstone, muddy 3 OBSERVATIONS IN RESEARCH AREAS
siltstone of Badong group of Triassic period (Wang,
et al., 2005). Three units can be roughly selected 3.1 Rayleigh wave prospecting
in the cross-section — loam with gravel, clay and At Qianjiangping landslide, the surface wave sur-
sandy clay with gravel, respectively, as shown in vey is conducted from SK4 boring hole, which are
Table 2. about 360 m away from the sliding cliff (Figure 4).

308
Table 2. Physico-mechanical parameters of main units.

Natural
Specific water Natural Void
Name gravity content density ratio
Serial of soil
number -layers (Gs) (%) (g/cm3 ) (eo )

1 Loam
with gravel 2.63 19.89 1.63 0.93
2 Clay 2.70 15.76 2.02 0.55
3 Sandy clay
with gravel 2.70 15.76 2.02 0.55

Figure 3. Shuping landslide consisting of two blocks at the

main stream of the Three Gorges area.

At Shuping landslide, the surface wave survey is con-

ducted from ZG86 GPS point on II-II line of Block-A,
its elevation is about 355 m downward Shahuang high
way (Figure 5). The data are acquired on a 58-m survey
line with a receiver interval of 2 m for each site. Sledge- Figure 4. Measurement points at Qianjiangping landslide.
hammer is used as a vertical source for data acquisi-
tion, and five shot gathers with source offsets for 4 m,
8 m, 12 m, 16 m and 20 m were recorded. Because 4 ANALYSES AND RESULTS
of limited space, source is only placed on one end of
the survey line. Data were acquired with 1 ms sample 4.1 Rayleigh wave prospecting
rate, and total of 2048 measurements are recorded. For dispersion analysis, first, the cross-spectrum of
each trace with a reference trace is calculated to form
a cross-spectrum gather for each shot location, and
3.2 Micro tremor observations
these are then stacked over the five shots to improve
The micro tremor observations were conducted at both the signal-to-noise level.
sites — from SK4 borehole at the Qianjiangping land- Then, in order to construct a 2D profile over the
slide (Figure 4) and from ZG86 GPS point on II-II line line, spatial windows of 12 traces are analyzed and
of Block-A of the Shuping landslide (Figure 5). The moved along at 2 m intervals, with the nominal posi-
total length of survey line for the latter is about 200 m tion taken as the center of the trace window. Figure 6
with an interval of about 10 m, thus summing up to 40 is the example of the dispersion analysis. These are
observation points. The data were recorded for 5 min- interpreted as the fundamental and higher modes. As
utes with 1 ms sample interval. The records of the hori- the figure shows, we can clearly see the higher-mode
zontal and vertical components of short-period micro dispersion curves, as well as the fundamental-mode
tremors are obtained using a three components high- curve.
sensitive seismometer, which has a one second natural Finally, we estimate the Vs structure by inverting the
period. The potential noise sources such as machinery, phase velocity data (Feng et al., 2005). Figure 7 shows
vehicles traffic or pedestrians, near the seismometer the borehole log results and the estimated 2D shear-
are avoided during the measurement time. wave velocity model from the phase velocity inversion.

309
Figure 5. Measurement points at Shuping landslide.
Figure 6. Example of the dispersion analysis at test site.

The model given by the multi-mode inversion is in bet-

ter agreement with soil types. At the Qiangjiangping amplitude spectra, spectra ratio and relative variation
landslide Vs section (Figure 7a), there are 3 main shear amplitude are computed. The spectra ratio of vertical
wave velocity (Vs) boundaries, corresponding to the component to the horizontal one (H/V, Nakamura tech-
main geological boundaries according to the drilling nique) of micro tremor observed at the surface ground
core. Especially, the boundary between Vs = 600 m/s are widely used to evaluate the site dynamic response
and Vs = 800 m/s units on the Vs section, agrees (Nakamura, 1989), which is considered the transfer
very well with the slide surface. At the Shuping land- function Sr = SHs /SHB = SHs /SVs .
slide (Figure 7b), the Vs section shows 2 clear velocity In this study, the spectral ratio (H/V— average type
boundaries inferring that the slide mass consists of suggested by Japanese Education Ministry was used)
3 layers, which shear wave velocities (Vs) are 400 m/s, from the five sets of 2048 digital data, proposed by
600 m/s, and 800 m/s, respectively. Nakamura was used to identify the predominant fre-
quency (fp : Hz) of the research ground (Iwatate, et al.,
1996). A sample H/V of micro tremor measurements
is shown as Figure 8.
4.2 Micro tremor observations
In order to verify the results obtained from the micro
The micro tremor is a kind of geophysics informa- tremor data, ground structures based on the actual
tion with abundant intension, which is related to borehole logs (SK-4 and SPZK-1) were revealed. One-
site structure, corresponding to engineering geolo- dimensional seismic response analyses of horizontally
gical condition. The frequency characteristics of micro layered soils using Multiple Refraction Theory were
tremor can be obtained by spectral analyses of its performed (Enomoto, et al. 2002). The values of shear-
signals, which can be used to probe the dynamic wave velocity were determined from the Rayleigh
characteristics of a studying area ground. From the wave prospecting. The transfer functions of the model
recorded data of micro tremor measurements, five ground were calculated. Compared with the predom-
sets of 2048 digital data at lower noise periods are inant frequency values from H/V of micro-tremor
selected to use for FFT analysis. The velocity Fourier measurements and analytical results, it shows a good

310
 Figure 8. Transfer function H/V of micro tremor measure-
Figure 7. 2D shear-wave velocity model from the phase ments.
velocity inversion compared with boring logs.

agreement (Figure 9). Therefore, it can be consid-

ered that the dynamic characteristics of the ground
in the research area can be evaluated by the spectral
ratios (H/V) from micro-tremor measurements. The
predominant frequencies obtained from the Qianjiang-
ping ground model clearly show the effect of slip band
that are relatively identical to that as obtained from the
micro tremor results (Figure 9a).
Figure 10 shows the H/V of all the measurement
points. The result shows that the predominant frequen-
cies obtained from the micro tremor data show clear
three peaks at 2 Hz, 4 Hz and 7 Hz. Compared with
the predominant frequency values from H/V of micro
tremor measurements and analytical results (Figure 9),
it shows a response of the sliding zone in the Qianjiang-
ping landslide. The change of predominant period of
ground micro tremor is closely related to the forma-
tion of site ground structure, and mutually corresponds
to the change of engineering geological conditions
of the site. Therefore, it can be considered that the
dynamic characteristics of the ground can be evalu-
ated by the spectral ratios (H/V) from micro tremor Figure 9. Compared with the transfer function from H/V of
measurements. micro-tremor measurements and analytical results.

311
 4. The predominant frequencies obtained from the
Qianjiangping clearly show the effect of slip band in
both analytical results and the micro tremor results.
The change of predominant period of ground micro
tremor is closely related to the formation of site
ground structure, and mutually corresponds to the
change of engineering geological conditions of the
site. So it can be considered that the ground struc-
ture in the landslide can be identified by the micro
tremor.

REFERENCES

Wang, F.W., Zhang, Y.M., Huo, Z.Y., Matsumoto T. & Huang,

B.L. 2004. The July 14, 2003 Qianjiangping Landslide,
Three Gorge Reservoir, China, Landslide, 1(2):157–162.
Gan, Y.B., Sun, R.X., Zhang, Y.Q. & Liao, S.Y. 2004. Urgent
investigation report on the Shuping landslide in Shazhenxi
town, Zugui Country, Hubei Province (in Chinese).
Bard, P. Y. 1998. Micro tremors measurements, a tool for site
effect estimation. ESG 98 symposium.
Yamanaka, H. & Ishida, H. 1996. Application of genetic
algorithm to an inversion of surface-wave dispersion
Figure 10. Predominant frequencies from H/V of micro data: Bulletin of the Seismological Society of America,
tremor measurements. 86:436–444.
Luo Xianqi, Jiang Qinghui, Ge Xiurun & Liu Sifeng, 2001.
Study of the stability of Qianjiangping slope in Huan-
5 CONCLUSIONS glashi slope group, Chinese Journal of Rock Mechanics
and Engineering, 21(1):29–33.
The surface wave survey and micro tremor observa- Wang Fawu et al. 2007. Landslide caused by water level
tion are conducted on the ground of the typical sites changes in Three Gorges Reservoir Area, China, Chi-
in the Qianjiangping landslide and Shuping landslide. nese Journal of Rock Mechanics and Engineering, Vol. 26,
The phase velocity of surface wave is calculated from No. 3, pp. 509–517.
Fawu Wang, Gonghui Wang, Kyoji Sassa, Kiminori Araiba,
the observed surface wave survey, and the subsurface
Atsuo Takeuchi, Yeming Zhang, Zhitao Huo, Xuanming
shear wave velocity structure is estimated by phase peng & Weiqun Jin. 2005. Deformation monitoring and
velocity inversion. The microtremors are observed the exploration on Shuping landslide induced by inm-
simultaneously on the ground surface. To study the poundment of the Three Gorge Reservoir, China, Annuals
application of this method to landslide geotechni- of disas. Prev. Res. Inst. Kyoto Univ., No. 48B.
cal investigation, a series of field experiments and Shaokong Feng, Takeshi Sugiyama & Hiroaki Yamanaka,
analyses are carried out. The main results are as 2005. Effectiveness of multi-mode surface wave inversion
follows. in shallow engineering site investigations, Butsuri-Tansa
(Vol. 58, No. 1), Mulli-Tamsa (Vol. 8, No. 1), Exploration
1. The Vs structure from the phase velocity inversion Geophysics 36, 26–33.
is in good agreement with soil structure. Nakamura, Y. 1989. A method for dynamic characteristics
2. In Qiangjiangping landslide, the result shows that estimation of subsurface using micro tremor on the ground
there are 3 main shear wave velocity (Vs) bound- surface. OR or RTR1, 30, 25–33.
aries on the Vs section, corresponds to the main Iwatate, T., Akira, O. & Koji, A., 1996. Surface Ground
Motion Characteristics of Zushi-Site. Eleventh World
geological boundaries of the drilling core. Espe-
Conference on Earthquake Engineering (11 WCEE),
cially, the boundary between Vs = 600 m/s and Acapulco, MEXICO, June 23–28.
Vs = 800 m/s on the Vs section, agrees very well Enomoto, T. & Iwatate, T. 2002. Site-effects evaluation
with the slide surface. H/V spectra comparing micro-tremor with strong motion
3. In Shuping landslide, the Vs section shows 2 records observed at ground surface and basement using
clear velocity boundaries infers that the slide mass borehole. 12th European Conference on Earthquake
consists of 3 layers. Engineering.

312
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

A case study for the landslide-induced catastrophic hazards

in Taiwan Tuchang Tribute

C.Y. Chen
National Chiayi U., Dept. of Civil & Water Resources Eng., Chiayi City, Taiwan, China

W.C. Lee
National Science & Technology Center for Disaster Reduction, Sindian City, Taipei County, Taiwan, China

ABSTRACT: A landslide-induced flow failure caused 15 residents dead, and 24 houses buried during a
Typhoon Aere in 2004. The landslide mechanism by a GIS spatial analysis was studied for clarifying the
reasons attributed to the catastrophic hazards for geological, hydrological, and topographical characteristics.
The evacuation and rescue response for the hazards are being discussed for enhancing the emergency response
in mountainous areas during harsh climatic conditions. After analysis, it was later found that the landslide was
associated with post-seismic behaviour, torrential rainfalls, steep slopes and unfavorable geological conditions.
More attention to the earthquake-induced slopeland ground surface cracking is needed during post-seismic
rainfall events.

1 INTRODUCTION

Landslides have frequently occurred during recent

years with the visibly abnormal climate change. One
of these catastrophic landslide hazards occurred in
the Leyte Island, Philippines, on 17 Feb., 2006. The
rainfall induced a giant landslide which caused more
than 1000 people and 300 houses to be buried in
the mass of debris. In Taiwan, the climate-change
induced landslide increased after the M7.6 Chi-Chi
earthquake in 1999. The landslide hazard in Tuchang
Tribute was speculated as having occurred on 25 Aug.,
9:00 AM, 2004, followed by a landslide dam formed Figure 1. Site location of the Tuchang Tribute in Wufeng
Village, Hsinchu County.
and breached at 10:00 AM according to the commu-
nication record from the buried police station. There
were 15 residents dead and 24 houses buried by the
situated in a mountainous area with an elevation about
landslide masses, including 3 policemen and the police
780 m. The tribute is a gate to the high mountainous
station. More attention is needed for the reaction
area and a police station for personnel management.
of historical landslides induced by earthquakes after
torrential rainfalls in the mountainous areas.
3 LANDSLIDE MAGNITUDE

2 SITE LOCATION The landslide covered an area of 0.58 km2 , with a

length of 490 m, and a base width of 200 m. The
The Tuchang Tribute is located at Wufeng Village, landslide area included in the affected area was about
Hsinchu County, northern Taiwan (Figure 1). The trib- 1.5 km2 by the measure of the post-landslide aerial
ute has a population of approximately 100 people and photo, shown in Figure 2. The estimated landslide vol-
more than 20 buildings. The landslide site is an imme- ume was about 1.0 million m3 . Figure 3 depicts the
diate neighbor to the center of the tribute. The tribute is buried area of the tribute by the debris masses, which

313
 4.1 Historical hazards
There were 77 historical landslides (source: Soil &
Water Conservation Bureau, called SWCB hereinafter,
taken in 2003 by Spot 5 Satellite), 22 potential debris
flow torrents, as show in Figure 5 (source: SWCB,
2003). There were 21 debris flow hazard spots in
the village from the Typhoon Area (source: NCDR,
http://www.ncdr.nat.gov.tw/). The landslide is a reac-
tive site for a visible landslide scarp after the M7.6
Chi-Chi earthquake, as shown in Figure 6.
Figure 2. Aerial Photos for the Tuchang Tribute pre- and
post-landslides (scale: 1/5000, photo by the Forestry Bureau,
http://www.forest.gov.tw/).

Figure 5. Historical Hazards in the Tuchang Tribute.

Figure 3. Buried area by the debris masses.

Figure 4. Landslide deposition area in the Tuchang Tribute

(taken on 30 September, 2004).

includes the position of the police station. Field inves-

tigation was made after the traffic was available on
30 Sep., 2004, as shown in Figure 4.

4 REASONS ATTRIBUTE TO THE HAZARDS

The reasons attributed to the landslide hazards were

studied for their landslide history, geology, hydrol- Figure 6. Historical scarp in the landslide site (scale:
ogy, and topographic characteristics, as shown in the 1/5000, photo by the Forestry Bureau, http://www.forest.
following. gov.tw/).

314
Figure 7. Geology map for the Tuchang Tribute.
Figure 10. Elevation map for the landslide area (unit: m).

Figure 8. Rainfall time history at the nearby rain-

gauge station (source Central Weather Bureau, http://www.
Figure 11. Slope map for the landslide area (unit: %).
cwb.gov.tw/).

station at the time in which the landslide dam breached,

shown in Figure 8. The rains were over the debris-flow
warning threshold of 250 mm in this area, coded by
the Soil & Water Conservation Bureau in 2007.

4.4 Topographic characteristics

The averaged slope for the landslide area was about
35.5◦ and an elevation changed 285 m from the
top at 1100 m to the base of 815 m as shown in
Figures 9, 10 and 11. Its slope length was 490 m
(source: the 9 m DTM from the Forestry Bureau,
http://www.forest.gov.tw). The aspect of the landslide
was north-northwest.

Figure 9. Surface map for the landslide area (unit: m).

5 EMERGENCY RESPONSE FOR THE
4.2 Geologic condition LANDSLIDE HAZARDS

The geological map shows that the landslide is located Four stages of the responses for the landslide haz-
at the margin of bedrocks for argillite, sandy shale ards relief are separated herein as (1) actions before
sandstone, and sandstone and shale, as shown in typhoon landfall, (2) actions during typhoon land-
Figure 7. fall, (3) emergency relief before disaster, and
(4) emergency rescue after disaster. The follow-
ing documented statements were sourced from the
4.3 Hydrological condition
National Disaster Prevention and Protection Commis-
The accumulated rainfall was 1340 mm and the rain- sion (http://www.ndppc.nat.gov.tw/), and the National
fall intensity was 27 mm/hr at the nearest rain-gauge Fire Agency (http://www.nfa.gov.tw/).

315
5.1 Actions before typhoon landfall • 25 Aug. 06:10. The bridge connected to the police
station was inundated by a flash flood (it was
The Central Emergency Operation Center was estab-
speculated as to the time of flood).
lished on 23 Aug. for preparing the Typhoon Aere
• 25 Aug. 09:10. The police station was covered by
landfall in Taiwan. The Council of Agriculture on
the landslide (it was speculated as to the time the
17:40 23 Aug., gave information to those residents
landslide was initiated).
who live near to the potential debris flow torrents to
• 25 Aug. 09:50. The policemen replied to the local
pay attention to the increased rainfall and to prepare to
emergency operation center in Wufeng Village by
evacuate for the mountainous areas in Hsinchu County.
wireless phone and reported that they had evacuated
to nearby houses.
5.2 Actions during typhoon landfall • 25 Aug. 09:55. Connection failed with the three
policemen by the wireless phone (it was speculated
At 08:00 AM on 24 Aug., the commander of the as to the time the landslide dam breached and debris
Central Emergency Operation Center for the Typhoon flow initiated).
Aere asked the Council of Agriculture to announce to
the public the warning areas for debris flows before
15:00 PM, especially for the tributes in the moun- 5.4 Emergency rescue after disaster
tainous areas. The information was provided to the
On 25 Aug. the commander of the Central Emergency
Council of Indigenous Peoples and the Department of
Operation Center asked the vice-commander, dele-
Social Welfare for the relief and preparedness of resi-
gates in the Council of Indigenous Peoples, and the
dents’ evacuation in the debris-flow warning areas for
Soil & Water Conservation Bureau to set up the For-
the commander at the local governmental emergency
ward Command Post near the village to aid the Hsinchu
operation center. Then, at 14:40, the commander asked
County Government in the emergency rescue.
the Council of Indigenous Peoples, the Council of
Starting on 26 Aug., the Central Emergency Oper-
Agriculture, and the Department of Social Welfare
ation Center asked the National Search and Rescue
to enforce the local government to evacuate the res-
Command Center, the rescuers at the Fire Depart-
idents at the debris-flow warning areas before 17:30.
ments in local areas, the non-governmental emer-
On 24 Aug. 16:23, the Wufeng Village was listed as
gency relief organizations, and the operators for
a warning area for debris-flow, and the commander
engineering machines, to help the local emergency
asked those residents who live nearby the potential
operation center at the Hisnchu County Govern-
debris flow torrents to evacuate. There was a strong
ment for rescue actions. The excavation by heavy
suggestion given to the village head of Wufeng Vil-
machines began on 1 Sep. 2004 for the dead in
lage on 24 Aug. 17:33 to evacuate the residents, and
the disaster. The rescue actions were performed as
three times to inform the local government to pay more
follows:
attention to this area. On 24 Aug. 22:10, the village
head rendered that he had asked the policemen to help • Seven special search and rescue teams from the
evacuate the residents, and three shelters in the village National Fire Agency were assigned to seven trib-
were opened for resident accommodation. utes in the mountainous areas in Hsinchu County,
Then, at 08:00 AM on 25 Aug., the commander for the aids of the rescue actions.
asked the Council of Indigenous Peoples, the Soil & • Seven international maritime satellite telephones
Water Conservation Bureau, and the Department of for emergency communication to the seven tributes
Social Welfare help the local governments to force in the mountainous areas were provided.
residents to evacuate in debris-flow warning areas and • The Ministry of National Defense sent soldiers to
to make a record for responsibility. Hsinchu County to help in the rescue actions.
• The National Search and Rescue Command Center
sent helicopters to deliver foodstuffs and transporta-
5.3 Emergency relief before disaster
tion for residents.
The communication records for the sacrificed police • The Council of Agriculture, the Council of Indige-
men before the disaster are list below: nous Peoples, and the Ministry of Transportation
& Communications were responsible for debris
• 24 Aug. 16:30. The policemen got the command flow field investigations and emergency road
to force the residents in the Tuchang Tribute and repair.
in other lower land areas to evacuate to a place of
safety at 17:00. Three bodies were found at the landslide site on
• 24 Aug. 17:40. The police station was inside the 5 Sep. 08:00. The search actions were continued and
warning areas. Three policemen left their station supported by 174 men, 19 hawks and 19 trucks, on
and sought safety at a nearby house. 11 Sep. 2004.

316
6 DISCUSSION AND CONCLUSION REFERENCES

A landslide dam breach, which induced debris flow Central Geology Survey, Ministry of Economic Affairs,
hazards, occurred on 25 Aug. 2004. The landslide retrieved date August 1, 2007, from http://www.moeacgs.
mechanism was examined for its historical hazards, gov.tw/.
geology, hydrology, and topographical characteristics. Central Weather Bureau, retrieved date August 1, 2007, from
http://www.cwb.gov.tw/.
The aids of field investigation, aerial photos, and Chen, C.Y., Lee, W.C. & Yu, F.C. 2006. Debris flow hazards
final communication records with the dead from the and emergency response in Taiwan. First International
disaster help to clarify the reasons attributed to the Conference on Monitoring, Simulation, Prevention and
hazards. The landslide mechanism was speculated as Remediation of Dense and Debris Flows, 7–9 June 2006,
a flood-induced riverbank erosion resulting in a land- Rhodes, Greece, 311–320.
slide, followed by a landslide dam, which formed and Chen, C.Y. 2007. Landslide Characteristics and rainfall
breached a flow of debris. distributions in Taiwan. Landslide and Climate Change-
There seems to have been a weak link during the Challenge and Solution, 20–24 May 2007, Ventnor, Isle
emergency operation for the central departments and of Wight, UK, 35–40.
Council of Agriculture, Executive Yuan, retrieved date
local government before the landslide, causing an August 1, 2007, from http://www.coa.gov.tw/.
inability by authorities to evacuate the residents to Council of Indigenous Peoples, Executive Yuan, retrieved
a place of safety early on. There also were numer- date August 1, 2007, from http://www.apc.gov.tw/.
ous difficulties during the emergency rescue actions Department of Social Affairs, Ministry of Interior, retrieved
because of cut-off roads from landslides, causing an date August 1, 2007, from http://www.moi.gov.tw/.
inability to communicate with the landslide-initiated Forestry Bureau, retrieved date August 1, 2007, from
mountainous area, which resulted in lack of informa- http://www.forest.gov.tw/.
tion to this region in time. It seems need more efforts Ministry of Transportation and Communications, retrieved
on enhancing the hazard management and a more effi- date August 1, 2007, from http://www.motc.gov.tw/.
National Disaster Prevention and Protection Commission,
cient field mechanism for improving road blockage in retrieved date August 1, 2007, from http://www.ndppc.
the mountainous areas is needed (Chen et al., 2006; nat.gov.tw/.
Chen, 2007). More attention to the seismic-induced National Fire Agency, Ministry of Interior, retrieved date
ground cracking is also needed to prevent a similar August 1, 2007, from http://www.nfa.gov.tw/.
case hereafter. National Science and Technology Center for Disaster Reduc-
tion (NCDR), Sindian City, Taiwan, retrieved date August
1, 2007, from http://www.ncdr.nat.gov.tw/.
ACKNOWLEDGEMENTS Soil & Water Conservation Bureau, Council of Agriculture,
Executive Yuan, retrieved date August 1, 2007, from
http://www.swcb.gov.tw/.
The author appreciates the National Science & Tech- The Ministry of National Defense, R.O.C., retrieved date
nology Center for Disaster Reduction (NCDR), the August 1, 2007, from http://www.mnd.gov.tw/.
Central Geology Survey, the National Fire Agency,
the Soil & Water Conservation Bureau, and the Coun-
cil of Agriculture for providing valuable materials for
the study.

317
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Pir3D, an easy to use three dimensional block fall simulator

Y. Cottaz
Geociel, Décines, France

R.M. Faure
Centre d’Etude des Tunnels, Bron, France

ABSTRACT: We present here the last release of a falling block simulator, called Pir3D. The main advance
is the user friendly front-end that allows a fast definition of the digital elevation model, from usual map to an
accurate 3D representation of the ground. New bouncing algorithms are now available and give the possibility
of comparisons of models. With these enhancements fence position is easy to define using mouse drag when
trajectories are stopped in real time.

1 INTRODUCTION The computer stores all this points with 3D coor-

dinates. The ground model is so very quickly deter-
Populated areas are increasing and in mountainous mined: it takes less than 60 minutes for 1000 validated
regions falling blocks from cliffs or steeped slopes are points. The point density is chosen by the operator
a real threat for buildings, roads, tunnel portals and that can add points where it is necessary, and any
lives. For risk assessment the new release of the 3D point can be added if necessary. When clicking on
simulator Pir3D, is really a jump in the use of it, with the ‘‘mesh’’ icons the convex contour of the set of
a new user-friendly front end and new algorithms for points is calculated and a Delaunay algorithm builds
bouncing calculation. (Faure et al, 1995). a triangular mesh and immediately the 3D model is
The paper will present, in a first part, the simula- displayed. The Delaunay algorithm ensures that each
tor following a real study, showing practical steps and circle built on the three points of a triangle does not
giving information about the consumed time. include any other point. With this algorithm the oper-
In a second part, focus is put on bouncing aspect, ator can tighten the points where necessary and have
developing the two algorithms used for the geotechni- more large triangle in other part of the ground model.
cal approach. (Faure et al, 1995). (see figure 1).
Comparisons between these two approaches and
others codes are given as conclusion.

2 GEOMETRY OF PIR3D

Three steps can be identified in the process: giving

topography, starting lines of trajectories and fence
position.

2.1 Topography and ground zones

The topography of the studied area is obtained from a
usual map or from an ‘‘Autocad’’® set of points, that
topographer usually use.
When one uses a map, the interesting area is
scanned as an image. The scale of the picture is given
through two identified points on the map. On the com-
puter screen, the operator gives the altitude of a level Figure 1. Following the level lines the operator introduce a
line, and with the mouse clicks along this line. set of points (1000 points in 60 minutes).

319
 Using a kind of ‘brush’, the operator paints the tri- 2.2 Starting lines
angles as to determine the kind of soil they represent.
For block fall simulation, the starting point of tra-
Figures 2, 3 and 4 describe this sequence.
jectories is usually not well known as it may occur
When the set of points is an ‘‘Autocad’’® file the
somewhere in the cliff.
operator only gives the name of the file and the 3D
A segmental line defined (mouse clicking) by the
model appears.
operator determines the starting points of the blocks.
Along this line a random algorithm gives the initial
point of each trajectory.
The easy to use front-end of Pir3D gives the possi-
bility to define several lines and move these lines with
a drag of the mouse.
When ready, Pir3D asks the operator for the num-
ber of blocks, usually, between 1000 to 10000, and
the initial speed. Mainly the initial speed is a vertical
falling of 1 meter high above the starting point, or an
accurate value giving the three speed components, vx ,
vy , vz , in case of seismic shock or blasting (Figure 5).

2.3 Fences
One of the most powerful possibilities in Pir3D, is the
simulation of fences, that will stop the block. The posi-
Figure 2. The triangular mesh appears when Delaunay tion of the fence is given by mouse dragging and its
algorithm is called, and the operator can determine the type efficiency is computed in real time and on the block
of grounds that are shown in different colors. diagram, trajectories intercepted by the fence don’t
appear beneath the fence. The operator can change
also the height, and very quickly, he can select on the
map the best position of the fence and determine its
height.
With this possibility the operator quickly deter-
mines the best position and the height of the fence
and for a better knowledge of the fence efficiency two
other graphics are drawn by Pir3D. The first one is
the fence elevation with the position of all the block
impacts, the colour of them indicate the energy of the
block following a colour scale. It is so possible to deter-
mine the kind of fence we can use. The second graphic
is an histogram of the energy impact, or the height of
Figure 3. A block diagram of the studied area. The map the impact on the fence. Figures 6, 7, and 8 illustrate
is mapped on the ground. the use of fence, and on figure 8, one can see the zoom
effect, available at any time when using Pir3D.

Figure 4. The same block diagram with soils differentia-

tion and level lines. Figure 5. Simulation with three starting lines.

320
Figure 6. The role of the fence (blue line with transpa-
rency and height).

Figure 9. Soil parameters showing uncertainty use on

Figure 7. The two diagrams giving more information on the
bouncing and sliding, for uniform distribution (The code can
fence system.
be used in French and English).

Figure 9. Lot of soils can be stored inside Pir3D, so

we have for identifying a soils its name and its color.
The uncertain on bounce parameters is defined as
a function of the speed. Usually variations are greater
at low speed and less important at high speed. It is
why when speed is close to zero we have ‘‘maximum
variability’’ and over a ‘‘speed limit’’, the variability
called ‘‘residual variability’’ is used. Between these
two speeds a linear function gives the value to be used.
(User manual, 2007).
The same table is defined for normal distribution.
The bouncing parameters are: energy restitution
Figure 8. A zoom for a better view on the simulation. with a distinction between vertical and horizontal
restitution, vertical and horizontal angle, and if the
angle after the bounce is less than a value, the
motion is a slide on the ground, with constant decele-
3 BOUNCING SIMULATION ration.
The computer code is written with object repre-
The bouncing model is based on point movement, but sentation and optimized, so 10000 trajectories are
with uncertainties on soils, on block shape, on vegeta- computed within a minute. Pir3D appears as a very
tion, we introduce uncertain parameters giving to the efficient tool that can be easily compared to other sim-
operator the choice between uniform distribution and ilar tools as results, but can’t be compared as use, its
normal distribution. Soil parameters are presented on user friendly interface give to it some advance.

321
4 USE FOR SPECIFIC STUDIES Each time a block flies over a pixel of a given
area at the foot of the slope, a counter notices it. For
In some case, for research purposes (back analysis each pixel of the given zone, the operator obtains the
of registered trajectory and soil parameter determina- probability of being reached by a block, and maps of
tion) or in case of comparisons Pir3D allows studies on probability, can be drawn. This representation is full of
2D profiles. In fact, these profiles are automatically interest when discussions with populations occur. For
transformed in 3D block diagram by extension of the engineer the display of energy received by each facet
profile. The uncertainty due to the ground vanishes, (total or maximum) is an appreciated information for
and fitting parameters is simplified. Figure 10 and 11 determining the best protection.
show this way.
Each trajectory can be studied in details; the opera-
tor can reach all the parameters of each bounce for any 6 CONCLUSIONS
trajectory. When some hundred of trajectories are plot-
ted, the selection is made using the mouse, and with up As a conclusion we summarize the hypothesis used in
and down touch the operator analyses all the bounces Pir3D following:
of the trajectories. (Labiouse et al, 2001) (Aliardi, Air resistance is not taken into account.
Crosta, 2003). Rocks have not a form. The influence of the form
is represented by random parameters.
Rock rotation on itself is not taken into account in
the calculations.
5 RISK AREAS DETERMINATION Rolling is not directly taken into account in calcu-
lations. Our software ignores this aspect as it depends
After simulating the trajectories of more than 10000 on the shape of the rock or stone in question. We have
blocks, the risk areas, at the foot of the slope can replaced this by a notion of rock sliding on the ground,
be determined in several manners and maps are below a minimum bounce angle.
automatically drawn. There is no interaction between rocks. Only isolated
rock fall situations and independent mass landslides
are considered.
Rock fragmentation is not considered. This has as
a consequence the dissipation of a large quantity of
energy; the worst scenario is thus represented by a
rock remaining intact.
Conditions relating to rock face breaking are not
taken into account. Initial rock fall conditions are
determined by the user.
These simplifications, used with uncertainty as
defined in Pir3D, give results very closed of reality,
and the use of Pir3D increases.

Figure 10. A extended profile showing the influence of REFERENCES

lateral uncertainty.
Faure R.M. Fayolle G. Tartivel F., 1995, PIR3D, a code which
simulate falling blocks in three dimensions. European
Conference of soils mechanics, Copenhagen. Vol 6 pp
27–32.
Labiouse V., Hendereich B., Desvarreux., Viktorovitch M.,
Guillemin P., 2001. Confrontation de logiciels trajec-
tographiques dans le cadre d’un programme Interreg IIC
consacré aux instabilités de falaises. Convegno Inter-
nazionale su opera di diffesa da caduta massi, Siusi, Italia,
Ottobre 2001. pp 13–23.
Agliardi F., Crosta G.B., 2003. High resolution three dimen-
sional numerical modeling of rockfalls. Int. J. of Rock
Mechanics and Mining Sciences 40, pp 455–471.
User manual, 2007, www.geociel.fr/pir3d.php.

Figure 11. Trajectories.

322
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Characterization of the fracture pattern on cliff sites combining geophysical

imaging and laser scanning

J. Deparis & D. Jongmans

LGIT, Université Joseph Fourier Grenoble, Grenoble Cedex 9, France

B. Fricout & T. Villemin

LGCA, Université de Savoie, Le Bourget du Lac Cedex, France

O. Meric & A. Mathy

SAGE, Gières, France

L. Effendiantz
CETE Lyon, Bron Cedex, France

ABSTRACT: Fracture characterization of potentially unstable cliffs is a crucial problem which can only partly
be solved by geological measurements at the surface. In this study we combine Laser Scanning (Lidar) and
geophysical techniques for obtaining the best possible image of the fracture pattern at the surface and inside
the rock mass. Two limestone cliff sites around Grenoble (French Alps), exhibiting different geometrical and
geotechnical features, were investigated in order to show the potential and the limits of the methods. Processing
of the Dense Digital Surface Models (DDSM) derived from Lidar data allowed fracture analysis which compares
well with field observations. For the site where abseiling was possible, performing Ground Penetrating Radar
(GPR) on the cliff face turned out to be the most effective method for accurately imaging the fractures inside the
rock mass. On the other site, seismic and electrical imaging techniques were used on the plateau for mapping
the fractures seen on the cliff.

1 INTRODUCTION co-registered images. In this approach, the point

clouds obtained from laser scanning are projected
Rock fall hazard assessment is a complex problem in on the image which keeps its original geometry and
high cliff areas, due to the difficulty of measuring and resolution. Each pixel can be localized and the 3D
mapping geological structures on the cliff face, and to orientation of fractures observed on the image can be
the little information about the internal discontinuities
provided by these conventional mapping techniques.
Within the framework of a French project funded by
the ‘‘Ministère de l’Ecologie et du Développement
Durable’’, we have developed a new approach which
aims at giving a more detailed description of the 3D
structure of the potentially unstable rock mass, for vol-
umes from a few thousands to a few hundred of thou-
sands of cubic metres. This methodology combines
two types of investigations: a structural analysis based
on remote measurements and geophysical imaging.
Airborne and terrestrial laser scanning techniques
are increasingly used for mapping in landslide detec-
tion and rock engineering (Lemy & Hadjigeorgiou
2004; Feng & Röshoff 2004; Schulz et al. 2005).
Bornaz & Dequal (2003) developed the concept Figure 1. Location map of the two cliff sites: Le Rocher du
of Solid Image combining laser scanning data and Midi and Le Ravin de l’Aiguille.

323
automatically computed. Solid images are a new tool Arnaud et al. 1978, Philippe et al. 1998). It is
offered to structural geologist for quickly, easily and separated from the mass on one side by a 1 m wide
safely having access to measurements. NW-SE fracture F1. The potential unstable volume
The geophysical methods have been increasingly was estimated at 50,000 m3 . A structural study of all
used for cliff investigation, both on the plateau above the nearby outcrops and of the vertical cliff has evinced
the cliff (Busby & Jackson 2006) and on the cliff the presence of three sets of discontinuities: the near-
face itself (Dussauge et al. 2003; Roch et al. 2006; horizontal bedding and a system of two conjugate near
Jeannin et al. 2006; Deparis et al. 2007). The use of vertical fracture families, Fa and Fb , striking N20◦ E
GPR (Ground Penetrating Radar) on the cliff face was and N125◦ E, respectively. Two electrical tomography
found to be the most valuable tool in terms of res- profiles (E1 and E2, Figure 2a) were conducted on
olution for investigating a rock mass (Jongmans & the plateau. As the rock offers the required quality for
Garambois 2007). Two limitations of GPR for cliff abseiling, four vertical GPR profiles (P1 to P4) were
investigation are safety requirements for abseiling and performed on the cliff face.
the penetration depth which was found to be lower The second site (Le Ravin de l’Aiguille, Figs. 2c
than 30 m with 100 MHz antenna in the limestone and d), composed of Tithonian limestone (Philippe
rocks around Grenoble (France). et al. 1998), is a 100 m wide and 170 m high tetrahe-
This study aims at combining the laser scanning dron, down pointed and limited by two large fractures,
and geophysical imaging techniques for characteriz- Fc and Fd striking N65◦ E and N130◦ , respectively. The
ing the fracture pattern on two cliff sites located in tetrahedron exhibits current signs of instability, with
the French sub-alpine limestone Massifs of Vercors frequent rock falls. The potential unstable volume was
and Chartreuse (Figure 1). On each site, the geomet- originally estimated at 2 × 105 m3 . Due to the danger
rical and mechanical characteristics of the cliffs allow of abseiling, GPR acquisition is impossible on this
adaptation of the investigation survey. cliff face and geophysical prospecting methods (seis-
mic and electrical profiles) were applied on the plateau
in order to delineate the internal limits and fractures of
the unstable rock mass.
2 STUDIED SITES CHARACTERISTICS

The first site (Le Rocher du Midi, Figures. 2a and b)

is a 100 m high column made of massive limestone, 3 STRUCTURAL ANALYSIS OF THE CLIFF
dated back to Lower Cretaceous (Urgonian limestone, USING SOLID IMAGES

For both sites, terrestrial laser scanning was made

from two different locations in order to have the most
complete coverage of the cliff (Figure 2).
For the ‘‘Ravin de l’Aiguille’’ site, the two acqui-
sitions points are located north of the cliff face. The
first one looked at the scarp with a very low incidence
angle while the second one had an incidence of 30◦
on average. Figure 3a shows a view of the ‘‘Ravin de
l’Aiguille’’ scarp from the first scanning point. The
dots have a color computed from digital images of
the site.
For the ‘‘Rocher du Midi’’ site, the first scanning
was performed from the North of the site at a distance
of about 150 m on average (Figure 2b). From this point
the scanner looked at the outcrop in a N140◦ E ± 10◦
direction with a dip ranging from +5◦ to −50◦ relative
to the horizontal plane. The second scanning point,
south of the cliff, looked at the outcrop in a N30◦ W ±
20◦ direction with a dip of 0◦ to −60◦ .
The process chain of the laser raw data was as
follows:
Figure 2. Helicopter views of the two sites and schematic
surface maps showing the location of the main observed frac- • Orientation of each point cloud: we placed on
tures and of the geophysical profiles. (a) and (b) Rocher du the cliff high reflectivity targets that were easily
Midi, (c) and (d) Ravin de l’Aiguille. The scanning points distinguished in the point cloud. The location of
are shown with white stars. these targets was measured in the field and used

324
Figure 3. Illustrations of the process chain for laser data Figure 4. Stereographic projection (lower hemisphere) of
a) View of the Ravin de l’Aiguille site after orientating of the fracture planes measured at the ‘‘Ravin de l’Aiguille’’
point clouds and filtering b) Triangular Irregular Network (top row) and the ‘‘Rocher du Midi’’ (bottom row). The left
(Rocher du Midi) and c) Synthetic 3D view with photographic column represent the data collected directly in the field while
draping (Upper south corner of the Ravin de l’Aiguille). data remotely measured on the solid images are on the right.
The large rock face corresponds to the N130◦ E limit of the
tetrahedron.
fracture planes have a surface ranging between
10−1 m2 to a few m2 . Most of them correspond to
to transform the laser coordinates to geographical
fractures or to small faults.
coordinates. After this step the point clouds were
On the ‘‘Rocher du Midi’’ site, our measurements
northerly and vertically oriented.
show two main families of near-vertical fractures
• Filtering of outliers for eliminating isolated points
(Figures. 4c and d). One strikes N10◦ E to N40◦ E
that are significantly far from the scarp.
while the second one has a direction ranging between
• Filtering of points corresponding to the vegetation
N110◦ E to N140◦ E. The two techniques (direct and
in order to keep only the points on the rock face
remote measurements) yield similar results although
itself.
the remote technique gives a less scattered diagram
• Co-registration of points clouds with the images of
(Figure 4d). The second family has an orientation sim-
the sites. This allows development of solid images.
ilar to the one of the large fissure F1 behind the site.
All points of the clouds are re-projected in the
On the ‘‘Ravin de l’Aiguille’’ site, results from both
images, allowing the field location of each pixel
methods are similar (Figs. 4a and b) although with
to be fixed.
slightly different orientations. Outcrop data are rare.
• Triangulation of the point clouds. Each point cloud
The measurements also show two families of vertical
was triangulated independently in a spherical geom-
fractures, N65◦ E ± 15◦ and N140◦ E ± 20◦ , the orien-
etry (Figure 3b), using only laser (azimuth, dip)
tation of which correspond to the ones of the two large
coordinates (Alberts 2004). This processing chain
fractures Fc and Fd defining the tetrahedron, north and
results in a TIN (Triangular Irregular Network), a
south respectively.
model of the site close to the real surface at a res-
olution of a few cm. The TIN was then used for
viewing in combination with the images of the site
(Figure 3c). 4 GEOPHYSICAL MEASUREMENTS

A structural analysis was first performed on the Geophysical experiments were conducted on the
accessible outcrops in the vicinity of each site. The plateau of the two sites, while GPR vertical investi-
fractures were measured with a clinometers-compass gation on the cliff face was only performed for the
and are shown in Figure 4 (a and c). A remote analy- ‘‘Rocher du Midi’’ site. Indeed, the cliff face of the
sis was done using the solid images on which fracture Ravin de l’Aiguille did not meet the safety require-
planes were manually delineated. If the selected areas ments for abseiling, due to the presence of numerous
are planar, the best fit plane is computed. The selected unstable blocks.

325
4.1 Rocher du Midi 4.1.2 GPR profiles
Four vertical GPR profiles (P1 to P4), 30 m to 60 m
Geophysical experiments included two electrical
long, were recorded on the cliff wall (Figure 2). These
tomography profiles on the plateau and four GPR ver-
Transverse Electric mode profiles were acquired with
tical profiles on the cliff face, the location of which is
a trace spacing of 20 cm, using unshielded anten-
given in Figure 2a.
nas of 100 MHz which offer a good compromise
between resolution and penetration in this limestone
4.1.1 Electrical Resistivity Tomography (ERT) (Jeannin et al. 2006). Reflecting targets were placed
Electrical profiles (E1 and E2, Figure 5) were car- on the cliff during the acquisition in order to locate
ried out perpendicularly to the cliff face (Figure 2a) precisely the GPR profiles. Due to higher reflectiv-
in order to check the continuity of the open fracture ity of these points compared to the limestone rock,
F1. The Wenner alpha array configuration was cho- their location is easily derived during the Lidar data
sen for its robustness (Dahlin & Zhou 2004), with processing. GPR data processing consisted of band-
an electrode spacing of 1 m and 2 m for profiles E1 pass filtering followed by a zero-phase band-pass
and E2, respectively. Inversion of apparent resistivity filter and a time to depth conversion using velocity
values was made using the software RES2DINV with of 10 cm/ns, which is derived from CMP (Com-
the L1 norm (Loke & Barker 1996). The influence of mon Midpoint Profile) analysis. To amplify the late
the 150 m high cliff on apparent resistivity measure- reflected events which were highly attenuated, an
ments was not corrected, as this effect would regularly AGC (Automatic Gain Control) process was used.
increase the resistivity values by a factor between two Finally, a static correction was performed to take into
near the cliff edge and one at the farthest distance account the topography of the cliff face. Non migrated
(Sahbi et al. 1997). Thus, strong lateral resistivity filtered profiles are displayed in Figure 6 for the
contrasts, which are the targets of this study, are little 100 MHz antenna. As expected, the penetration depth
affected by the presence of the cliff. The two profiles of 100 MHz antenna is about 25 m, except along pro-
(Figure 5) shows electrical resistivity values ranging file P3 (Figure 6c), located in a weathered conductive
from 50 .m in the highly weathered clayey zones, to zone where waves are strongly attenuated (Reynolds
more than 5000 .m in the open fractures, with a mean 1997). The theoretical resolution at 100 MHz of an
resistivity of 500 to 1500 .m within the rock mass. On open fracture location is about 25 cm but the detec-
profile E1, the F1 fracture (0.5 – 1 m wide) is clearly tion power of a thin bed could reach about 1 cm
displayed by a continuous 3 m wide vertical resistive for air filling (Jeannin et al. 2006). The GPR pro-
band, the thickness of which increases with depth. On files display numerous reflected waves, corresponding
profile E2, the fracture whose aperture is less than to near vertical or inclined fractures affecting the
10 cm at the surface, appears as a resistive spot at the rock mass.
surface. These results highlight the limited resolution Two horizontal GPR profiles acquired at the bottom
of the electrical tomography method for low-aperture of two vertical ones (P1 and P2) give the fracture ori-
fractures and the decrease of resolution with depth. entation. The near vertical fractures mainly exhibit two
orientations (N20◦ and N130◦ ) corresponding to the
two families Fa and Fb observed at the cliff surface.
A strong continuous reflector (labelled F2 ) dipping
inside the massif and parallel to the cliff face was
shown on the four GPR profiles. This major disconti-
nuity affecting the rock mass was not detected during
the initial surface investigation.
To validate the GPR results, a detailed investiga-
tion of the karstic network affecting the rock mass
was performed. The vertical sketch made from the
ground observations (Figure 7) shows the existence
of a major open fracture dipping to 45◦ at a depth
of 20 m and to 70◦ between 25 m and 40 m depth.
These observations are consistent with the geome-
try of the reflector F2 shown on the close P2 pro-
file (compare Figures. 6b and 7). This fracture is
parallel to the well-known and large Montaud fault
located 250 m from the site. These results high-
light the interest and the power of GPR methods
for characterizing the discontinuity pattern inside a
Figure 5. Rocher du Midi. Electrical profiles E1 and E2. rock mass.

326
 Figure 7. Schematic SW-NE cross-section drawn from
observations in the karstic network (Rocher du Midi site).

Figure 6. Rocher du Midi. GPR interpreted data for the

four profiles acquired with 100 MHz antenna a) P1, b) P2,
c) P3 and d) P4. Two horizontal profiles were conducted at
the bottom of P1and P2. Thin black lines show the interpreted Figure 8. Ravin de l’Aiguille. Block diagram showing the
fractures. 9 ERT profiles (E1 to E9). White lines delineate the dihedrons
(see text).

This fracture is clearly mapped by the electrical pro-

4.2 Ravin de l’Aiguille
files down to 20 m deep, particularly on the E1 to E5
The size of the Ravin de l’Aiguille site is far greater profiles. To the south and in the middle, the resistiv-
than the Rocher du Midi. As the cliff face conditions do ity variations show the presence of smaller dihedrons
not allow abseiling, nine ERT profiles (labelled E1 to imbricate into the main one (Figure 8). These results
E9 in Figure 2c) and two seismic profiles (S1 and S2) are validated by field observations. Two seismic pro-
were conducted on the plateau, parallel to the cliff face. files were carried out along the E2 and E5 electrical
Electrical images (Figure 8) exhibit resistivity values profiles. We used 48 vertical geophones at 3 m apart
ranging from 1000 to more than 14000 .m. The dihe- with a natural frequency of 4.5 Hz. Seismograms and
dron affected by open fractures is characterized by high first arrival picking are shown in Figure 9 for an end
resistivity values of a few thousands .m bounded by shot (a) and a fan shot (b) for the S1 profile.
conductive zones (less than 400 .m). To the north, Figure 9a clearly shows a change of the wave field
the narrow conductive zone coincides with the major characteristics, as well as a strong decrease of the sig-
fracture Fc limiting the dihedron and filled with clay. nal to noise ratio, at a distance of about 40 m, which

327
 the remote analysis and the one determined from direct
field measurement is good. These results highlight the
potential of using laser scanning techniques on high
cliffs, which yield a gain in time and in safety. For
the site where abseiling was possible, the Ground Pen-
etrating Radar (GPR) on the cliff face turned out to
be the most effective method for imaging the frac-
tures inside the mass, with a detection power of a few
cm with an antenna of 100 MHz and a penetration
of maximum 30 m. The traces of the profiles were
located on the solid images and the 3D coordinate of
Figure 9. Seismic profile S1 recorded for (a) an end shot the trace computed and used for topographical correc-
at 141 m and (b) a fan shot (b) (Figure 2). Black bars are the tions. Combining vertical and horizontal GPR profiles
picked first arrivals and grey bars (b) are the theoretical times on the cliff face allowed the 3D geometry of the frac-
computed for a weak anisotropic medium. R: reflection on tures to be mapped in homogeneous resistive limestone
the cliff face. rock mass. In other geological formations (e.g in shale,
micaschist and marly limestone), penetration can be
limited by the high rock electrical conductivity which
corresponds to the trace of fracture Fc . A more sub- attenuates radar waves. When abseiling was impossi-
tle modification of the signal frequency is observed ble due to the instability of rock blocks, seismic and
at 120 m, close to the Fd trace. A fan shot was made electrical imaging techniques were used on the plateau
at the dihedron edge (Figure 2c). P-waves (Figure 9b) for mapping the fractures seen on the cliff. Only the
exhibit a nearly constant arrival time (around 0.04 s), ones with an opening larger than 1 m were detected
in spite of the distance variations. The observed time using these methods, with a penetration limited to a
values are compared in the same figure with theoret- few tens of meters. On the ‘‘Ravin de l’Aiguille’’ site,
ical times computed in a weak anisotropic medium these techniques helped to delineate the geometry of
(Thomsen 1986) characterized by a N65◦ orientated the potential unstable rock masses and to determine its
fracture. P-waves are clearly delayed between the main volume.
fractures Fc and Fd , with a maximum value of about
0.01 s. These results highlight the fracturing effect
within the dihedron, the extension of which is approx-
ACKNOWLEDGMENTS
imately delineated by the time anomaly. Similar results
were obtained along the S2 profile.
This work was funded by the French national
project RDT (Risques, Décisions, Territoire, http://
www.rdtrisques.org/projets/camus/) from the Ministry
5 DISCUSSION AND CONCLUSION of the Ecology and Sustainable Development. We
thank the federal organization VOR (Vulnérabilité des
Fracture characterization on cliffs is a complex prob- Ouvrages aux Risques) for its financial support.
lem due to the difficulty of performing geological
observations and structural mapping on the cliff face,
and to the lack of information about the rock mass
provided by these techniques. This study aimed at eval- REFERENCES
uating the potential of applying laser scanning and
Alberts, C.P. 2004. Surface reconstruction from scan paths.
geophysical imaging techniques. Two sites showing Future Generation Computer Systems, 20:1285–1298.
potential instability problems with different charac- Arnaud, H., Bravard, Y., Fournier, D., Gidon, M. &
teristics of volume (50,000 and 200,000 m3 ) and of Monjuvent, G. 1978 Carte géologique à 1/50000, feuille
shape (column and wedge) were chosen in the cliffs de Grenoble. Tech. rept. BRGM Ed.
surrounding the town of Grenoble (French Alps) as Bornaz, L. & Dequal, S. 2003. The solid image: a new con-
test sites. In such high cliffs, terrestrial laser scan- cept and its applications. In The International Archives
ning is only appropriate when the cliff topography of the Photogrammetry, Remote Sensing and Spatial
allows the site to be correctly seen from one point Information Sciences, volume XXXIV.
or more, located along the crest at a short distance Busby, J. & Jackson, P. 2006. The application of time-
lapse azimuthal apparent resistivity measurements for
(typically <300 m) and covering the cliff with a sig- the prediction of coastal cliff failure. Journal of Applied
nificant incidence angle. The obtained point clouds Geophyiscs, 59:261–272.
can be combined with optical images to obtain a solid Dahlin, T. & Zhou, B. 2004. A numerical comparison of 2d
image on which a structural analysis can be performed. resistivity imaging with ten electrode arrays. Geophysical
The comparison of the fracture pattern deduced from Prospecting, 52:379–398.

328
Deparis, J., Garambois, S. & Hantz, D. In press. On the poten- Loke, M.H. & Barker, R.D. 1996. Rapid least-squares
tial of Ground Penetrating Radar to help rock fall hazard inversion of apparent resistivity pseudosections bya quasi-
assessment: a case study of a limestone slab, Gorges de la newton method. Geophysical Prospecting, 44:131–152.
Bourne (French Alps). Engineering Geology. Philippe, Y., Deville, E. & Mascle, A. 1998. Thin-skin
Dussauge-Peisser, C., Wathelet, M., Jongmans, D., Hantz, D., inversion tectonics at oblique basin margin: example
Couturier, B. & Sintes, M. 2003. Seismic tomography and of the western Vercors and Chartreuse Subalpine mas-
ground penetrating radar applied on fracture characteriza- sifs. Mascle A., Puigdefμabregas C., Luterbacher H.,
tion in a limestone cliff, Chartreuse massif, France. Near Fernμandez M. (eds) Geological Society, London, Special
surface geophysics, 1:161–172. Publications. Pages 239–262.
Feng, Q.H. & Röshoff, K. 2004. In-situ mapping and docu- Reynolds, J.M. 1997. An introduction to applied and envi-
mentation of rock faces using full coverage 3d laser scan- ronmental geophysics. John Wiley & Sons, Chichester,
ning technique, International Journal of Rock Mechanics England.
and Mining Sciences, 41, 1, 139–144. Roch, K.H., Chwatal, E. & Brückl, E. 2006. Potential of mon-
Jeannin, M., Garambois, S., Jongmans, D. & Grégoire, C. itoring rock fall hazards by GPR: considering as example
2006. Multiconfiguration gpr measurements for geomet- of the results of salzburg. Landslide, 3:87–94.
ric fracture characterization in limestone cliffs (alps). Sahbi, H., Jongmans, D. & Charlier, R. 1997. Theoret-
Geophysics, 71:B85-B92. ical study of slope effects in resisticity surveys and
Jongmans, D. & Garambois, S. 2007. Surface geophysical applications. Geophysical prospecting, 45(5):795–808.
characterization and monitoring: a review. Bull. Soc. géol. Schulz, T., Lemy, F. & Yong, S. 2005. Laser scanning
France, 178:101–112. technology for rock engineering applications. In Optical
Lemy, F. & Hadjigeorgiou, J. 2004. A Field application 3-D Measurement Techniques VII (Eds: Grün, Kahmen),
of laser scanning technology to quantify rock fracture Vienna.
orientation. In EUROCK 2004 & 53rd Geomechanics, Thomsen, L. 1986. Weak elastic anisotropy, Geophysics, 51,
435–438. 1954–1966.

329
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

In situ characterization of the geomechanical properties

of an unstable fractured rock slope

C. Dünner & P. Bigarré

INERIS, Ecole des Mines de Nancy, Nancy, France

F. Cappa & Y. Guglielmi

Géosciences Azur, CNRS-UNSA-IRD-UPMC, Sophia-Antipolis, France

C. Clément
LAEGO ∼ INERIS, Ecole des Mines de Nancy, Nancy, France

ABSTRACT: Rockfalls and landslides are recognized as a major natural hazard across the Alps with strong
economical and social impacts on regional land settlement and transportation policies. While numerous Alpine
zones and valleys are prone to gravitational risks, climate change is becoming an important issue on whether
or not the situation facing landslide hazards could become worse for decision makers. Along the Tinée Valley,
a steep fractured rock slope known as the Rochers de Valabres is currently being investigated as a Pilot Site
Laboratory (PSL), aiming to develop scientific knowledge on the physical processes of rock slope instabilities.
An original field characterization protocol coupling stress measurements and acoustic logging in a borehole was
conducted in order to improve site geomechanical properties and potential instability estimations. The authors
present an overview of this research site and objectives, as well as some preliminary results collected from
subsurface investigations and an assessment of the knowledge gained at this stage.

1 INTRODUCTION by INERIS and GEOSCIENCES AZUR Laboratory,

among other partners.
Rockfall is a slope failure mechanism for which The recent history of the PSL destabilization
both the preparatory and triggering processes are still marked by two important rockfalls that occurred in
poorly understood. Identification and quantification May 2000 (around 2,000 m3 of blocks and fragments)
of numerous predisposing factors—based on reliable and October 2004 (about 30 m3 of material) and caused
field data—is a challenging task while early-warning much damage. The first incident seriously damaged
signs of failure, or short term precursory mechanisms, one pile of a bridge over the Tinée River and caused a
still require research and development work. road traffic interruption, lasting several few weeks,
At present, it is difficult to correctly characterize, due to the considerable renovation works required.
on a deterministic and exhaustive basis, the rock mass Subsequent geological reconnaissance was urgently
fabric and the whole set of physical properties needed, undertaken and provided clear evidence of other poten-
mainly due to the variety of heterogeneities usually tial rockfalls. Since then, a research program has
encountered and to the differing scales of such het- been initiated and progressively developed as a long-
erogeneities. Correct assessment of rock mass initial term, extensive and multi-disciplinary program, the
volumetric conditions as internal boundaries, in situ description of which is beyond the scope of this
stresses, stiffness and moreover strength degradation paper.
and/or stress build up mechanisms with time requires The VAL-STRESS3D specific field experiment is
considerable effort and patience. part of this program. It is based on subsurface inves-
In order to improve knowledge about all of these tigations coupling in situ stress measurements and
issues relating to rock slope failure mechanisms, acoustic logging in a shallow borehole set in the
a large-scale fractured rock slope—the Rochers de lower part of the slope. The geological context of the
Valabres located in the Southern French Alps—is site, the preliminary results of stress measurements
being investigated as a Pilot Site Laboratory (PSL), and acoustic measurements are firstly presented.

331
Then, we discuss a better way to estimate slope proper-
ties and stability from a comparison between the three
approaches.

2 SETTING OF THE ROCHERS

DE VALABRES PSL

2.1 Geological features

The Rochers de Valabres PSL is located on the right
bank of the Tinée Valley, which is situated at the north-
western edge of the Argentera Mercantour crystalline
massif (French Alps). The basal portion of the PSL
is at an elevation of 900 m and the top of the slope
culminates at an elevation of 2,254 m. The topographic
surface is characterized by considerable roughness and
has a slope angle ranging from 40◦ to 60◦ (Figure 1).
This basement unit consists of metamorphic rocks
that have undergone a polyphased tectonic and meta- Figure 1. Frontal view of the Rochers de Valabres PSL.
morphic evolution during the Variscan and Alpine
orogeneses. The slope is mainly composed of two
hard mica gneisses with a sub-vertical foliation strik-
ing from N110◦ to N140◦ . The metamorphic foliation
appears undulated and micro-folded. Four sub-vertical
fault zones of decametric thickness cut the slope in
three zones. In these zones, the rock mass is intensively
fractured near the fault zones, with fracture density
decreasing rapidly while moving away from the faults
(Figure 2). These fractures can be seen to be par-
tially closed or opened with millimetric to centimetric
apertures.
Considering both the topographic and structural
complexities of the rock slope, subsurface investi-
gations were undertaken to provide more quantita-
tive information regarding stability analysis and the
hypotheses to be constrained.

2.2 Geological borehole logging

At the end of 2005, a 19 metres deep sub-horizontal Figure 2. Stereographic projection of the different sets
over-cored borehole was sunk in a compartment block of fractures (lower hemisphere). Sub-vertical fractures are
of interest, underlying a large planar failure flagstone clearly dominant. Dip directions of the Tinée and the Gorges
(Figure 1) to measure stresses, obtain samples and run de Valabres are indicated (after Gunzburger et al., 2004,
surveys for refined characterization of the rock fabric. 2005).
Except for the very first metres, RQD resulted in excel-
lent, numerous collected cores exceeding 0.5 metre in
length. A geological analysis of the drill cores as well
as a borehole endoscopy were performed (Clément
et al., 2006; Marignac et al., 2006).
The lithological analysis of the drill cores (Figure 3)
shows rather homogeneous intact rock corresponding
to highly migmatized gneisses with rough foliation
planes. For depths ranging from 9 m to 11 m, foli-
ation planes appear more regular and characterized by Figure 3. Drill core from over-coring test N◦ 5, at a depth
a significant increase in biotite content. of 15.75 m.

332
 Table 1. Synthesis of the elastic parameters used for inver-
sion of the gneisses of the Rochers de Valabres PSL.

Over-coring test 1 to 4 and 6 5

Transverse Isotropy Marked –
E1 (GPa) 25.2 28.6
E2 (GPa) 35.9 28.6
ν12 0.14 0.06
ν23 0.15 0.06
G12 (GPa) 12.4 13.5
G23 (GPa) 15.7 13.5
VP (m/s) 4500 3900
RC (MPa) 76.4 88.8

Figure 4. Log of location of over-coring tests and fractured

zones in the investigation borehole.

elastic parameters that had to be considered in stress

determinations.
Endoscopy reveals nine fault zones (Figure 4), six
Although biaxial testing does not allow the trans-
of these as a few millimetres-thick fault planes and
verse isotropic rock parameters to be determined, it
three 0.5 to 0.8 metre-thick breccia zones, located
can qualitatively highlight such parameters and, in
respectively at depths of 5 m, 13 m and 15 m.
any case, provides average values. Laboratory mea-
Faults are oriented sub-vertical and N40◦ to N70◦
surements on intact rock samples selected from drill
dip oriented, in full accordance with previous geo-
cores were thus conducted to complete and refine the
logical analyses (Figure 2). A significant number of
determination of elastic properties related to the pre-
fracture planes display a rough and oxidized surface
identified transverse isotropy of the gneiss. The two
(calcite infillings). Signs of water circulation were
zones sampled for this were those nearest to the tests
apparent although no actual water was found when
zones where a biaxial test could be carried out, in order
visual inspections were carried out.
to allow an easy quantitative correction of expected
scale effects.
Geomechanical results, along with scrutiny of the
3 IN SITU STRESS MEASUREMENTS drill cores and lithological analysis, confirmed a rather
homogeneous transverse isotropy of the deformability
3.1 Experimental protocol and the shear modulus due to the foliation (Table 1)
If rockfalls are disruptions originating from unbal- except in the fifth test zone, where foliation tapers off.
anced forces exceeding rock strengths, the local stress
field in steep fractured rock slopes remains difficult to 3.3 Results of stress measurements
assess. One objective of the VAL-STRESS3D experi-
ment was to perform a 3D in situ stress profile based The over-coring data was processed with SYGTGEO-
on the over-coring method. stress software, developed by INERIS, featuring
The over-coring method is based on CSIRO cell data inversion including anisotropic materials. Figure
measurements that make it possible to measure a full 5 presents the results of stress measurements with
3D strain tensor estimation and thus to back-calculate (σ1 , σ2 , σ3 ) magnitudes plotted as a function of bore-
a 3D stress tensor (σ1 , σ2 , σ3 ). Successive incremental hole depth. The results show essentially:
over-coring tests while drilling is progressing enable – a first zone between 0 and 1 m, which appears as
sharp stress gradients and locally heterogeneous stress rather fractured. Note that the first target zone was
fields with good accuracy to be assessed. intended to be placed at 1 m depth from the free
Six tests were conducted ranging from 2.5 m down surface, but disking to very shallow and dense fis-
to 18.5 m depth along the borehole (Figure 4). More suring meant the first measure had to be moved back
details of the experimental protocol can be found in to 2.5 m. This pre-existing fissuration is assumed
Dünner et al., 2007. to be most likely blast-induced;
– a second relatively homogeneous zone, between 2
and 6.5 m (over-coring tests N◦ 1, 2 and 3). This
3.2 Laboratory testing of the intact rock
zone is characterized by a major stress σ1 of approx-
Two cores from over-coring tests, located at depths imately 6 MPa and an intermediate stress σ2 of
of 2.5 m and 15.75 m, were completed by laboratory 5.4 MPa. Principal stresses (σ1 , σ2 ) are contained
biaxial tests in order to characterize the rock’s local in a sub-vertical plan. This orientation—which does

333
 and some coupling agent, while a push-rod system
enabled the probe’s depth position to be controlled.
Measurements provided two values of apparent veloc-
ity of P-waves associated respectively with each
receiver, VP1 and VP2 (Contrucci et al., 2007).
Data acquisition was based on automatic stack-
ing while analyses consisted in manually picking first
break arrival times in order to evaluate the P-wave
velocity profile versus depth. Eventually, a total of
156 measurements were collected.

Figure 5. Profile of principal stresses (σ1 , σ2 ) versus depth.

4.2 Results of velocity logging
It is worth noting that the ultrasonic logging was car-
not correspond in the outcrop to that of the geometry ried out more than one year after the borehole drilling.
of the average slope of the rock slope—is most likely Visual inspection showed a significant alteration of
influenced by the vertical discontinuities identified the borehole wall-sides with visible slight breakouts
afterwards. Minor stress σ3 presents a value of about and disseminated rock fragments, requiring a cleaning
2.6 MPa; it is horizontal, close to the drill hole axis, operation. This was quite unexpected when consider-
in other words oriented sub-perpendicularly to the ing the very smooth cores obtained initially and high
free surface at the borehole collar level; values of the strengths of the intact rock compared
– a third zone beyond 6.5 m (over-coring tests N◦ 4, to the very moderate level of stresses measured as
5 and 6). This zone has a stronger level in term of described here-above. As previously noted, no trace
average stresses, with a shift in the orientation of the of water was found in the borehole.
principal stresses. It presents a sharp stress gradient Two loggings were carried out along the borehole
at 10 m depth, for all three components (σ1 , σ2 , σ3 ); 90◦ from each other, first in the vertical plane and
– a local singularity at 15 m depth, with stresses in secondly in the horizontal plane.
full continuity to those measured in the first 6 m, Vertical logging was not conclusive since recurrent
followed by a maximum stress level at 18 m depth problems of coupling related to significant localized
with σ1 exceeding 12 MPa; and superficial alteration of the borehole wall-side
– except for measurement points 4 and 6 located at impeded correct emission of the mechanical stress
10 m and 18 m depth, (σ1 , σ2 ) is located in a sub- wavefront and introduced much dispersion in the data.
vertical plane, whereas locally the slope is 60◦ . This Data was found to be emergent, with low signal to
is driven probably by the different sets of vertical noise ratio, and not repeatable, while intensive use
fractures cutting through the rock mass. of coupling agent could not improve measurement
quality and reliability.
A second successful attempt was run 90◦ clock-
4 ULTRASONIC BOREHOLE LOGGING wise towards the right walls of the borehole. The depth
of investigation focused on the first 8 metres, show-
4.1 Ultrasonic acquisition set up ing good signal-to-noise ratio and repeatability of the
measurements.
In a second phase, a geophysical survey was carried This profile is presented in Figure 6, featuring:
out, based on high resolution ultrasonic logging, in
order to investigate the geophysical properties of the – in situ velocities VP1 ∼3,650 m/s and VP2 ∼
rock fabric. 4, 150 m · s−1 , with VP2 /VP1 ratio equal to 1.13.
The probe used for this logging was equipped These values are consistent with the velocities
with three ultrasonic transducers (1 transmitter and obtained from laboratory testing (Table 1);
2 receivers) operating in a 50–700 kHz frequency – a first zone between 0 and 2.5 m, where VP1 is
range centred at 150 kHz. Working with a dominant slightly increasing while VP2 is almost constant.
wavelength of a few centimetres offers a high resolu- This means that the first couple of metres of bore-
tion in detecting slight changes in physical properties hole walls may even be slightly damaged also along
and mineralogy, but most of all in detecting joint alter- this generatrix while free-surface and stress-release
ation states and opened fractures. Three transducers, effects, measured by VP2 , are not perceptible. This
two receivers respectively 9 cm and 18 cm distant from low VP1 zone, measuring a damage zone induced
the emitter, were placed in contact with the smooth around the borehole, most likely originates from
borehole walls using a pneumatic clamping device the construction of the overhanging road;

334
Figure 6. Horizontal plane velocity profile versus depth.
The arrows indicate velocity drops respectively at 3.5 m,
4.1 m and 5.3 m corresponding to fault zones located in
Figure 3. Figure 7. Profile of stresses and location of fractured zones
and over-coring tests.

– beyond 2.5 m, VP1 and VP2 are almost constant and

VP2 ∼VP1 . This trend indicates fairly homogeneous term of modulus and orientation—along the profile.
rock properties except for some specific zones; Even when considering the topography and frac-
– three thin zones affected by velocity drops respec- tured setting of the rock slope, the in situ stresses
tively at 3.5 m, 4.1 m and 5.3 m where the borehole found in the first 0–18 m range of depth may be
is intersecting fault zones. considered unexpectedly high and discontinuous.
These results confirm that the local stress field is
rather complex, characterized by both an overall
pattern and variability revealing the largely pre-
5 DISCUSSION dominant role of pre-existing fractures and local
heterogeneities.
Observations from the three approaches allow the
borehole to be divided into the following geomechan- As shown in Figure 7, both the measured principal
ical zones: stress and acoustic velocities variations appear closely
related to fault zones. Both stress magnitudes and VP
– a first zone between 0 and 2 m, which is significantly
values of intact rock are a factor of 1.6 higher than in
altered and fractured. Disking, lithological analysis
fault zones. These results show that fault zones clearly
of drill cores video analysis and lower velocity con-
display very different mechanical properties compared
firmed this important set of fracturing, which made
to intact rock zones and that such contrasts in the prop-
it necessary to conduct the first over-coring test at
erties induce large stress variations in the subsurface
a depth of about 2.5 m. This altered zone was most
of the rock slope.
likely induced by artificial blasting when the road
The low VP values in macroscopic fault zones
was built and by natural long term surface effects
intersected in boreholes are in general related to low
(stress release and some weathering);
Young’s modulus values of the zone materials com-
– a second zone between 2.5 and 7 m depth that
pared to intact rock Young’s modulus values (Barton
appears as rather homogeneous, for intact rock, in
et al. 1992; Kuwarahara et al. 1995; Malmgren et al.
terms of geology, stress tensor and wave velocity,
2006). Young’s modulus can be related to values from
despite the presence of three fractures at 3.5 m,
the following Equation (1):
4.1 m and 5.3 m shown by endoscopy and distinctly
identified through velocity drops;
– a clear singularity at about 15 m depth highlighted E = VP2 ρ(1 + υ)(1 − 2υ)/(1 − υ) (1)
by stress measurements. The geological survey
carried out on borehole cores and the endoscopy Where E is the Young’s modulus, VP is the velocity,
showed a breccia zone between 15.5 m and 16.5 m. ρ is the rock density and υ is the Poisson’s ratio.
This zone presents a strongly altered quartzic and If the Poisson’s ratio and the density are assumed
hydrochloric matrix. The presence of calcite in the constant, Equation 1 shows that a factor of 1.6 increase
fractures and fluid inclusions proves the occurrence of VP between intact rock and fault zones corresponds
of temporary meteoric fluid circulation; to a factor of 2.5 increase in the Young’s modulus.
– the excellent precision of the deformation data input Then, for an average value of EIntact rock = 30 GPa,
made it possible to highlight stress variations—in it provides a rough estimation of EFault zone = 12 GPa

335
for a thickness of the zones of a few 0.5 centimetres, experiments are planned to refine these results and
much less for a thickness of a few centimetres. estimate these fracture zone permeabilities and the
Moreover, at low stress, the elastic parameters relationships between permeabilities and fracture
of fault zones depend on the Uniaxial Compressive mechanical properties.
Strength (UCS) of the infillings and the fault walls, the
stress applied normal to the fault walls and the shape of
fault pores (Barton 1976 and 1990). Moisture effects
in fault pores can be neglected because the pores were ACKNOWLEDGEMENTS
dry when the experiments were performed. In our case,
fault zones that are filled with breccia and altered by Special thanks are extended to the French Ministry of
temporary water flow clearly display a much higher Ecology, Sustainable Development and Planning for
porosity and lower UCS compared to intact rock that financial support, as well as to the French National
could well fit with the average elastic modulus value Research Agency, to the Provence-Alpes-côte d’Azur
estimated from Equation 1. Region and the Alpes Maritimes Département.
Finally, such a method based on coupled accu- All authorizations and help from the Mercantour
rate acoustic logging and stress measurements through National Park and Electricité de France are fully
over-coring appears promising since it allows the acknowledged.
in situ rough estimation at shallow depths of macro-
scopic fault properties (stiffness and strength) from
two independent approaches respectively based on
velocity variations and stress-strain measurements. REFERENCES

Barton, N. 1976. Rock Mechanics Review—The Shear

6 CONCLUSIONS Strength of Rock and Rock Joints. International Jour-
nal of Rock Mechanics and Mining Sciences, Vol. 13,
pp. 255–279.
Complementary to surface investigations, subsurface Barton, N., Bandis, S. 1990. Review of predictive capabilities
surveys including in situ stress measurements, removal of JRC-JCS model in engineering practice. Rock Joints.
and laboratory tests of rock samples and geophys- Barton & Stephansson (eds). Balkema, Rotterdam. ISBN
ical logging have been undertaken at the Rochers 90 6191 109 5, pp. 603–610.
de Valabres PSL, France. As part of a long term Barton, N., Zoback, M.D. 1992. Self-Similar distribution and
research program related to rock slope failure mecha- properties of macroscopic fractures at depth in crystalline
nisms, these surveys aim to quantify the predominant rock in the Cajon Pass scientific drill hole. J. Geophys.
role of the geological structure, through the presence Res. 97, 5181–5200.
Clément, C., Laumonier, B., Brouand, M. 2006. Mineralog-
of discontinuities, faults and fractures in the failure ical study of gneisses of the Rochers de Valabres PSL.
mechanisms. Internal memo referenced INERIS DRS-06-75754LN07,
In the zone of investigation, the in situ stress field 10 p.
was quantified accurately and found much higher than Contrucci, I., Cabrera, J., Klein, E., Ben Slimane, K. 2007.
expected when considering analytical or numerical Excavation Damaged Zone characterized by ultrasonic
modelling. The in situ stress profile exhibits signif- borehole logging in drifts of different ages excavated in
icant variability and complexity, most likely governed the argillaceous formation of the Tornemire experimen-
both in terms of orientation and amplitudes by the pre- tal station (Aveyron, France), 3rd International Meeting
existing sub-vertical discontinuities. Normal stress to on Clays in natural & Engineered barriers for radioac-
tive waste confinement, 17–20th September 2007, Lille,
the discontinuities is clearly reduced. It must be kept France.
in mind that recent rockfalls just overlying the inves- Dünner, C., Bigarré, P., Clément, C., Merrien-Soukatchoff,
tigation zone may have a key role in the readjustment V., Gunzburger, Y. 2007. Natural and thermomechanical
of stresses and could explain the complex stress field. stress field measurements at the ‘‘Rochers de Valabres’’
Measured stresses are still found one order of magni- Pilot Site Laboratory in France. In: Ribeiro E Sousa
tude lower than UCS estimated from laboratory tests et al. Editors. Proc. of the 11th Congress of the Inter-
on intact rock. national Society for the Rock Mechanics (ISRM), 9–13th
Video and ultrasonic logging show that disconti- July 2007, Lisbon, Portugal. Leiden, The Netherlands:
nuities contain infilling materials between intact rock Taylor & Francis Group; 2007, vol.1, pp. 69–72.
Gunzburger, Y., Merrien-Soukatchoff, V., Senfaute, G.,
walls that generate very sharp contrasts in stiffness and Piguet J.-P., Guglielmi, Y. 2004. Field investigations,
shear strength that can be coarsely estimated. monitoring and modeling in the identification of rock fall
Some temporary water flows were apparent in causes. In: Lacerda et al. Editors. Landslides: evaluation
the fracture zones, which could also correspond and stabilization. Proc. of the 9th International Sym-
to preferential groundwater paths. Further in situ posium on Landslides (ISL), 28th June—2nd July 2004,

336
 Rio de Janeiro, Brazil. London: Taylor & Francis Group; frequency s-wave vertical seismic profiling (VSP) and
2004, pp. 557–563. borehole logging. Geotherm. Sci. Technol. 5, 71–78.
Gunzburger, Y., Merrien-Soukatchoff, V., Guglielmi, Y. Malmgren, L., Saiang, D., Töyrä, J., Bodare, A. 2006. The
2005. Influence of daily surface temperature fluctuations excavation disturbed zone (EDZ) at Kiirunavaara mine,
on rock slope stability: case study of the Rochers de Sweden—by seismic measurements. Journal of Applied
Valabres slope (France). International Journal of Rock Geophysics 61, 1–15.
Mechanics and Mining Sciences, vol. 42, no.3, April 2005, Marignac, C., Clément, C., Laumonier, B. 2006. Petro-
pp. 331–349. graphic analysis of the drill cores of VAL-STRESS3D
Kuwarahara, Y., Ito, H., Ohminato, T., Nakao, S., Kiguchi, experiment. Internal memo referenced INERIS DRS-06-
T. 1995. Size characterization of in situ fractures by high 75754LN06, 7 p.

337
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Properties of peat relating to instability of blanket bogs

A.P. Dykes
School of Earth Sciences and Geography, Centre for Earth and Environmental Science Research,
Kingston University, UK

ABSTRACT: Instability and failure of hillslopes mantled with blanket peat occurs naturally in cool temperate
regions that favour the development of blanket bogs. Such peat failures are most common in the UK and Ireland,
where they pose potential hazards to properties and infrastructure. These blanket bogs are also particularly
susceptible to engineering-induced instability. This paper identifies an urgent need for new approaches to the
assessment of blanket peat stability, particularly through appropriate geotechnical characterisation, and examines
some of the principal issues identified during recent research that now need to be addressed including validation
or modification of existing methods of quantitative stability analyses for use with peat.

1 INTRODUCTION a type of peat deposit not previously of concern to

engineers.
Peat is a material formed from the accumulation of Properties of peat were examined and reviewed
waterlogged remains of dead and decaying plant mat- specifically for engineers by Hobbs (1986). How-
ter. It has traditionally been investigated by engineers ever, data presented in his review were primarily
primarily concerned with settlement following loading obtained from fen or transitional (fen-bog) peat, and
associated with construction of roads or other infras- the shear strength of peat was not addressed. More
tructure across peat deposits such as the Irish raised recent reviews have provided little additional mate-
bogs (e.g. Hanrahan 1964) or the Canadian Muskeg rial. In other words, properties and characteristics
(MacFarlane 1969). However, natural instability of of some peatlands that are particularly susceptible to
peat deposits had not been systematically investigated natural or engineering-induced failure have not previ-
until the last ten years, when initial detailed research ously been reported in the mainstream engineering or
on specific peat landslides (Kirk 2001, Dykes & geomorphological literature.
Kirk 2001) and on the general topic of peat failures The upland blanket bogs of the UK and Ireland in
(Mills 2002, Warburton et al. 2004) was followed the northern hemisphere and subantarctic islands in the
by an extensive research programme arising from the south fall into this category of susceptible peatlands,
peat landslide events of 2003 in the UK and Ireland although the latter locations are largely uninhabited
and founded on a new classification scheme for peat and peat failures generally do not constitute a hazard.
failures (Dykes & Warburton 2007a). In fact, over 80% of all known natural peat mass move-
On 19 September 2003, extreme rainfall triggered ments occurred in the British Isles (60% in Ireland and
40 landslides on a small mountain in western Ireland 20% in England and Scotland: Dykes & Kirk 2006),
that included 12 large peat failures and had severe most commonly involving upland blanket bog on hill-
impacts on the rural communities below the moun- slopes with gradients typically in the range 2–12◦ .
tain (Dykes & Warburton 2007b). On the same day, Windfarm construction and commercial forestry oper-
a separate extreme storm triggered around 35 land- ations on British and Irish uplands create additional
slides in Shetland, northern Scotland, including 20 potential hazards from failures of blanket peat.
large peat slides, which combined with floodwaters The aim of this paper is to highlight the problems of
to cause significant damage to local infrastructure assessing and analysing the stability of hillslopes man-
(Dykes & Warburton, in press). Then, a month later, tled with blanket peat. Specific issues are identified,
engineers preparing the site for a wind turbine in Ire- arising from the nature of peat as an engineering mate-
land triggered a 450,000 m3 peat flow that also had rial, that need to be addressed if the hazard from natural
environmental and socio-economic consequences. All blanket bog failures is to be adequately assessed and
of these events involved failures of upland blanket bog, engineering-triggered failures avoided.

339
2 TYPES AND MECHANISMS condition of undisturbed or engineered blanket bog
OF PEAT FAILURES covered hillslopes.
There are four principal difficulties associated with
Several distinct types of peat failures have been iden- stability assessments of blanket peat. These will be
tified and defined by Dykes & Warburton (2007a) addressed in turn.
primarily according to field morphology as controlled
by apparent failure position and mechanism, but with 1. All peats are highly heterogeneous, making the
additional divisions for types of peat deposit (e.g. classification of peat for engineering purposes
‘bog bursts’ exclusively refer to raised bogs) and difficult. Small-scale variations in the botanical
engineering-induced failures (‘peat flows’ being used composition and stratigraphy undoubtably give rise
to identify head-loaded failures). The latter designa- to widely ranging values for some geotechnical
tion makes no assumption about failure mechanism: and index properties (although these groups of peat
for example, the 450,000 m3 Irish windfarm peat flow properties have never yet been correlated), and in
began as a translational shear failure within the basal some cases the matrix properties may be less use-
peat caused, at least in part, by placement of exca- ful than the ‘peat mass’ (c.f. rock mass) properties.
vated peat up to 1 m thick on the in-situ blanket bog For example, the saturated hydraulic conductivity
(AGEC 2004). It is not known if any of the other (ksat ) of the lower ‘catotelm’ peat, i.e. below the
head-loaded peat flows failed by translational shear- level of maximum seasonal water table draw-down
ing or some other mechanism, but for any stability 0.2–0.5 m below the surface, could be between
assessment it would be reasonable to initially assume 10−3 and 10−10 m s−1 , higher in the horizontal
a shearing component. plane due to compression of plant fibres (Hobbs
Large and potentially highly damaging natural ‘bog 1986). On the other hand, laboratory measurements
slides’ involve translational shearing within the lower of ksat of undisturbed ‘small’ and ‘large’ core sam-
peat, whereas ‘bogflows’ are thought, on the basis of ples (50 or 100 mm long × 50 or 100 mm diameter)
eyewitness accounts in the early 20th century (e.g. of basal catotelm peat from several Irish bogflows
Delap et al. 1932), to begin due to in-situ loss of and bog slides, using a standard constant-head
strength (analagous to sensitive ‘quick’ clay) caused by technique, varied between 10−4 and 10−11 m s−1
high water pressures. However, there is as yet no evi- with either higher vertical values or no signifi-
dence for any specific failure mechanism for bogflows cant anisotropy (e.g. Yang and Dykes 2006, Dykes
so an initial shearing component cannot be ruled out. et al., in review, Fig. 2). Dissipation of excess
The potential for severe impacts from these events is pore water pressures due to loading, for example
perhaps demonstrated by their scale: whilst bog slides by placement of fill or spoil on intact peat over a
tend to involve several tens of thousands of cubic relatively short time (as occurred at the Irish wind-
metres of peat moving from 5–8◦ slopes, bogflows farm), will be limited not only because of the low
of up to 375,000 m3 have been recorded from 2–5◦ natural permeability but because the permeability
slopes in Ireland. Both types can produce extensive falls rapidly as the void ratio is reduced by loading
debris runout with a high capacity to cause physical (Hobbs 1986): the resulting potential for instability
damage and pollution. is clear. In natural blanket bogs, the permeabil-
ity of the peat matrix is much less important than
that of the peat mass, with inputs and subsequent
drainage of rainwater occurring by means of nat-
3 GEOTECHNICAL ISSUES IN BLANKET ural pipes, cracks and other macropores (Holden
BOG STABILITY ASSESSMENT and Burt 2003). Natural instability has commonly
been associated with heavy rainfall, the assump-
Investigations of twelve separate peat landslides at tion being that rainwater supply may be sufficiently
seven locations in northern and western Ireland have rapid to exceed the natural drainage capacity so
generated a set of data that describe the index prop- that excess water pressures were generated within
erties and some geotechnical properties of the failed the basal peat (Warburton et al. 2004). Indeed,
blanket bogs. These failed peats possessed much at some sites, the low permeability peat matrix is
higher water contents and liquid limits than fen or thought to have caused artesian water pressures in
transitional peats at the same degree of humification soil pipes beneath thin blanket bog, contributing
as reviewed by Hobbs (1986) (Fig. 1). Furthermore, to widespread peaty-debris slides (e.g. Dykes and
although the limited data do not appear to show Warburton 2007b).
any differences between the different failure types, 2. There are almost no data to show how peat
they do include preliminary geotechnical results that strength relates to any other index physical prop-
highlight the need for new or modified techniques erties or indeed botanical properties. Long (2005)
in order to be able to reliably assess the stability and Dykes & Kirk (2006) reviewed published peat

340
 10
Humification (vonPost scale)

0
0 200 400 600 800 1000 1200 1400 0 200 400 600 800 1000
Liquid Limit (%) Field Water Content (%)

Figure 1. Variations in liquid limit and water content with humification (after Hobbs, 1986). Black circles represent data
from fen-transitional peat in Shropshire, UK (50–70% organic content), open squares refer to a drained raised bog in County
Durham, UK (99% organic), and crosses indicate properties of Irish upland blanket bogs (> 95% organic). Both plots use the
same y-axis.

1E-09 1E-08 1E-07 0.000001 0.00001 0.0001 0.001 0.01 0.1 1

0
Horizontal,small core samples
-0.2
Vertical, small core samples
-0.4 Horizontal, large core samples
-0.6 Vertical, large core samples
Depth (m)

-0.8

-1

-1.2

-1.4

-1.6
Saturated hydraulic conductivity (m/s)

Figure 2. Saturated hydraulic conductivity of blanket peat samples from four landslides on Dooncarton Mountain, Ireland,
in 2003. No upper acrotelm layer could be distinguished in any of the peat profiles, so all samples are assumed to represent
catotelm peat.

strength data: Table 1 presents data from peat thought that peat does not fully (if at all) con-
deposits of similar type and/or similar strength form to the Mohr-Coulomb failure model, but it
characteristics to blanket bog. All of the results has not yet been established that conventional slope
were obtained from samples consolidated under stability analyses using M-C parameters cannot
specified normal stress (within the range 5 to adequately represent shear failures in peat. How-
50 kPa) prior to application of the compressive ever, the underlying difficulty is that shear strength
(triaxial) or shearing (direct shear) stress. The of peat cannot yet be reliably estimated from other
final three entries in Table 1 high-light two impor- properties. The shear strength of peat obtained
tant points: firstly, that application of identical from vane tests has been demonstrated to be unre-
(direct shear) test procedures on the same equip- liable because of the influence of fibres in the peat
ment demonstrates real differences between dif- (Helenelund 1967, Landva 1980), although vanes
ferent Irish blanket bogs, and secondly, that these may be useful for estimating relative strength vari-
shear tests overestimated the cohesion. In fact, ations. Helenelund (1967) considered the tensile
back-analyses of peat failures using, for exam- strength to be a potentially more reliable indicator
ple, the infinite slope model, consistently require of peat strength, and found tensile strength values
much lower shear strengths than laboratory tests of some Canadian and Finnish peats to be approx-
produce (e.g. Dykes et al., in review). It is now imately half of the corresponding vane strengths.

341
Table 1. Published values for the shear strength of peat (after Dykes & Kirk, 2006).

Cohesion, Internal
Original source Peat type/characteristics c (kPa) friction angle, φ

Adams (1965) Peat with low moisture contents 0 48◦

Hanrahan et al. (1967) Remoulded H4 Sphagnum peat 5.5–6.1 36.6–43.5◦
Hollinshead & Raymond (1972) No information 4.0 34◦
Landva & La Rochelle (1983) Sphagnum peat (H3 , mainly fibrous) 2.4–4.7 27.1–35.4◦
Farrell & Hebib (1998) Raised bog, 98% organic; undr. triax. compression: 55◦
direct and ring shear: 38◦
Kirk (2001) Ombrotrophic blanket peat, 0.35–1.25 m deep 2.7–8.2 26.1–30.4◦
Long (2005) Review of triaxial test data from studies of
Indonesian, Brazilian and Japanese peat 32–58◦
Long & Jennings (2006) Unspecified (blanket?) peat, 1–2.5 m deep, H4–5 to H7–8 3–5 35◦
Dykes & Warburton (2008) Ombrotrophic blanket peat, 1.4–1.7 m deep, H10 8–11 21–25◦
Dykes & Warburton (2008) As above, calculated from back-analyses of failures 2.0–6.2 21◦
Dykes (unpublished data) Ombrotrophic blanket peat, 2 m deep, H5–6 5.2 33.4◦

1600 Welsh bog

Welsh bog
1400
Shropshire fen

1200 Sizewell fen

WC = LL
Liquid Limit (%)

1000 WC = 0.875 LL
WC = 1.38 LL
800
Fen-Bog transition

600 Irish blanket bogs

400

200

0
0 200 400 600 800 1000 1200 1400 1600 1800 2000 2200 2400 2600
Field Water Content (%)

Figure 3. Variations in liquid limit with water content (after Hobbs, 1986). Hobbs identified fen peat below the horizontal
transition line, ‘amorphous bog peat’ immediately above this line where WC > 1.38 LL, and ‘fibrous peat’ higher up the
‘WC = 1.38 LL’ line.

This measure of strength may be worth revisiting (Yang and Dykes 2006). Hobbs demonstrated clear
in the context of blanket peat instability. relationships between LL and organic content in
3. The liquid limit (LL) of peat is known to sometimes fen and transitional peats, but that within any nar-
exceed the natural water content (wc) (Fig. 3) but row range of organic contents, the LL falls with
it is difficult to determine. The ‘loss on ignition’ increasing humification (Figs 1, 4). In Figure 4,
is often used as an alternative indicator of the sus- the latter control is best exemplified by the results
ceptibility of the peat, but this is unreliable as LL from bog peat (organic content > 90%), the highly
and wc vary with humification (and usually depth) humified Irish blanket peat having the lowest LLs
(Fig. 1) and with type of peat (Fig. 4). The LL may among this group (c.f. Fig. 1). Hobbs also
be a more fundamental indicator of peat properties suggested that bog peats can be readily distin-
and behaviour than is currently thought (Hobbs, guished from fen and transitional peats according
1986), and issues relating to its determination have to whether the natural water content is greater or
been examined using samples of Irish blanket peat less than the liquid limit, with LL = 700% being

342
 100
Cranberry Moss
Welsh bog
80 Shropshire & Sizewell fens
Avonmouth (buried peat)
Loss on Ignition (%)

60 Kings Lynn (buried peat)

Skempton & Petley (1970)
MacFarlane (1969)
40
Miyakawa (1960)
Shropshire & Sizewell fens
20 Irish blanket bogs

0
0 200 400 600 800 1000 1200 1400 1600
Liquid Limit (%)

Figure 4. Variations in liquid limit with organic content as indicated by ignition loss for different UK peatlands, including
summary relationships obtained from Japan (Miyakawa) and Canada (MacFarlane) (after Hobbs, 1986, Figs 16 & 17b).

identified as the fen-bog boundary (Fig. 3, also saturated then sheared in a 100 mm square shear-
applicable to Fig. 4). Hobbs explicitly excluded box. These samples were not consolidated prior
blanket peat from this analysis, although the data to shearing, and the normal loads of between 1
from Irish blanket bogs (Yang & Dykes 2006) do and 10 kPa were applied to each respective sample
plot within the ‘amorphous bog peat’ zone. It can as shearing commenced, possibly generating some
therefore be seen that engineering works involv- matrix pore water pressures broadly similar to the
ing blanket peat deposits cannot rely on published in-situ field condition. The samples were sheared
general relationships between index peat proper- rapidly, at 0.2 or 0.5 mm min−1 , to reflect the
ties because of their characteristically high degree field evidence of extremely rapid failure with the
of humification, negligible mineral content and associated likelihood of some undrained shearing
botanical composition associated with acidic plant effects, again better representing field conditions.
species (c.f. generally alkaline fen peats: Hobbs, The problems of this approach are recognised,
1986). Analysis and interpretation of natural fail- including the absence of pore water pressure moni-
ures in blanket bogs will also therefore need to be toring or control and the use of normal loads far
primarily based on data obtained from site-specific below the generally accepted minimum for this
peat samples. type of equipment. However, the results from two
4. Blanket bog is typically fully saturated all year, replicated sets of experimental tests were suffi-
thus in its natural state it exerts a negligible (or ciently similar to allow preliminary interpretation
even zero) normal stress on its basal layer. If the and to guide subsequent research. What the results
bog is slightly unsaturated, for example if the water appear to show is: (i) the natural basal peat has very
table falls up to 0.5 m below the surface, this repre- little in-situ strength in its undisturbed state with
senting an exceptional degree of drying in western minimal effective stresses, consistent with back-
Ireland, the basal peat may experience an effective analyses of stability and with field experience of
normal stress of up to around 5 kPa. The low to zero extracting samples for testing; and (ii) further low-
effective normal stresses that apply within natural stress direct shear and triaxial strength tests are
blanket bogs raise serious doubts about the validity warranted to investigate the validity and potential
of any shear strength data obtained using standard utility of this approach for use with existing or
procedures, in which the peat consolidates signifi- improved stability assessments and analyses.
cantly under the applied normal loads. Preliminary
experiments were therefore undertaken using a
direct shear apparatus to provide results for compar-
ison with data obtained previously from the same 4 CONCLUSIONS
blanket peats using standard procedures. Small
block samples of undisturbed basal peat extracted The possibility of more frequent natural peat fail-
from near the margin of an Irish bog slide were fully ures as a consequence of climatic changes, and the

343
 30
S1
SE5
25 SE5 low pre-consolidation
E6 drained (Kirk, 2001)
E6 no pre-consolidation
Peak shear stress (kPa)

20 E6 no pre-consol. (replicated)
Landva & La Rochelle (1983)
S1
15
SE5
SE5 low pre-consolidation
10 E6 drained (Kirk, 2001)
E6 no pre-consolidation

0
0 5 10 15 20 25 30 35 40
Normal stress (kPa)

Figure 5. Shear strength results obtained from direct shear tests of Irish upland blanket bog peat (S1, SE5—Dykes &
Warburton, 2008; E6—Dykes & Kirk, 2006), also showing the range of ring shear test results (upper and lower envelopes)
obtained from Canadian H3–4 Sphagnum peat by Landva & La Rochelle (1983). All samples were normally consolidated
prior to rapid first-time shearing except those identified as ‘no pre-consolidation’ (in which the normal loads were applied
as shearing commenced, with consolidation occurring during the initial shearing), ‘low pre-consolidation’ (consolidated
under 5 kPa normal loads prior to shearing, with the full normal loads applied as shearing commenced) and ‘drained’
(consolidated-drained tests using very low shear rates).

increasing numbers of windfarms and other activi- University. The contributions of many colleagues to
ties proposed for upland environments that require this work over several years are gratefully acknowl-
engineering such as construction of access roads, sug- edged.
gest that the need for improved methods for analysing
and modelling the stability of upland blanket bogs is
urgent. Conventional stability analyses of blanket bog
covered slopes may be appropriate, as has thus far been REFERENCES
assumed, but there are as yet insufficient data to ver-
Adams, J.I. 1965. The engineering behaviour of a Canadian
ify this or to permit the development of more reliable muskeg. Proceedings, 6th International Conference on
failure models for peat deposits. However, conven- Soil Mechanics and Foundation Engineering 1: 3–7.
tional geotechnical analyses of this material are clearly AGEC. 2004. Reports on the Derrybrien Windfarm—Final
inappropriate. The research priorities are therefore to Report on Landslide of October 2003. Unpublished report,
investigate the botanical controls on the geotechnical Applied Ground Engineering Consultants Ltd., Ireland.
properties of peat, to establish a reliable method for Delap, A.D., Farrington, A., Praeger, R.L. & Smyth, L.B.
determining the shear strength of peat, and to identify 1932. Report on the recent bog-flow at Glencullin, Co.
or develop a reliable method for analysing the stability Mayo. Scientific Proceedings of the Royal Dublin Society
20: 181–192.
of blanket bog covered slopes.
Dykes, A.P. & Kirk, K.J. 2001. Initiation of a multiple
peat slide on Cuilcagh Mountain, Northern Ireland. Earth
Surface Processes and Landforms 26: 395–408.
ACKNOWLEDGEMENTS Dykes, A.P. & Kirk, K.J. 2006. Slope instability and mass
movements in peat deposits. In I.P. Martini, A. Martínez
Cortizas & W. Chesworth (eds), Peatlands: Evolution
The Irish blanket peat data were obtained during and Records of Environmental and Climate Changes:
projects supported financially or otherwise by the UK 377–406. Amsterdam: Elsevier.
Natural Environment Research Council (Grant Refs. Dykes, A.P. & Warburton, J. 2007a. Mass movements in
NER/A/S/2003/00888–9), the Limestone Research peat: A formal classification scheme. Geomorphology 86:
Group, Fermanagh District Council & Huddersfield 73–93.

344
Dykes, A.P. & Warburton, J. 2007b. Geomorphological con- Kirk, K.J. 2001. Instability of blanket bog slopes on Cuilcagh
trols on failures of peat-covered hillslopes triggered by Mountain, N.W. Ireland. Unpublished PhD thesis, Univer-
extreme rainfall. Earth Surface Processes and Landforms sity of Huddersfield, UK.
32: 1841–1862. Landva, A.O. 1980. Vane testing in peat. Canadian Geotech-
Dykes, A.P. & Warburton, J. 2008. Failure of peat-covered nical Journal 17: 1–19.
hillslopes at Pollatomish, Co. Mayo, Ireland: analysis Landva, A.O. & La Rochelle, P. 1983. Compressibility
of topographic and geotechnical influences. Catena 72: and shear characteristics of Radforth peats. In P.M. Jar-
129–145. rett (ed.), Testing of Peats and Organic Soils: 157–191.
Dykes, A.P. & Warburton, J. In press. Characteristics of the Philadelphia: ASTM Special Technical Publication 820.
Shetland Islands (UK) peat slides of 19 September 2003. Long, M. 2005. Review of peat strength, peat characteri-
Landslides. sation and constitutive modelling of peat with reference
Dykes, A.P., Gunn, J. & Convery (née Kirk), K.J. In to landslides. Studia Geotechnica et Mechanica XXVII:
review. Landslides in blanket peat on Cuilcagh Mountain, 67–90.
Northern Ireland. Geomorphology. Long, M. & Jennings, P. 2006. Analysis of the peat slide at
Farrell, E.R. & Hebib, S. 1998. The determination of the Pollatomish, County Mayo, Ireland. Landslides 3: 51–61.
geotechnical parameters of organic soils. Proceedings MacFarlane, I.C. (ed.) 1969. Muskeg Engineering Handbook.
of the International Symposium on Problematic Soils, Toronto: University of Toronto Press.
IS-TOHOKU 98, Sendai, Japan: 33–36. Mills, A.J. 2002. Peat slides: morphology, mechanisms and
Hanrahan, E.T. 1964. A road failure on peat. Géotechnique recovery. Unpublished PhD thesis, Durham University,
14: 185–202. UK.
Hanrahan, E.T., Dunne, J.M. & Sodha, V.G. 1967. Shear Miyakawa, I. 1960. Some aspects of road construction in
strength of peat. Proceedings of the Geotechnical Confer- peaty or marshy areas in Hokkaido. Sapporo, Japan: Civil
ence, Oslo, Vol. 1: 193–198. Engineering Research Institute, Hokkaido Development
Helenelund, K.V. 1967. Vane tests and tension tests on fibrous Bureau.
peat. Proceedings of the Geotechnical Conference, Oslo, Skempton, A.W. & Petley, D.J. 1970. Ignition loss and other
Vol. 1: 199–203. properties of peats and clays from Avonmouth, King’s
Hobbs, N.B. 1986. Mire morphology and the properties and Lynn and Cranberry Moss. Géotechnique 20: 343–356.
behaviour of some British and foreign peats. Quarterly Warburton, J., Holden, J. & Mills, A.J. 2004. Hydrologi-
Journal of Engineering Geology 19: 7–80. cal controls of surficial mass movements in peat. Earth
Holden, J. & Burt, T.P. 2003. Runoff production in blanket Science Reviews 67: 139–156.
peat covered catchments. Water Resources Research 39: Yang, J. & Dykes, A.P. 2006. The liquid limit of peat and
1191, doi:10.1029/2002 WR001956. its application to the understanding of Irish blanket bog
Hollingshead, G.W. & Raymond, G. 1972. Field loading tests failures. Landslides 3: 205–216.
on Muskeg. Canadian Geotechnical Journal 9: 278–289.

345
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Stability problems in slopes of Arenós reservoir (Castellón, Spain)

J. Estaire & J.A. Díez

Laboratorio de Geotecnia, CEDEX, Mo de Fomento, Madrid, Spain

C. Olalla
ETSI Caminos, Canales y Puertos, Universidad Politécnica de Madrid, Madrid, Spain

ABSTRACT: This paper presents the analysis of the stability conditions of two slopes situated in Arenós
reservoir, taking too into account that a town, named Puebla de Arenoso, is located on the top of one of the
two slopes. With the information provided by the geotechnical investigation carried out, some soil profiles
were elaborated that made it possible to identify the types of soils existing in the slopes: sandy coluvial sed-
iments on a 10 m thick layer of altered marls. The interpretation of all the data available made it possible
to deduce that the slope was in a quite strict equilibrium situation that produced movements in the colu-
vial sediment layer amplified by the successive increases and decreases of water level in the reservoir. The
slope situated in front of Puebla de Arenoso was also studied. This slope suffered a great dimension slide
at the beginning of Quaternary, whose stability should be verified. In this case, the main problem that can
be originated by a reactivation of that huge slide was the formation of a wave that could affect some build-
ings situated in low parts of the town or even the dam. To face these problems, some soil treatments were
carried out.

1 INTRODUCTION shown in Figure 1, are slopes 1 and 2, that suffered

a sliding at the beginning of Quaternary and a small
Arenós dam and reservoir are situated in Castellón, reactivation in 1989, and slope 4 in whose top Puebla
at the east of Spain, on Mijares river. The slopes de Arenoso is situated. In the paper those slopes are
of the reservoir have presented, since dam construc- referred as ‘‘paleoslide’’ (slopes 1 and 2 and ‘‘Puebla
tion, instability problems and slides. Among those de Arenoso slope’’ (slope 4). Slope 3 was also stud-
slides, there are some of great dimensions that limit ied but it does not have any important problem of
the maximum volume of water that can be stored in stability.
the reservoir. The slopes studied in this paper, as

2 DESCRIPTION OF PROBLEMS

The potential instability of those slopes make neces-

sary to analyze the following aspects:

a. Determination of the possibility of a reactiva-

tion of the great slide occurred in Quaternary
that could affect almost all the surface of slopes
1 and 2.
b. Determination of the possible maximum volume
of the slides with highest probability of occur-
rence, taking into account the present geomor-
Figure 1. Diagram showing potential slides in the studied phological and hydrological conditions of the
area. slopes.

347
c. Determination of the velocity of the material flow
in its way to the reservoir, in case of a slide occur-
rence. That velocity is one of the factors with more
influence in the height of the waves produced in the
water stored in the reservoir.
d. Determination of the possibility that the material
fallen from the slopes could form an artificial dam,
whose failure could produce very big waves with
a great destructive potential, affecting mainly the
low parts of the town or the dam.
e. Determination of the stability condition of slope 4
and the definition of the remedial treatments, taking
into account its importance as there are some houses
inhabitant in its top.
Figure 2. Position of boreholes drilled in Puebla de
Arenoso.
3 GEOTECHNICAL INVESTIGATION

The works performed to analyze the stability of the 4 SOIL PROFILES

different slopes are described below.
Based on the data collected in the geological and
3.1 Paleoslide (slopes 1 and 2) geotechnical investigation, some geological soil pro-
files were made.
a. Geological study: it consisted in the interpretation
of stereoscopic pairs of photos and ‘‘in situ’’ ground
4.1 Paleoslide (slopes 1 and 2)
investigation, whose data were used to elaborate
geomorphological maps at 1:10.000 and 1:5.000 In slopes 1 and 2, situated in front of Puebla de Arenoso
scales. and in the left margin of Mijares river, there is a great
b. Campaign of geotechnical investigation: five bore- slide formed by a principal, old and of great dimension
holes were drilled, of more than 100 m length each. slide and other minor slides produced in its toe. The
Inclinometric tubes were installed in all the bore- principal slide possibly occurred at the beginning of
holes. Before installation, narrow and lengthwise Quaternary epoch, during Pleistocene, that means it is
incisions were made in the tubes to be used also as between 1, 8 and 0, 07 million of years old. Afterwards,
piezometric tubes to analyze the variations in water and during recent Holocene, other minor slides were
level. The surrounding hole was filled with fine generated in the slope toe due to the erosive action of
gravel to allow water go into the tubes. Mijares river. Nowadays that zone is covered by water
c. Control of water levels in the boreholes. stored in Arenós reservoir. In 1989 a rotational slide
d. Search of the eleven boreholes drilled in 1989, after occurred in the slope toe after a period of heavy rains.
the reactivation of a slide in the toe of slope 1 and Those slides can be seen in Figure 3.
analysis of their present state. The upper part of the great paleoslide is situated
e. Campaign of laboratory tests made mainly with between 750 and 850 m, above sea level, and its toe is
samples obtained in the altered rock substratum. in the bottom of Mijares river valley, at about 540 m,
above sea level, as can be seen in Figure 4.
It is interesting to remark that, according to different
3.2 Puebla de Arenoso slope
studies, the flow of this slide closed the valley and went
a. Control of the present state of the 21 boreholes up in the other margin. The rests of this flow form the
drilled in 1972, in the town, before the construc- area in which Puebla de Arenoso is situated, about
tion of the dam. The position of those boreholes 50 m above the bottom of the valley.
can be seen in Figure 2. The slope is formed by different slide masses that
b. Drill of other six boreholes equipped with inclino- suffered several readjustments. The slide material is
metric tube to be used also as piezometric tube. Its formed by an up to 90 m thick disorganized mass
position can also be seen in Figure 2. with metric and decimetric loose limestone blocks,
c. Inclinometric measurements were made, one per embedded in sandy-clay matrix.
month, during six months. The base of the paleoslide is formed by Aptiense
d. A map of cracks existing in the buildings and marls and, in the toe area, by Weald clays, that they
streets of the town was made to determine the main are in contact with the marls by a fault. In some bore-
directions of the ground movements. holes, metric and decimetric thick breccia zones were

348
Figure 3. General view of the paleoslide.

Figure 4. Soil profile of the paleoslide.

Figure 5. Aspect of the breccia material found in boreholes. Figure 6. General view of the town near the top of the slope.

detected, near the contact between the slide mate-

rial and the substratum. The slide movement very of the slope, the piezometric level coincided with the
probably occurred through those breccia zones which reservoir level.
are formed by angled fragments of marly limestone,
of milimetric and centimetric size, embedded in a
4.2 Puebla de Arenoso slope
marly-clay, blackish grey matrix, as it can be seen in
Figure 5. As it was said previously, Puebla de Arenoso is located
The piezometric level in the upper part of the slope on slide materials belonging to the toe of the paleoslide
was in the contact zone between slide material and occurred in the opposite margin of Mijares river, as it
substratum. In the boreholes situated in the low part can be seen in Figures 3 and 6.

349
 Table 1. Material geotechnical parameters deduced by
stability back analysis.

Cohesion Friction
Material (kPa) angle (◦ )

Slide materials 15 28.0

Substratum altered layer 0 12.5
Substratum Weald clays 0 25

Table 2. Different hypothesis made in the stability analysis.

Figure 7. Soil profile of Puebla de Arenoso slope.
Reservoir level (m) Phreatic level Safety factor

575 Flat 1.06

Figure 7 shows the geological soil profile of Puebla 575 High 1.08
de Arenoso slope. The average thickness of slide mate- 600 Flat 1.05
rials is about 50 m, with a detected maximum value of 600 High 1.08
56 m. Great blocks and rocky fragments in a chaotic 600 Fast drawdown 1.01
disposition, embedded in a marly-clay matrix, form
the slide materials. The substratum, formed by marls
that are altered in its upper part, appears below. to one corresponding to a partial reactivation of the
The points where the boreholes made in 1972 are cut slide occurred in Holocene.
are also drawn. Those points form a possible cinemat- The slight difference between the results obtained
ically sliding surface, whose top coincides with some is due to the fact that the volume of materials affected
of the cracks observed in that slope area. This sliding by an increase in the reservoir water level is relatively
surface goes through the mass of slope materials and small.
it has its toe some meters above the river level. The main conclusion of this second set of calcu-
lations is that present stability situation of the toe of
the slope is quite precarious, as the maximum safety
5 DATA ANALYSIS factor is below 1.1.

5.1 Paleoslide (slopes 1 and 2)

5.2 Puebla de Arenoso slope
The first step was to perform 2D stability back analysis
of the different slides occurred in the slope: the great The potential sliding surface drawn in Figure 7 seems
paleoslide occurred in Quaternary, the middle sized to indicate that the slope is in a very precarious equilib-
slide produced in Holocene and the last and smallest rium situation that produces readjustment movements
one, occurred in 1989. in the slide mass, increased by the successive oscil-
In these calculations, using Morgenstern-Price lations of the reservoir water level. However, it is
method, the sliding surface of the great paleoslide was important to remark that, since the drilling of the bore-
supposed to develop through the altered rock substra- holes in 1972, no important slide has occurred in this
tum layer. In the other slides, the sliding surface can slope.
be seen in Figure 4. The stability back analysis performed with that sur-
The results of those calculations were the geotech- face show that the slide material has the following
nical parameters of the materials existing in the slope strength parameters: cohesion between 0 and 10 kPa,
whose values are given in Table 1. combined with friction angles ranging between 20 and
The main conclusion of these first analyses is that 22◦ . Those values are in correspondence with the char-
the reactivation of the paleoslide is not probable as the acteristics of the matrix of the slide mass formed by
strength values of the altered substratum layer deduced sandy-clay coluvial with a slight cohesive component.
from stability calculations are clearly lower than the Those values could not be contrasted with laboratory
ones determined in laboratory (c = 10 kPa; φ = 18◦ ). tests due to the great size of the limestone blocks
The second step was to analyse the present slope embedded in the sandy-clay matrix.
stability situation, in the calculation hypothesis shown On the other hand, in the inclinometric measure-
in Table 2. ments some movements were detected in three of the
The first results of this second set of calculations is boreholes, at depths corresponding with zones near the
that the most probable future sliding surface is similar contact between the slide materials of the paleoslide

350
and the grey marls of the substratum. These measure- 7.1 Installation of auscultation
ments were performed during a period of draught and
14 GPS stations were installed to monitor the slope
with a slow and progressive decrease in the reservoir
movements, as it can be seen in Figure 8, (Solanes,
water level.
2007). Besides, the inclinometers, installed during the
geotechnical investigation campaign, are read period-
ically. By now, the movements detected do not exceed
6 CONCLUSIONS OF THE STUDY
5 mm and are quite slow.

6.1 Paleoslide (slopes 1 and 2)

7.2 Analysis of wave height
a. It is not probable that the great paleoslide has a
sharp and sudden reactivation, without some slow Some calculations were performed to determine the
movements appear previously in the slope that make wave height and the transient elevation of the reser-
it possible to take some remedial treatments. voir water level due to a more or less sudden
b. However small movements at the toe of the great entrance of material, coming from some of the
paleoslide can be produced, as it occurred in 1989. potential unstable slopes of the reservoir, (Segura,
Taking into account the precarious stability equi- 2007).
librium of the toe of the slope, a period of heavy As the velocity of the potential slopes depends on
rains can be an important activation factor for those some physical parameters that are difficult to quantify,
small movements to be produced. calculations were made with a wide range of times of
c. The estimated volume of that future potential slide material flow (3, 6, 10 and 30 minutes).
at the toe is about 2,1 millions of m3 . Based on Before that, a data recompilation from reference
geological considerations, it is estimated that the literature was made. The values of velocity of material
volume of material that could reach the reservoir flow in landslides similar to the one studied here are
would be 1,55 millions of m3 . represented in Figure 9 (Segura, 2007).
d. Taking into account the characteristics of the mate- It can be seen that the velocities used in this study
rials existing in the slope, it is not foreseeable that are clearly greater to the ones measured in most of the
its flow into the reservoir was going to be quick real cases referenced.
enough not to be detected by a intensive system of
auscultation, before a sharp and sudden movement
was produced.

6.2 Puebla de Arenoso slope

a. The analysis performed with the boreholes drilled
in 1972 makes it possible to draw a possible cine-
matically sliding surface, whose top coincides with
some of the cracks detected in that zone of the slope.
This sliding surface goes through the mass of slide
material existing in the slope and it has its toe some
meters above the reservoir water level.
b. The interpretation of the inclinometric measure-
ments performed in the boreholes of the town,
during a period of draught and with a slow and
progressive decrease in the reservoir water level,
made it possible to determine that the slope was
in strict equilibrium. Those measurements indi-
cated that movements are mainly produced in the
contact between the slide material and the marly
substratum.

7 DESCRIPTION OF SOLUTIONS

Taking into account the previous conclusions, some

remedial measurements were taken to try to solve the Figure 8. View of one of the GPS station installed in the
stability problems of the slopes. slope.

351
 This 190 m long structure is formed by a slab of a
variable thickness between 1.6 and 2.0 m, and verti-
cal walls of height between 8 and 10 m and thickness
between 1.2 and 2.0 m. Their main function is to
avoid water erosion of the stabilization fills placed in
both margins. To design this structure some numeri-
cal hydraulic simulations were performed and a scaled
physical model was tested.
Figures 12 and 13 show a general view of the works
performed in the Puebla de Arenoso slope.

Figure 9. Reference values of velocity of material flow in

landslides (Segura, 2007).

In spite of it, the obtained results indicate that the

maximum reservoir water level is not greater, in any
case, to 3 m, in Puebla de Arenoso slope, and to
1 m, in the dam. These values must be considered Figure 11. Section of the works performed in Puebla de
as acceptable as the existing clearance are greater in Arenoso slope.
both zones.

7.3 Remedial works

The solution for the stabilization of Puebla de Arenoso
slope consisted in the construction of fills made with
granular soils, protected by rockfills, in both margins
of the river and of a concrete structure to channel the
river (Solanes, 2007). A sketch of the solution can be
seen in Figures 10 and 11.
These fills, with stabilization effects in the toes of
both margins, are between 30 and 40 m high. Their
inclination is 2H:1V. They have draining layers to
dissipate water pressures in case of a fast drawdown.
In the riverbed, a layer of 5.5 m of granular mate- Figure 12. General view of the works performed in Puebla
rial was placed as a foundation layer of the concrete de Arenoso slope.
structure.

Figure 10. Plan of the works performed in Puebla de Figure 13. General view of the slope after the end of the
Arenoso slope. works.

352
ACKNOWLEDGEMENTS de Arenós (Castellón). Jornadas Técnicas sobre Estabil-
idad de Laderas en Embalses. pp 295–332. Zaragoza.
The authors of this paper want to acknowledge Conf. Hidrográfica del Ebro.
Mr. Julián Cuesta, geologist of Eptisa, due to his Segura N. & Fernández L. 2007. Estudio de la variación del
effort during the geological study, and Mr. Fernando nivel de embalse producida por la entrada de material
de laderas en el embalse de Arenós. Jornadas Técnicas
Solanes and Miss María Irles of Spanish Ministerio sobre Estabilidad de Laderas en Embalses. pp 455–476.
de Medio Ambiente (Ministry of Environment) due Zaragoza. Conf. Hidrográfica del Ebro.
to their support and help during the execution of the
works.

REFERENCES

Solanes F., Calderón P., Escuder I. & Martínez J. 2007. Esta-

bilización de la ladera de Puebla de Arenoso en el embalse

353
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

The 22 August, 2006, anomalous rock fall along the Gran Sasso NE wall
(Central Apennines, Italy)

G. Bianchi Fasani, C. Esposito, G. Scarascia Mugnozza & L. Stedile

‘‘Sapienza’’ University of Rome, Department of Earth Sciences and ‘‘CERI’’ Research Center on Geological Risks,
Rome, Italy

M. Pecci
IMONT, National Mountain Institute of Italy, Rome, Italy

ABSTRACT: It is described the rock fall event occurred along the Gran Sasso massif (Central Apennines, Italy)
on 22 August, 2006, when a limestone block, with an estimated volume of about 30,000 m3 , fell from the sub-
vertical NE wall nearby the Corno Grande peak, the highest peak of the Italian Apennines. Despite the small rock
volume involved in the landslide, the rock fall deposits covered an area of about 35,000 m2 , a giant and abrasive
dust cloud was generated by the atmospheric pressure waves (air blasts) induced by the rockfall impact and
determined destructive effects over an area of about 110,000 m2 at the base of the slope. Moreover the dust cloud
covered a distance of about 3 km, thus reaching the village of Casale San Nicola and the A24 motorway that was
temporarily closed for security reasons. The seismic noise generated by the rock fall was recorded by the National
Institute of Nuclear Physics seismometric devices located in the Gran Sasso underground laboratories (LNGS).

1 INTRODUCTION history with previous unusual rock falls (Morrissey

et al. 1999; Wieczorek et al. 2000) in order to predict
On August 22nd, 2006 a limestone block of about hazard scenarios potentially induced by future rockfall
30,000 m3 detached from the NE slope of the events along the Gran Sasso NE wall.
Gran Sasso d’Italia (the Apennines’ highest peak,
2912 m a.s.l.) at an elevation of 2,800 m asl and was
involved in a rock fall along the 1,500 m high cliff,
also known as the ‘‘Paretone’’ (big wall). Because 2 GEOLOGY AND CLIMATE
of the huge dust cloud generated by the rock fall,
the motorway A24, located few kilometres down- The unique landscape of this zone within the Apen-
hill, was closed for security reasons. The seismic nines is dominated by the Gran Sasso Massif that
noise generated by the rock fall until the final impact lies just like a ‘‘huge boulder’’ (from which the name
over the cliff base was recorded by the seismomet- itself) (Fig. 1) made of Triassic and Jurassic mas-
ric devices located in the underground laboratories sive limestone and dolomite over relatively smooth
(LNGS) run by the National Institute of Nuclear valley slopes incised in bedded marly-limestone and
Physics. The rock block involved in the fall con- sandy-marlstone.
sists of Triassic dolomitic-limestone which thrust The geomorphological setting strictly reflect the
marly-calcareous and marly-arenaceous formations of structural setting of the area: it is featured by numer-
Cretaceous-Tertiary age within a very complex struc- ous thrust sheets, superimposed one each other during
tural setting. Prompt site investigations allowed to the Pliocene compressive tectonic phases (Ghisetti &
estimate the volume of the felt rock block by means of Vezzani 1990, Calamita et al. 2002).
laser telemetry measurements from the nearest peak; The more evident sheet crossing the sub-triangular
the effects induced by the impact of the rock fall were wall of the Corno Grande (Fig. 1) is also the easier
also observed and the run out area at the steep cliff toe mountaineering route to reach the top. The so called
measured as well. In this paper we analyse the possi- ‘‘Jannetta route’’ overlaps the main overthrust, bring-
ble causes which led to the slope failure, the dynamics ing into contact the stratigrafical-structural Corno
of the fall and the mechanisms which generated the Grande sub unit (sector above the Jannetta route, con-
giant dust cloud. We also compare the present case taining the rock failure) with the same formation, but

355
 system elevated the Corno Grande structural high dur-
ing the extensional phase at the beginning of medium
Lias and have been re-activated during the Messinian
by the Apennine compressive tectonic phase.
Glacial and periglacial processes typify the summit
area of the massif which hosts the southernmost glacier
(Calderone) in Europe, in a fast reduction phase (Pecci
2007).
Extensive historic and prehistoric rockfall and debr-
isflow deposits have accumulated at the base of the
wall (D’Alessandro et al. 2003), where some vil-
lages and important lifelines are located. In fact,
earthquakes, snowmelt, freezing-melting cycles and
frost-wedging effects have caused rock falls in the area,
like the historically recorded 1897 event which made
Figure 1. Virtual 3D view of the Gran Sasso NE wall (from a scar known as ‘‘Farfalla’’ (Butterfly) due to its shape
the website http://maps.live.com). (see the circle in Figure 2). The NE wall is a very popu-
lar and very challenging climbing cliff, thus imposing
further risk conditions (Amanti et al. 1994)

3 DESCRIPTION OF THE EVENT

The event has been observed just after the release

until the impact (Fig. 3), while time histories (Fig. 4)
recorded by the seismic antenna at LNGS (Laboratori
Nazionali del Gran Sasso—National Laboratories of
the Gran Sasso) and field investigations helped in bet-
ter frame the dynamics of the rock fall. It occurred at
7.30 am GMT, when a 30,000 m3 rock block detached
from the ‘‘Guglia Bambù’’ (Bamboo Pinnacle) along
a slickensided fault surface.
After the failure, the already jointed rock mass
underwent a crumbling process while experiencing a
fall as high as 1,500 m along the steep cliff which
features this Gran Sasso slope. Along this section
the debris motion took place mainly by free fall and
bouncing. In addition, shortly after the detachment the
falling mass split into two parts: a first one fell down
along an almost straight, ballistic trajectory, while a
second one was channelled within a chute (Vallone
Jannetta; see the white arrows in Figure 2) following
a more winding path and entrapping rock fragments
and debris along the path.
In both cases, the fallen rock fragments reached
the base of the steep slope where they spread over an
area of about 35,000 m2 ; a part of the debris was then
channelled within the upper section of a deeply cut
valley. In addition, immediately after the rock impact
Figure 2. Frontal view of the landslide paths along the rock at the wall base, a huge and dense dust cloud devel-
slope. oped and rapidly moved downhill covering a distance
of about 3 km until the A24 motorway and the vil-
lage of Casale San Nicola (Fig. 3); because of this
belonging to the stratigrafical-structural Corno Pic- dense and rapidly moving dust cloud the motorway
colo sub unit in the sector below (Adamoli 1992). was temporarily closed for security reasons.
This structural element cuts all the eastern face and The rock fall occurred on a sunny day without
corresponds to an original normal faults system. Such any apparent trigger event such as an earthquake or

356
Figure 3. Photographics sequence of the rock fall event, showing the spreading of the dust cloud.

Figure 4. Seismogram recorded by the LNGS in concomi-

tance with the rock fall event.

Figure 5. Rock free face of the detached block from

different points of view.
heavy rainstorm. During the preceding days some
rainfalls were recorded, due to some afternoon thun-
derstorms, a typical phenomenon during this hot regards the geologic controls on the failure mecha-
period of the year. Dark streaks observable on the nism, the shape and volume of the detached block is
detachment free face indicated paths where water had clearly guided by the presence of tectonic slickensided
seeped through the jointed rock mass (Fig. 5). As surfaces (Fig. 5); the jointing conditions within the

357
rock mass are on their turn connected with the local
structural setting, featured by tectonic lines of regional
importance.
Laser-GPS telemetry measurements from the near-
est peak also allowed to determine the actual detached
rock volume (30,000 m3 ), once compared with images
showing the shape of the Bamboo Pinnacle before the
fall. The remainder of the pinnacle still in place is an
overhanging rock dihedral estimated to be 20,000 m3
in volume. Site investigations along the steep cliff
pointed out the numerous joints within the rock mass
and the presence of unstable rock blocks.

4 RECORDED EFFECTS

The most significant feature of the 22 August rock

fall is definitely represented by the high energy Figure 6. Aerial view showing the extension of the rock fall
dissipation derived from the impact and rock frag- deposit and the zone affected by air blast.
mentation at the slope toe and the so generated
dust cloud (Fig. 3). Actually, according to the
above reconstructed kinematics of the event and as
inferred from the time histories recorded at LNGS
(Fig. 4), we hypothesize two distinct impacts due
to a gap in the arrival times by the rock frag-
ments which followed the two different paths. As
regards the mechanisms involved in such highly dis-
sipative process, it can be assumed that the energy
of the fallen rock volume was transferred to both
ground and atmosphere, thus creating air pressure
waves (airblast effect) and a consequent dense dust
cloud.
The latter has been driven by the airblast with high
velocity and has been able to preserve an abrasive
effect over a quite large area (110,000 m2 ) (Fig. 6),
and reached the village of Casale San Nicola located
3 km downhill the impact area. Such an abrasive effect
is shown by the uprooting and snapping of hundreds of Figure 7. Abrasive air blast effect on the vegetation.
trees and bushes as well as by the debarking (Fig. 7).
Airblasting is usually generated by landslides involv-
ing rock volumes in the order of 106 m3 . Only few
cases have been previously reported (e.g. Yosemite, a reduction in rock element dimensions versus dis-
10 July 1996 event by Morrissey et al. 1999 and tance from impact point are recorded. The larger land-
Wieczorek et al. 2000) regarding relatively small lside debris channeled within a stream and came to
magnitude landslides able to generate air pressure halt 250 m far from the impact point, while the finer
waves. material blanketed areas within 3 km distance.
In both cases the airblast generation is due to the
high impact velocity and the two separated impacts,
4.1 Considerations about further risk scenarios
as inferred from the seismograph time histories. Dur-
ing the time interval between the impacts, the initial In addition, during the survey phases that were car-
atmospheric conditions were changed after the forma- ried out to better reconstruct the discussed rock fall,
tion of a first dust cloud. Such high density fluid could it was possible to preliminarily define the overall geo-
have favoured the sudden variation in air pressure morphic, geomechanical and, finally, slope stability
(Morrissey et al. 1999). conditions along the whole NE wall of the Gran Sasso.
The rock debris within the landslide deposit shows For this purpose, data derived from direct surveys in
grain size within fine sands and boulder (1 m3 ), and the accessible areas were coupled with data inferred

358
by remote techniques (such as laser telemetry), per- to be characterized by frequent, small-sized rock fall
formable through devices easy to transport on steep events, the potential occurrence of massive rock slope
and quite unstable tracks. As a result, other blocks failures involving the ‘‘pillars’’ under specific condi-
within the rock mass prone to detachment have been tions, could evolve in a dry granular flow. As a matter
recognized. Even if most of the so identified blocks of fact, the presence of deeply incised channels at the
have dimensions quite comparable with those of the base of the wall can represent the geomorphic ‘‘con-
hereby presented rock fall, it was possible to observe straint’’ able to convey the highly fragmented debris,
also some huge rock ‘‘pillars’’ with volumes rang- allowing for a long run-out (Fig. 9).
ing between 105 m3 and 106 m3 (Fig. 8). Furthermore,
field surveys pointed out the presence of even large-
sized (tens of cubic meters) limestone blocks also in 5 CONCLUDING REMARKS
some areas downslope of Casale San Nicola village
and the motorway, thus testifying the occurrence of As before the event no earthquake was recorded, no
past massive and catastrophic rock slope failure. seismic trigger can be invoked for the failure. The only
Based on this evidence, it is possible to hypoth- possible causes can be ascribed to the intense jointing
esize risk scenarios for both the motorway and the within the rock mass and to the frost (thermal) wedging
village. Even if the gravity-induced landscape evo- and to the possible permafrost degradation phenom-
lution under the present boundary conditions seems ena. The latter is justified by the progressive retreat
of the Calderone glacier nearby the source area at the
same elevation (Pecci 2007). According to recent stud-
ies in the Alps (Davies et al. 2001, Fischer et al. 2006),
rockfall events can be referred to climate changes in
high mountain environment, thus determining a signif-
icant permafrost or seasonal ice pattern modification.
Annual mean temperature can influence the stability
of rock masses with ice bearing joints. This can be
related to the water seepage evidences observed on the
free face of the 22 August event. As a consequence,
the presented case history could be considered as a fur-
ther example of rock slope failure induced by climatic
change in a very sensitive environment such as the
high peaks of the Gran Sasso Massif, the southernmost
glacial and periglacial environment in Europe.

REFERENCES
Figure 8. DEM adopted for the volume calculations of the
unstable blocks.
Adamoli, L. 1992. Evidenze di tettonica d’inversione
nell’area del Corno Grande-Corno piccolo (Gran Sasso
d’Italia). Bollettino della Società Geologica Italiana III:
53–66.
Amanti, M., Pecci, M., Scarascia Mugnozza, G. & Vallesi, R.
1994. Comparison and critical review of quick field data
collection methods on rock slopes: a contribution from
climbing techniques and experiences. Atti del Convegno
‘‘Man and mountain ’94’’, 20–24 giugno 1994, Ponte di
Legno (bs), pp. 189–198.
Calamita, F., Scisciani, L., Adiamoli, M., Ben M’Barek, M.
& Pelorosso, M. 2002. Il sistema a thrust del Gran Sasso
d’Italia (Appennino Centrale). Studi Geologici Camerti
1/2002: 19–32.
D’Alessandro, L., De Sisti, G., D’orefice, M., Pecci, M. &
Ventura, R. 2003. Geomorphology of the Summit Area
of The Gran Sasso d’Italia (Abruzzo Region, Italy).
Geografia Fisica e Dinamica Quaternaria 26: 125–141.
Davies, M.C.R., Hamzal, O. & Harris, C. 2001. The effect
Figure 9. Possible morphologic constrain able to produce of rise in mean annual temperature on the stability of
the channelling of granular flow deriving from larger sized rock slopes containing ice-filled discontinuities. Periglac.
landslide event. Process 12: 137–144.

359
Fischer, L., Kaab, A., Huggel, C. & Noetzli, J. 2006.Geology, Pecci, M. 2007. The shrinkage of the central Mediterra-
glacier retreat and permafrost degradation as controlling nean cryosphere in a changing mountain environment.
factors of slope instabilities in a high-mountain rock wall: Mountain Forum Bullettin VII Issue 2, ISSN 1029–3760
the Monte Rosa east face. Nat. Hazards Earth Syst. Sci 6: (http://www.mtnforum.org/rs/bulletins/mf-bulletin-2007-
761–772. 07.pdf).
Ghisetti, F. & Vezzani, L. 1990. Stili strutturali nei sistemi Wieczorek, G.F., Snyder, J.B., Waitt, R.B., Morrissey, M.M.,
di sovrascorrimento della Catena del Gran Sasso (Appen- Uhrhammer, R.A., Harp, E.L., Norris, R.D., Bursik, M.I.
nino Centrale). Studi Geologici Camerti vol. spec. 1990, & Finewood, L.G. 2000. Unusual July 10, 1996, rock fall
37–50. at Happy Isles, Yosemite National Park, California. GSA
Morrissey, M.M., Savane, W.Z. & Wieczorek, G.F. 1999. Air Bullettin 112: 75–85.
blast generated by rockfall impacts: Analysis of the 1996
Happy Isles event in Yosemite National Park. Journal of
Geophysical Research 104, n◦ B10: 23189–23198.

360
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

New formulae to assess soil permeability through laboratory identification

and flow coming out of vertical drains

J.C. Gress
Ecole Nationale des Travaux Publics de l’Etat, Lyon, France

ABSTRACT: Mastering a lot of landslides through deep dewatering drainage by one or more lines of drains,
being pumped either by siphoning pipes or by electropneumatic pumps® , it has appeared to us that water tests
were not reliable for different reasons. We suggest here a procedure through laboratory identification tests and
new formulae to assess the permeability, we can wait for, and a new formula to assess the flow which will come
out of a line of vertical drains in a slope.

1 INTRODUCTION – reach the aquifers to be drained,

– be bored with the good tools, in order to avoid the
For past twenty years, the mastering of landslides, in decrease of permeability,
France, has taken great benefit of the experience of – must be equipped like wells (proper slotted pipes,
a great number of works of deep drainage through filter).
siphon drains® or electropneumatic drains® .
But it has clearly demonstrate that it was difficult to The efficiency between adjacent drains must be
have a good estimation of the flow, we could wait for. designed; this will lead to a better assessment of the
After having analysed why, we propose hereunder new spacing between drains.
formulae, in order to have a better approach of deep If the wanted drawdown, under the soil surface, is
drainage efficiency through lines of vertical drains, less than eleven meters, then the drain will be pumped
formulae to assess soil permeability through identifi- through siphoning pipes, a hydraulic accumulator,
cation tests and a formula to assess flow coming out
of vertical drains in a slope.

2 BASIC PRINCIPLES OF VERTICAL DRAINS

Lines of vertical drains are placed, on the site, in order

to dewater the landslide and stop the instability.
These drains, with an average spacement of 5
meters must:

Manhole Ø 800

Siphoning pipe

Water table before treatment

Outlet manhole Automatically regulated
Drain flushing system
Sliding surface
Water tab
le
after tre
atment
Horizontal reference plane

Tank permanently full of water

Scale: 1 meter

Figure 2. Electropneumatic pump® and view of the com-

Figure 1. Cross section through a siphon drain® . pressor chamber.

361
equipped with a flushing system, regulating the flow The oedometer formula:
above the critical one.
If it is greater, then we use electropneumatic log k = Cke (2)
pumps® , air under pressure being fed by a compressor
placed in a chamber, the pump being equipped with a with Ck ≈ 0.5eo (Tavenas, F. et al. 1983) is giving
sensor, analysing if the pump is empty or full and then only a variation of k with the void index e for clayey
regulating the feeding with air. soils.
Then, we can pump up to depth of 100 meters. We have worked on correlations, we had through
water tests and identification tests of different soils and
on the works of Nagaraj et al. 1986 and Sivapullaiah
3 CLASSICAL MEASUREMENTS OF THE P.V., et al. 2000.
PERMEABILITY It appears that the formulae proposed hereunder fits
relatively well, correlating log10 k with WL = liquidity
In order to study the feasibility of such dewatering limit, when the particle of the soil have a maximum
scheme, there is always a preliminary geotechnical size of 400 μm; otherwise, VBS = methylene blue
and hydrogeological study with borings, samplings, value of the total soil; e = void index, %2μ = percent
laboratory tests and water tests on site and in the finer than 2 μm.
laboratory. The methylene blue value is very frequently mea-
It appears very often that water tests didn’t give a sured in France. The methylene blue value of the
good order of the real permeability, because of: 0/400 μm fraction is well correlated to the plasticity
index Ip and to the liquidity limit WL :
– bit to do the boring being not the good one,
– too small diameter, Ip ≈ 0.045 VB0400μ (3)
– water test run in an injection way and not in a
pumping way, WL ≈ 0.14 + 0.063 VB0400μ (4)
– had we to run a Lefranc or Nasberg test? the methylene blue value of a granulometric portion
o/d being linked to the percent finer than d through
These difficulties can be mastered. But it stays the
the formulae:
major difficulty, that is to say, that the permeability
is not homogeneous and can vary vertically through VBod x % od
the same geological layer, due to the variation of the
clayey fraction or to the superposition of layers of sand = VB0400μ x % 400μ = VB02μ x % 2μ (5)
and clay, the thickness of these layers being maybe less
than decimetric. When the maximum size of the soil particle is
To put this in evidence, the best thing is to do 400 μm, we propose the formulae hereunder, in order
intact continuous samplings and to analyse the vari- to have a rough estimation of soil permeability:
ation of density, granulometry and qualify the activity
of clay through Atterberg limits or methylene blue if WL < 0.25
tests. We have then tried to obtain formulae giving a log k = −(1.41 + 25.55 WL ) + (4.46 − 3.5 WL )e
rough assessment of permeability through the different (6)
parameters measured.
if 0.25 ≤ WL < 0.80

4 ASSESSMENT OF THE PERMEABILITY log k = −(5.23 + 9.2 WL ) + (4.6 − 4.11 WL )e (7)

THROUGH SOIL IDENTIFICATION
When the maximum size D of the particles is greater
PARAMETERS
than 400 μm, then we propose:
There are not much formulae allowing us to have a if VBS < 1.5
good assessment of the permeability, these formulae
working for gravely soils to clay. log k = −(4.99 + 1.61 VBS) + (3.97 − 0.22 VBS)e
The Hazen formula is well known: (8)

k = K(D10 )2 (1) if 1.5 ≤ VBS < 10

log k = −(6.52 + 0.58 VBS) + (4.03 − 0.259 VBS)e
k in cm/s and D10 being the diameter in cm of the (9)
screen allowing ten per cent in weight of the soil to go
through. But it works only for sands and sandy gravels. where VBS is the methylene blue value of the total soil.

362
 e
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 Horizontal initial
0
Water L piezometric surface
feeding line
-1 Qw
-2

-3
VBS = 0.1 2rw
VBS = 0.5
-4
VBS = 1
-5
H
-6 VBS = 1.5
hw
VBS = 3
-7

-8 VBS = 6

-9
VBS = 8
Figure 4. Cross section after dewatering (horizontal initial
-10 piezometric surface).
-11 VBS = 10.5
log(k)
-12

Figure 3. Values of log10 k with e and VBS.

water feeding line

Spring line
We can then with these formulae, through simple
laboratory tests, check the in-situ water tests and have a L
a relatively better knowledge of the different levels of
permeability.

Well (drain)
5 NEW FORMULA TO ASSESS THE FLOW
COMING OUT OF THE DRAINS
Figure 5. Overview of a set of drains.
For a line of drains, dewatering an horizontal piezo-
metric surface, for an unconfined aquifer, the flow of
each drain is given by:
  Initial
Parabola
2Qw L Qw a
H 2 − h2w = + Ln (10) q1=Qw/a
q0=poHok
ka πk 2πrw po

where Qw = flow coming out of each well; rw = H0

Initial

radius of each well; k = permeability of the aquifer;

upstream
hw flow

a = distance between each drain. X

q2

Each drain penetrating totally the aquifer, and a x 0

feeding line being located at a distance L: Surface of drawdown 0

In case of an inclined piezometric surface having

an initial po slope, we propose this formula: Figure 6. Cross section after dewatering (surface of draw-
down).
Ho2 − h2w
qw =  (11)
Ho +hw
a po k + π1k Ln 2πarw
The substratum is supposed to be horizontal.
where Ho = total initial height of the sheet of water; Example if:
hw = thickness of the dewatered sheet; po = initial Ho = 20 m
slope of piezometric surface; a = distance between hw = 10 m
each drain; k = permeability; rw = radius of the po = 8%
drain. a=5m
The initial upstream flow qo is equal to po Ho k. The
k = 5 × 10−6 m/s
flow coming out of the vertical drains per meter of line
rw = 0.08 m
is equal to q1 = Qw a .The downstream flow q2 is equal
to q0 − q1 . then Qw = 71 liters per hour.

363
6 CONCLUSIONS Clark, A.R. et al. 2002. The planning and development of
a coast protection scheme in an environmentally sensi-
Vertical drains, either drained by siphoning pipes® or tive area at Castlehaven, Isle of Wight. Proc. Int Conf
electropneumatic pumps® , have allowed the stabiliza- on Instability, Planning & Management, Thomas Telford,
tion of more than too hundred lanslides, these last 2002.
Clark, A.R. et al. 2007. Allowing for climate change;
twenty years. an innovative solution to landslide stabilisation in an
The hereabove proposed formulae of the perme- environmentally sensitive area on the Isle of Wight Inter-
ability and of the flow coming out of the drains should national Conference on ‘Landslides and Climate Change-
lead to an improvement in the design of the scheme Challenges and Solutions’ Ventnor, Isle of Wight, UK.
of drains, these formulea having to be adjusted to the Gress, J.C. 1996. Dewatering a landslip through siphon-
further experience of new works. ing drain® . Ten years experiences. Proc 7th International
Symposium on Landslide. Trondheim.
Gress, J.C. 2002. Two sliding zones stabilized through siphon
REFERENCES drains® . International conference on Landslide, slope
stability of infrastructures. Singapor.
Nagaraj, T.S. et al. 1993. Stress state—permeability relation
Bomont, S. 2002. Drainage with electropneumatic drains® . for fine grained soils. Geotechnical (43): 333–336.
Conference JNGG 2002. Nancy, France. Mitchell, J.K. 1993. Fundamentals of soil behavior. 2nd ed.
Bomont, S. 2004. Back experience from four landslides sta- John Wiley & Sons, Inc., New York.
bilized through lines of siphon drains® in Normandy, Pandian, N.S. et al. 1995. Permeability and compressibiliy
France. 9th International Symposium on Landslides 2004. behavior of bentonite—sand/soil mixes. Geotechnical
ISL Rio. Testing Journal (18): 86–93.
Bomont, S. et al. 2005. Two applications for deep drainage Sivapullaiah, P.V. et al. 2000. Hydraulic conductivity of
using siphon and electropneumatic drains® . Slope works benton. Canadian Geotechnical Journal (37): 406–413.
for Castlehaven Coast Protection Scheme, Isle of Wight Tavenas, F. et al. 1987. State of the Art on Laboratory and
(UK) and slope stabilisation for the Railways Agency, in situ stress strain time behavior of soft clays. LAVAL
France. In, Proceedings of the International Conference University.
on Landslide Risk Management. 18th Annual Vancouver
Geotechnical Society Symposium.

364
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Structure-controlled earth flows in the Campania Apennines

(Southern Italy)

F.M. Guadagno, P. Revellino, G. Grelle, G. Lupo & M. Bencardino

Department of Geological and Environmental Studies, University of Sannio, Italy

ABSTRACT: Slow-velocity landslides predominate in the area of the Province of Benevento, due to the
prevalent clay nature of its outcropping deposits. An analysis of these instabilities, via detailed inventory mapping,
has shown a pervasive diffusion of earth flows, characterized by a reactivation tendency. Their evolution, in terms
of activity and kinematic mechanisms, is structurally-controlled and can be generally connected to three principal
controls: i) bedding; ii) stratigraphic or tectonic contact between lithologically differentiated sequences; and
iii) zones of intense fracturing linked to folds and faults. Differences in pattern and controlling factors led to
their grouping into recurrent types characterised by a different style of evolution.

1 INTRODUCTION stony flysches, resulting from the deformation of

paleogeographic domains (Patacca & Scandone 1989).
Like other sectors of the Italian Apennines, the area of As a consequence, lithologically differentiated
the province of Benevento (Campania region, South- sequences belonging to marine sedimentary rocks,
ern Italy) is characterised by recurring events of slope ranging from Cretaceous to Pliocene, and continen-
instabilities, in time and space, determining conditions tal deposits of Pleistocene, outcrop in the area. They
of high risk. The geological and structural setting of can be grouped as follows (Pescatore et al. 2000):
the Campania Apennines is the basis of landsliding 1) Platform carbonate Units: limestone and dolomitic
processes. Moreover, recent occurrences demonstrate limestone forming the higher slopes; 2) Pre-orogen
and confirm that human activities can have a deter- basin Units: basinal facies and clayey, quarzarenite
minant role upon the morphologic evolution of the and arenite deposits; 3) Synorogen and lateorogen
slopes. The changes linked to increasing urbanization Units: arenaceous and arenaceous-clayey flysch-like
of morphologically complex settings and agricultural successions; 4) Pyroclastic deposits: a) generally
and forestry practices seem to have an important effect. lithoid and coherent, grey-yellowish flow deposits (the
Further problems connected to landslide and ero- Campanian Ignimbrite); and b) incoherent or weakly
sion phenomena are linked to the responses of natural cemented air-fall deposits; 5) Continental, fluvial and
slopes to rainfall regime modifications induced by cli- detrital deposits: debris and colluvial fans at the toe
matic changes, which could produce changes in the
spatial-temporal recurrence.
The principal objective of this paper is to high-
light the influence of the geological and structural
settings on the occurrence of earth flows in the study
area, by using data from a ‘‘Landslide Inventory Map’’
(Guadagno et al., 2006) of the province. These con-
trols affect the evaluation of the susceptibility required
in hazard assessment.

2 THE GEOLOGICAL ENVIRONMENT

The territory of the province of Benevento is located

in one of the most geologically complex areas in
Italy. The Southern Apennines consists of thrust-
belt structures of carbonate terrain and clayey and Figure 1. DEM of the province of Benevento.

365
Figure 2. Map of the lithotechnical sequences outcropping in the province of Benevento, as shown in table 1. Legend:
1) Clayey-silty sequences; 2a) Clayey-marly sequences; 2b) Clayey sequences; 3a) Calcareous sequences; 3b) Conglomeratic
sequences; 3c) Tuffaceous sequences; 4a) Calcareous sequences (calcareous s.s. and calcareous-clayey); 4b) Arenaceous-
clayey and arenaceous-conglomeratic sequences; 5a) Sandy- arenaceous sequences; 5b) Alluvial sequences; 5c) Fluvial
sequences; 5d) Cemented detrital sequences; 5e) Incoherent detrital sequences.

of the carbonate slopes; generally terraced, Quaternary according to geological-technical and geomechani-
alluvial fan deposits along the river valleys. cal features (Figure 2). This procedure permitted the
The structural setting, resulting from the tectonic obtainment of data regarding the relationships between
phases, strongly influences the morphological con- landslides and involved deposits and consequently on
figurations on the area. The western portion of the the principle causes of landsliding.
province is characterized by the presence of calcareous The tectogenetic phases have induced complex tec-
mountains, while the eastern sector is characterized by tonic settings in the area, testified by typical structures
a hilly morphology (Figure 1). In contrast, the central of a ductile-type and fragile-type tectonics, which
area corresponds to the depression where marine and completely deform and displace the sequences. Sets
continental clastic Pleistocene deposits outcrop and of joints are connected to these, which pervasively
along which the main rivers flow. affect the masses and show high frequency near the
Bearing in mind the aims of this paper, it was most important tectonic lineations.
felt opportune to group the deposits outcropping in Therefore, most of the above-indicated sequences
the province into successions, characterized by a can be defined as structurally complex formations
likely lithotechnical homogeneity. In other words, (AA.VV. 1985, Picarelli 1986) as consequence of
these successions are constituted by similar lithotypes lithostratigraphical and tectonical features.

366
Table 1. Geomechanical and lithological characteristics of the sequences outcropping in the province of Benevento.

Group of sequences Lithotechnical sequences Competence Setting

1 Prevalently pelitic—from Clayey-silty (Ag-L) Incoherent Stratified, generally

low-degree to medium- monoclinalic
degree of tectonization

2 Prevalently pelitic—high- Clayey-Marly (Ag-M) Complex. From mildly to intensely

degree of tectonization Clayey (Ag) Prevalently folded. Scaly clay.
incoherent
3 Stony—from low-degree Calcareous (Ca) Stony Well stratified strata and banks
to medium-degree (from 30–40 cm to 5–10 m)
of tectonization Conglomeratic (Cg) Stony Sometimes stratified
Tuffaceous (Tf) Weakly lithified Up to 30 m banks. Columnar
jointing

4 Stony and complex—high- Calcareous Calcareous s.s. Stony Well stratified and highly
degree of tectonization (Ca) tectonized
Calcareous-clayey Complex, prevalently Well stratified strata and banks
stony Intensely jointed and folded

Arenaceous-calyey Complex Arenaceous and conglomeratic

Arenaceous-conglomeratic bank, thin layers of marl,
(Ar-AgCg) clay and conglomer ate
Intensely jointed and folded.

5 Coarse clastic and/or Sandy- arenaceous (S-Ar) Non-homogeneous Generally stratified and well
non-homogeneously lithification stratified
lithificated
Alluvial (Al) Incoherent or weakly Unclear bedding
Fluvial (Fl) cemented
Cemented detrital (Dc) Non-homogeneous Irregular bedding
lithification
Incoherent detrital (Ds) Incoherent or locally Irregular bedding
weakly cemented

On the basis of qualitative and quantitative obser- 3 LANDSLIDES OF THE BENEVENTO

vations, together with data collected from site and PROVINCE
laboratory tests (Guadagno et al. 2006), the masses
were classified from a geo-mechanical point of view. In order to investigate the distribution and type of
Table 1 shows the grouping of the various successions failures and their correlation with the lithostruc-
as a function of their common mechanical and litho- tural and morphological setting, a landslide inventory
logical characteristics. The main groups identified can map of the province was compiled on a 1:75,000
be listed as below: i) Group 1, Prevalently pelitic scale (a copy of the map can be requested by
sequences (from low-degree to medium-degree of tec- e-mail—guadagno@unisannio.it—from the authors).
tonization); ii) Group 2, Prevalently pelitic sequences The surveys were performed on the basis of a 1:25,000
(high-degree of tectonization); iii) Group 3, Stony scale topographic map, whereas, specific areas, char-
sequences (from low-degree to medium-degree of acterized by the presence of typical landslides and
tectonization); iv) Group 4, Stony and complex suc- towns or infrastructures, were surveyed on more
cessions (high-degree of tectonization); v) Group 5, detailed maps (greater than 1:10,000 scale).
Coarse clastic and/or non-homogeneously lithificated Analyses were carried out by means of an interpre-
sequences. tation of historical aerial photos from different time
The map in Figure 2 shows the areal distribu- periods, dating from 1954, together with geological
tion of the above described sequences. It is possi- surveys carried out between the years 2001 and 2005.
ble to observe that wide areas are characterised by It should be noted that, 2003 and 2005, two heavy
the presence of both clayey and stony tectonization rainfall events triggered the reactivation of hundreds
formations. These areas correspond to the sectors of landslides.
which are the most involved in erosion and landslide The classification criteria adopted refer to well-
processes. know classification systems (Varnes 1978, Hutch-

367
inson 1988, Cruden & Varnes 1996, Hungr 2001).
Furthermore, they are based on some specific char-
acteristics of landsliding processes which are worth
pointing out.

3.1 Classification criteria

Six different typological groups (numbered from 1
to 6 below) were distinguished, taking into account
the survey results: i) (1) falls and topples; ii) (2)
rock and debris avalanches, and debris flows; iii) (3)
translational, rotational and composite slides, and; iv)
earth flows distinguished in (4) single (Figure 3), (5) Figure 5. Example of basin affected by coalescent earth
multi-source (Figure 4) and (6) coalescent (Fig. 5). flows.
This latter typological choice derives from the
necessity to provide morpho-evolutive elements for
the earth flows, addressed, on the one hand, to
wide differences in the landsliding processes involving
understanding the landslide mechanisms and, on the
the slopes.
other, to defining hazard management. The use of
In relation to coalescent earth flows, the surveys
diversified typological contexts permits to stress the
highlighted that different sectors, mainly characterized
by the outcropping of pelitic and complex sequences
from a medium to high degree of tectonization, have
been and are still affected by several landslide gen-
erations that mobilise interacting among themselves.
These groups of landslides, which can be considered
as a unicum from a morphological point of view, are
localized inside unitary sub-catchments (Figure 5).
Multi-source earth flows reflect the complex litho-
structural settings. In particular, in specific areas,
litholologically differentiated sequences alternate or
they are in contact, inducing differentiated erosion
zones. The result is the formation of articulated source
areas, composed of several branches.
The structural setting may also influence the mor-
Figure 3. Oblique aerial view of the Sant’Agata de’ Goti phology of the channels, as well as that of the
landslide of 11 January ’97. accumulation zones.
These aspects are important in defining the distri-
bution of activity, which contributes to the assessment
of the evolution of the landslide body and, therefore,
to the obtainment of data on landslide susceptibility
evaluation.

3.2 Areal distribution of the earth flows

In the study area, 3,160 landslides were inventoried,
covering an approximate area of 358 km2 , equal to
about 18% of the whole provincial surface. This value
refers to parts of the territory already affected by land-
slides and mappable on the scale of representation. It
does not include those areas where predisposing con-
ditions have been verified and which indicate possible
Figure 4. Multi-source earth flow from a segment of the
Landslide Inventory Map (Guadagno et al. 2006). Legend: future landslide evolutions.
1) Distribution of the activity in the landslide area: a. Ret- Moreover, the Landslide index, calculated as the
rogressing; b. Enlarging; c. Advancing; 2) Source area; percentage of area affected by landslide events per grid
3) Movement direction; 4) Lithostructural element control- of 1 km2 , reaches values even greater than 75% where
ling the evolution of the landslide mass. clayey sequences are outcropping.

368
Figure 6. Percentage distribution of areal landslide phe-
nomena, computed for number and area, in dimensional
classes. Figure 7. Percentage ratio between area involved in land-
sliding (SF) and total outcropping area (ST) for each
lithotechnical sequence. Refer to Table 1 for the symbol-
Figure 6 shows the percentage distribution of the ogy used.
landslides in dimensional classes. The data refers to
the single events, even if they make up part of the
coalescent groups. It is interesting to note that land- Cruden & Varnes 1996) of the landslides and therefore
slides <10 ha constitute over 50% of the inventoried their evolution in time (Figures. 8, 9). In particular,
phenomena, whereas the few larger events (ha >25) the evolution of earth flows, in terms of activity and
represent over 20% of the areas affected by landslides. kinematic mechanisms, can be generally explained by:
i) bedding of the homogeneous and complex
sequences; ii) stratigraphic or tectonic contact between
3.3 Controlled evolution of earth flows
deposits with a different competence; iii) zone of
The litho-structural setting determines favourable con- intense fracturing linked mainly to the presence of axes
ditions for the development of landslides. To verify of folds and faults.
from a quantitative point of view, analyses of data were These settings constitute fundamental geological
carried out both for the 5 groups of sequences and for elements in understanding the control mechanisms in
the single lithotechnical sequences (cf. Table 1). Con- the earth flow source areas and channels. The haz-
sidering the landslide type, analyses were carried out ard analysis and, in particular, aspects connected to
evaluating the formations specifically involved in the the spatial prediction, imply a careful evaluation of
source areas. the geostructural conditions that become a fundamen-
As shown in Figure 7, the percentage of out- tal element in the areas where first-order events are
cropping surface affected by landslides is up to developing.
46% of the total. The prevalently pelitic and com- Where geological bodies with different competence
plex sequences at a high-degree of tectonization are outcrop, the lithostructural control is recurrent. A typ-
those mostly affected by landslides, totalling more ical example is the instability in Figure 9. The presence
than 80% of the inventoried phenomena. Addition- of a bedding plane (N125◦ /40◦ ) of a stony sequence
ally, over 40% of the areas where clayey forma- directs the shape, the orientation and the evolution
tions crop out, are also affected by landslides. In of both the source area and the channel of the land-
particular, lithotechnical sequences named clayey, slide, inducing a lateral retrogression of the source in
clayey-marly and arenaceous-clayey—arenaceous- comparison with the flow channel.
conglomeratic are those in which most of the land- Areas of severe structural control are the source
slides occur. areas of multi-source earth flows (more than 150),
Once the initial failure is identified, landslide devel- where separation into source branches is a direct con-
opment can be guided by the presence and orientation sequence of the local lithological and structural setting.
of the structural control elements, which influence Figures 4 and 10 show the morphological characters
the style and the distribution (cf. WP/WLI 1993, of multi-source instabilities and the distribution of the

369
 A
Figure 10. Example of multi-source earth flow. For the
legend see Figure 4.

Figure 8. Earth flow structurally controlled by a marly-

clayey hill. During motion, the landslide body affected some
houses: A) Image of November 2003; B) Image of March
2005.

Figure 11. Examples of basin affected by coalescent flows;

Legend: 1) Coalescent earth flows and basin boundary;
2) Flow direction.

of the surface is involved in mass movements on aver-

age, with extreme values of 72%, whereas the number
of events with interconnected evolution varies from
5 to 36.

4 CONCLUSION

The structural control of the earth flows of the Province

Figure 9. Earth flow from a segment of the Landslide of Benevento was investigated by means of a land-
Inventory Map (Guadagno et al. 2006). For the legend see slide inventory map. A morphological evolution con-
Figure 4. ditioned by active landslide processes affects a large
part of the territory. Considering the prevalent clayey
nature of the outcropping deposits, slow-velocity land-
activity by means of the empty triangles. The peculiar slides predominate, having in some areas a pervasive
shaping of the source areas are fixed by the presence of diffusion.
stony sequences overlapping highly-tectonized, pelitic Landslides are characterized by a reactivation ten-
terrains. dency. They are linked to the repetition of triggering
In contrast, conditions of relative lithological homo- conditions which are correlated to the occurrence of
geneity can be attributed to the basins affected by seismic events and rainfall with specific characteris-
coalescent earth flows. In particular, about 50 basins tics, in terms of intensity and duration.
(Figure 11) have been identified that include over 390 The diffusion of the earth flows led to their group-
events (size <10 ha). For each single basin, about 60% ing into three types. On this basis, the definition of

370
the distribution of the activity seems to be relevant in Guadagno, F.M., Focareta, M., Revellino, P., Bencardino, M.,
order to obtain basic data with the aim of ascertaining Grelle, G., Lupo, G. & Rivellini, G. 2006). La carta delle
landslide susceptibility. frane della provincia di Benevento. Sannio University
This distinction seems to be a basic aspect for Press.
the comprehension of landslide evolutive mechanisms Hungr, O., Evans, S.G., Bovis, M. & Hutchinson, J.N. 2001.
Review of the classification of landslides of the flow type.
and, consequently, for evaluations related to prediction Environmental and Engineering Geoscience, 7(3): 1–18.
of future events. It also opens a classification strategy Hutchison, J.N. 1988. General Report: Morphological and
for earth flows. Geotechnical Parameters of Landslide in Relation to Geol-
ogy and Hydrogeology.- Proc., Fifth International Sym-
posium on Landslide (C. Bonnard, ed), A.A. Balkema,
ACKNOWLEDGEMENTS Rotterdam, Netherlands, 1, 3–35.
Patacca, E. & Scandone, P. 1989. Post-Tortonian mountain
building in the Apennines. The role of the passive sinking
The research was financially supported by the PRIN of a relic lithospheric slab. In A. Boriani, M. Bonafede,
2005 project (prot. 2005047032_003, resp. F.M. G.B. Piccardo, G.B Vai (Eds.): The lithosphere in Italy.
Guadagno) and by the Provincial Council of Ben- Advances in Earth Science Research. It. Nat. Comm. Int.
evento and Benevento Chamber of Commerce. We Lith. Progr., Mid-term Conf. (Rome, 5–6 May 1987), Atti
would like to thank Italo Abate and Donato Tornesiello Conv. Lincei, 80: 157–176.
for the photographs in Figures 3 and 8a. Pescatore, T.S., Di Nocera, S., Matano, F. & Pinto, F. 2000.
L’Unità del Fortore nel quadro della geologia del settore
orientale dei Monti del Sannio (Appennino Meridionale).
Boll., Soc., Geol., It. 119, 587–601.
REFERENCES Picarelli, L. 1986. Caratterizzazione geotecnica dei terreni
strutturalmente complessi nei problemi di stabilità dei
AA.VV. 1985. Geotechnical Engineering in Italy. An pendii. Proc. 16th Conv. Ital. di Geotecnica, Bologna,
overview. Published on the occasion of the ISSMFE 3, 155–170.
Golden Jubilee. Roma, A.G.I.-Ass.Geotecnica Italiana: Varnes, D.J. 1978. Slope movements, type and processes. In:
414 pp. Schuster R.L. & Krizek R.J. (Eds.), Landslides analysis
Cruden, D.M. & Varnes, D.J. 1996. Landslide Types and Pro- and control. Washington Transportation Research Board,
cesses. In ‘‘Landslides: Investigation and Mitigation’’, Ed. Special Report 176. National Academy of Sciences, WA,
Turner A.R. e Shuster R.L.Sp.Rep. 247, Transportation 11–33.
Research Board, National Research Council, National WP/WLI 1993. A Suggested Method for Describing the
Academy Press, Washington D.C.: 36–72. Activity of a Landslide. Bulletin of the I.A.E.G, 47, 53–57.

371
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Geotechnical and mineralogical characterization of fine grained soils

affected by soil slips

G. Gullà & L. Aceto

CNR-IRPI_Sede di Cosenza, Rende (CS), Italy

S. Critelli
Università degli Studi della Calabria, Dip. Scienze della Terra, Rende (CS), Italy

F. Perri
Università degli Studi della Basilicata, Dip. Scienze Geologiche, Potenza (PZ), Italy

ABSTRACT: Soil slips affect essentially the degraded or weathered covers of soil, and could be particularly
dangerous because of their kinematic mechanism and their spreading over wide areas during rainstorms. The high
incidence of soil slips in an area of Central Calabria (Southern Italy) prompted a research aimed at geotechnical
characterisation of fine-grained soils involved in this kind of instability. This paper illustrates the geotechnical
and the mineralogical characterization carried out on samples coming from three sites representative of the wider
study area, where sedimentary terrain (Plio-Pleistocene) crops out. Classification and direct shear tests have been
conducted on undisturbed specimens and artificially degraded specimens. The results supply a reference frame
of the soils physical-mechanical characteristics. Correlating the geotechnical analysis with the mineralogical-
geochemical investigations allows a wider characterization of the sediment properties.

1 INTRODUCTION landslides in Calabria (Southern Italy). Above all we

compare in this paper the geotechnical and mineralo-
Shallow and fast sliding-flow instabilities, soil slips gical characters of natural (intact) silt and clay soils
in this paper (Campbell 1975), are dangerous phe- with the same soils artificially degraded in the labora-
nomena produced by frequently shallow landsliding tory (Gullà et al. 2004, Gullà et al. 2006).
events related to rainstorms (Govi & Sorzana 1980,
Brand 1984, Antronico & Gullà 2000, Wieczorek
et al. 2001, Antronico et al. 2002, Antronico et al. 2 GEOLOGICAL SETTING
2004, Sorriso-Valvo et al. 2004).
The presence of many structures and infrastruc- The Calabria-Peloritani Terrane (CPT) is a fault-
ture creates very high risk conditions for wide areas bounded, allochthonous terrane located between
affected by shallow landsliding events. the NW-SE-trending Southern Apennines and E-W-
Furthermore all lithotypes involving thick weath- trending Sicilian Maghrebides (Bonardi et al. 2001).
ered or degraded soil profiles can be affected by soil The CPT is characterized by a pre-Mesozoic crys-
slips. Shallow landslide density, and then the magni- talline basement and it shows evidence of pre-Neogene
tude of the shallow landsliding events, depends on soil tectonism.
cover saturations conditions before the triggering rain- The CPT comprises several Oligocene-to-Recent
storms (Antronico et al. 2004), and to the geomaterial sedimentary units covering about half of its emerged
characteristics (Antronico et al. 2002). bedrock area. These sedimentary deposits are mostly
In particular, the density of soil slips in fine-grained equally distributed along the Ionian and Tyrrhenian
soils is probably related to the saturation-desaturation margins of Calabria and in mostly subsided areas in
cycles that affect their degradation and reduce their the central portions of the Calabrian terranes, one of
shear strength (Gullà et al. 2004, Gullà et al. 2006). these latter subsided areas is the Catanzaro Graben ori-
To better understand this predisposing-triggering ented orthogonal to the overall structural grain of the
effect, we discuss in this paper the geotechnical and CPT. The Catanzaro graben is filled with deposits of
mineralogical characteristics of some silt and clay Tortonian to Quaternary age, composed of marginal
samples collected in a study area affected by shallow fanglomerates, evaporites and shallow marine facies,

373
 equipped with a solid-state Si (Li) detector. The min-
eralogical composition of the <2 μm grain-size was
determined from a thin section of highly oriented
aggregate. The presence of expandable clays was
determined after treatment with ethylene glycol at
25◦ C for 15h. Elemental analyses for major and some
trace elements concentrations were obtained by X-ray
fluorescence spectrometry (Philips PW 1480). Total
loss on ignition (L.O.I.) was determined, after heating
the samples for three hours, at 900◦ C.
The morphological change and secondary min-
eral formation were examined by field emission
scanning electron microscopy (ESEM Philips Elec-
tronics QUANTA 200F with EDX GENESIS 4000).
Secondary mineral compositions were determined
qualitatively by energy dispersive X-ray spectrometer
(EDS).

Figure 1. Geological outline of the study area and location

of the sample sites. 1) alluvial deposits; 2) continental con-
glomerates and sand; 3) marine conglomerates and sands; 4 GRAIN SIZE DISTRIBUTIONS AND INDEX
4) mostly sandstone deposits; 5) mostly clay deposits; PROPERTIES
6) mostly conglomerate deposits; 7) evaporitic level;
8) Castagna Unit (two mica gneiss); 9) Polia-Copanello
Unit (garnet and sillimanite-bearing gneiss); 10) sedimen- Based on the grain size distribution envelope of 54
tary deposits which overlay Polia-Copanello Unit; 11) Stilo clay samples collected in the studied area, it is possi-
Unit (phyllites); 12) normal fault; 13) transcurrent fault; ble to classify the soils of this area as clay with silt to
14) landslide; RO = Roccella site. From Gullà et al. 2004, silt with clay and sand (Fig. 2a). The grain size dis-
modified. tribution curves of the clay samples investigated by
mineralogical analyses (Monte Frasso site) are shown
in Figure 2a.
characterized by fine grained sediments (Fig. 1). Fine- The grain size distribution envelope of 99 silt soil
grained sediments are widespread in the Catanzaro samples (Vallone Solleria and Murano sites) show a
Graben and are mostly Pliocene to Pleistocene in age. composition that varied from sandy silt with clay to
slightly gravelly clayey sand with silt (Fig. 2b). Figure
2b shows the grain size distribution curves of the silt
3 EXPERIMENTAL PROGRAM AND samples used for mineralogical analyses.
MEASURING TECHNIQUES Based on some index properties, it is possible to
observe that there are differences between clay and
Laboratory tests were carried out to detect the charac- silt soils, even if the ranges of overlap values are
teristics of the fine-grained soils present in the study comparable. The average values of all clay samples
area, and to investigate the variations in mechani- are: unit weight of solid particles 26.7 kN/m3 , unit
cal characteristics due to the wetting-drying-freezing- weight of dry soil 15.1 kN/m3 , void ratio 0.778; for
thawing cycles (Antronico & Gullà 2000, Gullà & the clay samples used for mineralogical analyses the
Antronico 2001, Gullà et al. 2004, Gullà et al. 2006). average values are: unit weight of solid particles 26.6
The experimental program, that has been carried kN/m3 , unit weight of dry soil 14.9 kN/m3 , void
out on samples collected in three different sites of ratio 0.786. The average values of all silt samples
the study area (Monte Frasso, Vallone Solleria and are: unit weight of solid particles 26.6 kN/m3 , unit
Murano; Fig. 1), is based on classification tests (parti- weight of dry soil 16.0 kN/m3 , void ratio 0.674; the
cle size distribution analysis and index test), direct and average values of the silt samples used for mineralog-
ring shear tests on intact, degraded and reconstituted ical analyses are: unit weight of solid particles 26.7
specimens. Twenty-nine fine grained soil samples kN/m3 , unit weight of dry soil 16.3 kN/m3 , void ratio
were analyzed by X-ray diffraction (XRD) and X-ray 0.639.
fluorescence spectrometry (XRF). Plasticity and activity characteristics of the exam-
For the chemical and mineralogical analyses, the ined soils are illustrated in Figure 3. It is possible to
whole-rock samples were first dried and then crushed observe that the clay soils are classifiable as inorganic
by hand in an agate mortar. Randomly-oriented whole- inactive, with medium-high plasticity, whereas the silt
rock powders were run using a Scintag X1 apparatus soils are classifiable as inorganic inactive or normally

374
active, with medium-low plasticity. Clay and silt sam- 5 MINERALOGY AND CHEMISTRY
ples used for mineralogical analyses are representative
of the total analysed samples. Plasticity and activ- The XRD patterns of whole-rocks show that the
ity characteristics of the clay and silt soils are quite analyzed samples are rich in clay minerals asso-
similar. ciated with significant amounts of calcite, quartz
and feldspars, whereas minor concentrations of
dolomite have been identified in some XRD patterns
(Figs. 4–5). The <2μm grain-size fraction of all the
samples is composed predominantly of illite, followed
by illite-smectite mixed layers, chlorite and kaolinite.
The differences among silts and clays are related to
the clay mineral contents (Fig. 4). The clay soils show a
higher percentage of clay minerals and lower content
of feldspars, dolomite and quartz. These differences
are minor, and do not significantly affect the plasticity
and activity features (Fig. 3).
The XRD patterns of natural (intact) silt samples
and silt samples degraded in the laboratory do not show
substantial mineralogical differences (Fig. 5); only the
<2 μm fraction of these samples shows a little variation
related to the illite and illite-smectite mixed layers (I-S)
content (Fig. 6). These slight differences are proba-
bly related to the laboratory degradation treatments
(cycles of saturation, drying, wetting and freezing).

Figure 2. Grain size distribution envelopes for clays (a) and

silts (b) in the study area, and grain size distribution curves
for the samples used for present study (MF = Monte Frasso,
VS = Vallone Solleria; MU = Murano).

Figure 4. The XRD patterns of silt and clay whole-rocks

fine grained soil samples.

Figure 3. Plasticity and activity charts for the clays (a)

and silts (b) in the study area (solid symbols), and for the
samples used for mineralogical analyses (empty symbols)
(MF = Monte Frasso, VS = Vallone Solleria; MU = Figure 5. The XRD patterns of silt samples degraded in
Murano). laboratory and natural (intact) silt samples whole-rocks.

375
Figure 6. The XRD patterns of the air dried (AD) and
ethylene-glycol solvated (GLY) <2 μm fraction specimens Figure 7. Major element distributions normalized to PAAS
of silt samples degraded in laboratory and natural (intact) silt (Post-Archean Australian Shales; Taylor and McLennan,
samples. 1985) of silt and clay samples.

The <2 μm XRD pattern of natural (intact) silt sam-

ples and silt samples degraded in the laboratory shows
variations in the expandable clays, in which a different
peak shape is observed in the 5–10 ◦ 2θ range, related
to different content of the I/S mixed layers. This is
probably associated with chemical and mineralogical
variations in response to different cycles of degra-
dation. The silt samples show a content of smectite
layers in the I-S mixed layers in a range of 40–50%
(R ordering = 0; Reickeweite number), suggesting
a shallow diagenetic stage. The same characteristic
seems detectable too in the clay samples.
Major and trace element variations and chemical Figure 8. Major element distributions normalized to PAAS
weathering indices are related to a different behaviour (Taylor and McLennan, 1985) of silt samples degraded in
and distributions of chemical elements, principally laboratory and natural (intact) silt samples.
controlled by the degree of artificial degradation and
weathering. The elemental distributions of the studied
samples have been normalized to the PAAS standard Al2 O3 , Fe2 O3 , TiO2 , MnO, MgO and K2 O have a
(Post-Archean Average Shale; Taylor and McLennan positive correlation whereas CaO has a negative cor-
1985). relation. In fine-grained sediments Al2 O3 commonly
Major element composition of the samples is quite monitors (indicates) clays and this behaviour may thus
homogeneous, whereas the trace element concentra- account for the competition between mica-like clay
tion is relatively variable. minerals and carbonate phases. Most of the trace
Chemical analyses show similar variations between elements show positive linear trends against Al2 O3 ,
silts and clays (Fig. 7) and between natural (intact) silt Fe2 O3 , K2 O and TiO2 , indicating clay mineral control
samples and silt samples degraded in the laboratory on their contents.
(Fig. 8).
The soil samples are characterized by narrow com-
positional changes for all elements which have concen-
trations rather weakly depleted relative to the PAAS.
Mg alone has concentrations close to those of the 6 MINERALOGICAL SIGNATURES OF
PAAS, whereas Ca is strongly enriched. Sr, similar DEGRADATION AND WEATHERING
to Ca, is enriched relative to the PAAS. In standard-
normalized variation diagrams, for the natural (intact) The scanning electron microscope (SEM) and appli-
silt samples and silt samples degraded in the labo- cation of X-ray diffraction (XRD) have been used
ratory, only the Na ratio weakly decrease during the to better understand the behaviour of grain min-
course of artificial degradation. erals and their morphological variation during the

376
course of sample degradation. Among common rock- 7 SHEAR STRENGTH
forming minerals, quartz is the most resistant mineral,
while plagioclase and mica (usually biotite) are easily The direct shear tests (peak and residual) carried out
replaced by secondary phases. on specimens used for the present study gave results
In Figure 9 SEM photomicrographs show the grain that are congruent with the shear strength envelopes
feature, coating products and interlayer sites in the proposed by Gullà et al. (2004).
mica minerals. In order to investigate the mechanisms that pro-
The difference between Figure 9a and 9b are related duced a reduction of the shear strength, we have been
to the grain-size particles; Figure 9a shows the silt fea- reproduced in the laboratory a lot of wetting-drying-
tures characterized by more feldspar and quartz grains, freezing-thawing cycles with a period that range from
whereas the clay sample (Fig. 9b) is clearly marked by one to 90 days (Gullà et al. 2004, Gullà et al. 2006);
abundant clay minerals. using this technique, we have been studied the degra-
The slightly weathered mica (Fig. 9c) of the natu- dation of the soils induced by saturation-desaturation
ral (intact) silt sample shows a smooth basal surface cycles.
with etch pits of submicrometer size, whereas sec- The results of the direct shear tests carried out on
ondary minerals are attached to the edge. During the natural (intact) and artificially degraded specimens,
course of weathering and the degradation processes, at the same vertical stress, are shown in Figure 10.
mica minerals have a tendency to open the reticular In both cases, for clay and silt, we can see the same
sheets and gradually lose K+ ions toward the top part. failure mechanism, and find significant reductions of
This result indicates that illitization has occurred and shear strength for the degraded specimens relatively
such K-depleted inter-layers in the mica minerals will to the shear strength of the natural (intact) specimens:
be referred to hereafter as illite/smectite-like inter- for clay 74% and 35% respectively after one day and
layers. The mica mineral of the silt sample degraded two days of degradation; for silt 42% after one day,
in the laboratory (Fig. 9d) is slightly deformed, with 32% after 14 days and 28% after 30 days (after 60 and
secondary minerals present in the lower part and at
the edge of the grain. One of the secondary min-
erals is identified qualitatively as an illite and/or
smectite-like silicate, as indicated by its morphology
and composition (revealed by SEM-EDS).
These deformation mechanisms is the mica min-
erals are probably partially related to the artificial
degradation carried out in the laboratory.

Figure 9. SEM images from grains of four different fine

grained soil samples: a) silt sample, b) clay sample, c) natural Figure 10. Stress-displacement curves of intact and artifi-
(intact) silt sample, d) silt sample degraded in laboratory. cially degraded specimens for clay (a) and silt (b) soils.

377
90 days of degradation there is no further reduction in maximum peak shear strength after thirty days of
shear strength). degradation. The same feature is found for the reduc-
For the silts, the previously illustrated results allow tion curve of the clay studied by Gullà et al. (2006) at
us to assume that the decrease of shear strength for the Roccella site (Fig. 11). It is possible to assume
degradation is not due to mineralogical and/or chem- similar features for the other envelope curves for
ical changes, according to the results obtained from the clay and silt studied in this paper (reduction of
Gullà et al. (2006) for the soils outcropping in an the maximum peak shear strength after thirty days
other area of the Calabria region (Roccella site, RO of degradation: between 20% and 45% for clay and
in Fig. 1). The reduction of shear strength for the silts between 27% and 35% for silt).
tested in this study, according to the studies by Gullà
et al. (2006), is probably related to the soil structure
changes induced from the degradation cycles. How- 8 CONCLUSIONS
ever, other enquiries are necessary in the study case in
order to confirm this hypothesis. This paper has illustrated an integrated study, based on
Figure 11 gives a first reference frame for the shear mineralogical and geotechnical investigations, aimed
strength data related to artificially degraded speci- to improve the characterisation of fine grained soils
mens. Figure 11 shows the ratio values among shear involved in soil slip phenomena. This approach is
strength of the degraded specimens and maximum crucial to better understand the magnitude and dis-
peak shear strength of the undisturbed specimens, tribution of shallow landsliding events in such soils.
related to the time of degradation and for all examined The degradation cycles simulated in the laboratory
vertical stresses. do not produce mineralogical and chemical changes
The examined vertical stresses are: from 13 kPa to in tested clay. However cycles of wetting-drying-
98 kPa for clays, with cycles of degradation until seven freezing-thawing have disturbed the natural structure
days; from 24.5 kPa to 147 kPa for silts, with cycles of the soil, producing changes in fabric and bonding.
of degradation until 30 days. After seven days of degra- The stress-displacement curves obtained from
dation it is possible to observe shear strength decreases direct shear tests on intact silts and clays show a well-
between 55% and 20% for clays, and between 50% and defined peak strength in contrast to similar tests on
30% for silts (Fig. 11). the degraded specimens. Peak shear strength shows a
Figure 11 shows an asymptotic trend of the curves reduction during one month of degradation; after this
for silts, that indicate a reduction to 27% of the period no further meaningful change is observed. The
reduction of the peak shear strength due to the degra-
dation induced in the laboratory is comparable to the
residual conditions (Gullà et al. 2004).
In conclusion, the approach used and the results
proposed in this paper represents valuable support for
the study of shallow landslide susceptibility in fine
grained soils.

REFERENCES

Antronico, L. & Gullà, G. 2000. Slopes affected by soil slip:

validation of an evolutive model. Proc. of the 8th Intern.
Symp. on Landslides, Cardiff 1: 77–84.
Antronico, L., Gullà, G. & Terranova, O. 2002. L’evento plu-
viometrico dell’8–10 settembre 2000 nella Calabria Ionica
Meridionale: Dissesti sui versanti e processi in alveo.
Proc. of the Symp. ‘‘Il dissesto idrogeologico: inventario
e prospettive’’, XIX Giornata dell’Ambiente, Accademia
Nazionale dei Lincei, Roma, 5 Giugno 2001: 67–79.
Antronico, L., Gullà G. & Borrelli, L. 2004. Shallow instabil-
ities for sliding flow: regional influence and area effects.
Proc. of the 9th Intern. Symp. on Landslides, Rio de
Janeiro 2: 1381–1387.
Bonardi, G. Cavazza, W. Perrone & V. Rossi, S. 2001.
Calabria-Peloritani terrane and Northern Ionian Sea. In:
Figure 11. Effects of degradation on the maximum peak Vai G.B., Martini I.P., Eds. Anatomy of an orogen: the
shear strength. MF = Monte Frasso; MU = Murano; VS = Apennines and Adjacent Mediterranean Basins, 287–306.
Vallone Solleria; RO = Roccella. Dordrecht/Boston/London, Kluwer Academic Publishers.

378
Brand, E.W. 1984. Landslides in Southeast Asia: A State- Gullà, G. Mandaglio, M.C. & Moraci, N. 2006. Effect of
of-the-Art Report. Proc. of the 4th Intern. Symp. on weathering on the compressibility and shear strength of
Landslides, Toronto: 17–59. natural clay. Can. Geotech. J. 43: 618–625.
Campbell, R.H. 1975. Soil slip, debris flow and rainstorms in Sorriso-Valvo, M., Antronico, L., Gaudio, R., Gullà, G.,
the Santa Monica Mountains and vicinity, Southern Cal- Iovine, G., Merenda, L., Minervino, I., Nicoletti, P.G.,
ifornia. U.S. Geological Survey Professional Paper 851: Petrucci, O. & Terranova, O. 2004. Carta dei dissesti
1–51. causati in Calabria meridionale dall’evento meteorologico
Govi, M. & Sorzana, P.L. 1980. Landslide susceptibility as dell’8–10 settembre 2000. CNR-GNDCI, Pubblication
a function of critical rainfall amount in Piemont Basin 2859, Geodata 45, Rubbettino Publisher, Soveria Man-
(North-Western Italy). Studia Geomorph. Carpatho- nelli, Italy.
balcanica 14: 43–61. Taylor, S.R. & McLennan, S.M. 1985. The Continental
Gullà, G. & Antronico, L. (eds.) 2001. Le instabilità superfi- Crust: Its Composition and Evolution. Oxford, Blackwell
ciali per scorrimento-colata nella Stretta di Catanzaro. In Scientific.
‘‘Linee Guida per Interventi di Stabilizzazione di Pendii Wieczorek, G.F., Larsen, M.C., Eaton, L.S., Morgan, B.A. &
in Aree Urbane da Riqualificare’’, CNR-IRPI, Regione Blair, J.L. 2001. Debris-flow and flooding hazards associ-
Calabria-UE, POP 1994/99. ated with the December 1999 storm in coastal Venezuela
Gullà, G. Aceto, L. & Niceforo, D. 2004. Geotechnical char- and strategies for mitigation. U.S. Geological Survey,
acterization of fine-grained soils affected by soil slips. Open File Report 01–144, pp 40.
Proc. of the 9th Intern. Symp. on Landslides, Rio de
Janeiro 1: 663–668.

379
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Vulnerability of structures impacted by debris flow

E.D. Haugen & A.M. Kaynia

Norwegian Geotechnical Institute, Oslo, Norway

ABSTRACT: As debris flow impacts a structure it sets the structure in vibratory motion. This article presents
a simple method for prediction of damage in a structure impacted by a debris flow of known magnitude. The
method uses the principles of dynamic response of simple structures to earthquake excitation and fragility curves
proposed in HAZUS for estimation of the structural vulnerability, by the damage state probability. The method
was tested by applying it to a debris flow site in Italy where the reported structural damages ranged from light to
complete devastation. The vulnerability of six of the impacted structures was assessed by the proposed model
and compared with the real damage states.

1 INTRODUCTION makes the comparison of approaches difficult. In addi-

tion, many vulnerability or risk assessment models
In areas susceptible to landslides, it is becoming more depend on empirical data from past damages (e.g.
and more common to assess the risk in two parts: Wong et al. 1997, Remondo et al. 2005) and on local
1) Assessing the hazard, which is defined by: ‘‘The historical databases which contain the socio-economic
probability that a particular danger (threat) occurs damage values.
within a given period of time.’’ 2) Assessing the vul- This paper presents a suggestion to a quantitative
nerability, which is defined by: ‘‘The degree of loss model for vulnerability assessment of structures hit
to a given element or set of elements within the area by debris flow, where the vulnerability is measured by
affected by a hazard’’ (ISSMGE, 2004). the probability of reaching a certain damage state. The
The area of risk management in geohazards is fairly method is based on the premises that the debris flow is
complex. Much work has been done on making ade- not so large as to completely wipe out the structure, but
quate models for the probability of occurrence of an impacted by a force, which leads to structural vibration
adverse event and developing landslide hazard maps and results in a specific state of damage.
(e.g. Remondo et al. 2005). But within the area of
vulnerability many conflicts exist; vulnerability anal-
ysis has commonly been integrated directly in the risk
2 THEORY
analysis and no unique or simple method is found for
landslide vulnerability assessment (Glade, 2003).
2.1 HAZUS damage state probabilities
Many suggestions have been made to vulnerability
assessments of structures struck by landslides. Car- The vibrations from an impact will damage a structure
dinali et al. (2002) assessed the vulnerability as a in approximately the same way as a ground vibration
function of qualitative landslide intensity and type. will damage a structure standing on the ground. From
Leone et al. (1996) made use of the damage intensity as this it may be assumed that a debris flow impact force,
a vulnerability measure. Heinimann (1999) estimated modelled as an impulse, may be compared to ground
rock fall vulnerability by using the building type and acceleration from an earthquake. This allows one to try
rock fall magnitude as input. to incorporate existing literature on structural damage
As one may see, none of these methods cover the from earthquake vibrations in landslide vulnerability
whole scope of the vulnerability problem. Each of the assessment.
methods has used important parameters, such as land- The Federal Emergency Management Agency
slide intensity, volume and type, but not one of them (FEMA) of the United States together with the
incorporates completely all the aspects of vulnerabil- National Institute of Building Sciences (NIBS), have
ity. Glade (2003) states that most approaches do not developed the HAZUS programme (short for Hazard
distinguish between types of processes or magnitudes, US). HAZUS is used in USA to map the expected
that absolute values of vulnerability vary significantly losses due to earthquake events. HAZUS quantita-
and these values are spread over a wide range; this tively estimates the losses in terms of direct cost

381
for repair, direct cost associated with loss of func- flow using HAZUS principles. This means that a struc-
tion, casualties, displacement from residences, debris ture’s maximum lateral displacement caused by debris
quantity and regional economic impact (HAZUS, flow impact may be taken as the spectral displacement
2006). and used in the fragility curves. The result would be
In contrast to the conventional vulnerability grading a quantitative estimation of vulnerability—that is; the
from 0–1 for a given element at risk, HAZUS defines probability of certain damage states occurring for a
five damage states: None, slight, moderate, extensive given debris flow impact.
and complete.
These damage states are displayed by fragility
2.2 Spectral displacement
curves, which portray the damage-motion relation-
ship. The fragility curve for a given building type A debris flow hits a structure with an impulse force, the
describes the probability of a specific damage state dynamic impact force, Pdy . This will set the structure
arising as a function of the spectral displacement. The in vibrations and give it a lateral displacement, which
spectral displacement is the maximum displacement is increased by the hydrostatic force, Pst . Assuming
under earthquake of a single-degree-of-freedom sys- a single-degree of freedom, the maximum displace-
tem, with the same natural frequency as the building ment of a structure, which may be set as the spectral
relative to the ground motion. HAZUS has classified displacement, is:
36 building classes, and the vulnerability for each
of these may be assessed with the fragility curve of umax = ust + udy,max (2)
each class. An example of fragility curves for unre-
inforced old masonry buildings (class 34) is seen in The displacement from the hydrostatic force, ust ,
Figure 1, which shows how the damage state proba- may be found directly from the static force, Pst , and
bility may be found from the spectral displacement of the structure stiffness, k:
the building.
Each fragility curve is a log-normal curve, char- Pst pst A ρdf gbh2
ust = = = (3)
acterised by a median spectral displacement (Ŝd,ds ), k k 2k
which corresponds to the threshold of the damage
state, and a log-normal standard deviation (βds ), the where pst is the average hydrostatic pressure, A = bh
variability associated with the damage state. is the area of the structure hit by the debris flow, ρdf
The probability of a certain damage state, ds, given is the debris flow density, g the gravitational constant
a spectral displacement, umax , is defined by: (9.81 m/s2 ), b is the width of the flow and h is the
height of the flow.
   The maximum dynamic displacement, udy,max ,
1 umax must be found from the impulse force, Pdy , and
P(ds|umax ) =  ln (1)
βds Ŝd,ds structure parameters such as the natural period of the
structure, Tn , and the stiffness, k. Using a shock spec-
trum, see Figure 2, the displacement may be found
where  is the standard normal cumulative distribu- via Rmax . Rmax is the response ratio which is the
tion function. ratio of the dynamic displacement to the static
The objective of this study is to provide a framework
for assessment of structural vulnerability to debris

Figure 1. HAZUS fragility curves for building class 34, Figure 2. Shock spectrum for three force pulses of equal
unreinforced masonry buildings. amplitude (Chopra, 2001).

382
displacement, caused by the impulse force amplitude The natural period of the structure, Tn , can be esti-
Pdy,max (Clough, 1993): mated by the structure’s height, Hstr (NS 3491-12,
2004) as follows:
  For concrete frames:
udy,max Pdy,max udy,max
Rmax =   = (4)
ust Pdy,max Pdy,max /k 3/4
Tn = 0.075 · Hstr (8a)

In the shock spectrum the shape of the impulse For wooden frames:
force, Pdy (t), is assumed with a amplitude Pdy,max . 3/4
Three impulse shapes, rectangular, half-cycle sine and Tn = 0.050 · Hstr (8b)
triangular, are plotted for τ /Tn versus Rmax , where τ is
the impulse duration. For debris flow impact a conser-
vative assumption is a rectangular impulse force shape, 3 APPLICATION AND IMPLEMENTATION
so that Pdy,max = Pdy . OF METHOD
As seen from Equation (4) the maximum lateral
displacement imposed by a dynamic impulse impact On 5–6 May 1998 several devastating debris flows
may be found from the response ratio for a known occurred in the Sarno area of the mountain Pizzo
amplitude of the dynamic force and stiffness of the d’Alvono, in southern Campania, Italy. The area was
building. hit by tens of debris flows. 160 people were killed, over
The dynamic force must be estimated from para- 150 buildings were totally destroyed and over 500 dam-
meters which may be approximated. A function of the aged seriously or partially. The value of the damage to
debris flow velocity at impact, vdf , the flow density, structures was estimated to approximately 34 million
ρdf , and the area hit by the flow, A, is amongst oth- Euros (European Commision, 2003). Zanchetta et al.
ers proposed by Hungr et al. (1984), Armanini and (2004) mapped the extent of the debris flows in the
Scotton (1992) and Lin et al. (1992). The dynamic Sarno area.
force suggested by Hungr et al. is used: The debris flows originated from shallow soil slips
on slopes with an angle of 30–35 degrees. Nine days
of continuous rainfall had saturated the volcaniclastic
Pdy = pdy · A = ρdf · vdf
2
·A (5) soil, which failed on the boundary between soil and
bedrock. Progressive failure and liquefaction enlarged
Most debris flow pulses will last longer than Tn /2, the mass as it flowed downhill. The debris flows
which gives τ /Tn > 1/2. For these pulses the max- followed the stream channels until they reached the
imum deformation occurs during the impulse, and alluvial fans, with an inclination less than 12 degrees,
the pulse shape plays a bigger role than for pulses and spread out into the inhabited areas and caused
shorter than Tn /2 (Chopra, 2001). In these cases the damage and casualties.
shock spectrum may be used directly, inserting τ /Tn , Zanchetta et al. have estimated the physical and
and finding the response ratio from the corresponding dynamical properties of the debris flow which are
pulse shape. For structures with less than 20 stories needed to find the impact pressure and mapped the
and impulse duration equal to that of a flood impact, structural damage. This data is very convenient for
which FEMA (1995) proposes to be τ = 1 s, rectangu- using in the proposed vulnerability model, as all the
lar impulse forces give a response ratio with a constant input factors are available, and the damage is mapped.
value of Rmax = 2. Figure 8 in Zanchetta et al. gives the variation of
The stiffness of the building is also needed to find the dynamic and hydrostatic pressure and the dam-
the lateral displacement. It may be approximated by: age done to the buildings by three debris flows named
Ep-6, Lav-2 and Ep-3b. The damage varies from light
to complete.
4π 2 mstr
k= (6) Using the data found from the Sarno debris flows in
Tn2 the vulnerability model, the vulnerability of two build-
ings in each debris flow are assessed; these buildings
where mstr is the mass of the structure. are named Ep-6-1, Ep-6-2, Lav-2-1, Lav-2-2, Ep-3b-1
Using Equations (5) and (6) in Equation (4) and and Ep-3b-2.
rearranging it, gives the maximum lateral displace-
ment caused by the dynamic impact: 3.1 Impact force
For the three debris flows the impact pressures of
Pdy,max ρdf v2 A
udy,max = Rmax = Tn (7) the structures 1 and 2 in every debris flow are found
k 2π 2 mstr directly from Zanchetta et al. (2004). Table 1 shows

383
Table 1. Debris flow impact and damage at different spectrum as a function of impact force shape and τ /Tn.
structures. By assuming a rectangular impulse force shape and
using τ /Tn = 4 in Figure 2, gives the maximum dis-
Dynamic Static placement ratio Rmax = 2. For the less conservative tri-
force force Structural
MN MN damage
angular and sinusoidal impulse shapes, the maximum
displacement ratio is 1.0 and 1.15 respectively.
Ep-6-1 3.32 1.03 Complete Continuing with Rmax = 2, the displacement of
Ep-6-2 1.01 0.45 Heavy each structure due to the dynamic impact may be found
Lav-2-1 6.12 1.99 Complete by Eq. (7), and the displacement from the static force
Lav-2-2 1.08 0.81 Moderate-light from Eq. (3). The displacements for the various struc-
Ep-3b-1 3.93 1.14 Complete tures are calculated and listed in Table 3, where the
Ep-3b-2 0.43 0.34 Complete total displacement is taken as the spectral displacement
in the following steps.

3.4 HAZUS fragility curves

the slide properties, the dynamic force and the static
force and the damage of each structure. The displacements found in Table 3 are used as input in
the HAZUS fragility curves. But first the appropriate
fragility curves must be selected by finding the FEMA
3.2 Natural period and stiffness of structure building class of the buildings in the area, and the
The natural period of the structure depends on the type belonging curve parameters. The two types of build-
and mass of the structure. The buildings in the area ings in the area may be classified in the FEMA system
hit by debris flows were both old country houses and as URML (class 34) and RM2L, which respectively
new, reinforced concrete houses, mostly one- and two- refer to unreinforced old masonry buildings and new
story masonry residences with basement. Since it has reinforced masonry buildings, both with 1–2 stories
not been possible to find out the structure type of the and basement (HAZUS, 2006b).
six structures selected, the vulnerability assessment is As the buildings have not been seismically
performed for both types of structures. Assuming the designed, the fragility curve parameters may be found
structures in question all have two stories, which gives from Table 5.9d of the HAZUS Earthquake Tech-
them a height Hstr = 5 m, the natural period is found nical Manual (HAZUS, 2006b). These are listed in
from Eq. (8) to be Tn = 0.25 s. Table 4.
Making assumptions on the structural dimensions Having found the parameters for the standard cumu-
and using a concrete density, 2300 kg/m3 , gives the lative normal distribution of the fragility curves, and
approximated mass, mstr , in Table 2. The stiffness, the maximum displacement of the structures, the prob-
found by Eq. (6) with the structure mass and the natural ability of the different damage states may be found by
period, is also found in Table 2. Eq. (1).
The tables in the following section give the damage
state probabilities for the spectral displacements found
3.3 Spectral displacement in Table 3 for both un-reinforced masonry houses
(URML) and reinforced masonry houses (RM2L).
Using the impulse duration τ = 1 s, one sees that
τ > Tn /2 = 0.13 s, and therefore the maximum dis-
placement ratio may be found directly from the shock

Table 3. Displacements of the structures 1 and 2 in debris

Table 2. Approximated mass and stiffness of structures in
flows Ep-6, Lav-2 and Ep-3b.
flow paths.
Dynamic Static Total
Flat area Mass Stiffness
displacement displacement displacement
m2 kg MN/m m m m
Ep-6-1 105 97 · 103 61 Ep-6-1 0.110 0.017 0.127
Ep-6-2 170 156 · 103 98 Ep-6-2 0.021 0.005 0.026
Lav-2-1 160 147 · 103 92 Lav-2-1 0.133 0.022 0.155
Lav-2-2 320 294 · 103 185 Lav-2-2 0.012 0.004 0.016
Ep-3b-1 140 129 · 103 81 Ep-3b-1 0.097 0.014 0.111
Ep-3b-2 140 129 · 103 81 Ep-3b-2 0.011 0.004 0.015

384
Table 4. Fragility curve parameters (the mean Ŝd, ds in Table 9. Damage state probabilities for building Ep-3b-1,
meters). where complete structural damage was observed.

URML RM2L Slight Moderate Extensive Complete

Ŝd,ds βds Ŝd,ds βds
URML 0.989 0.946 0.797 0.550
Slight damage 0.0081 1.15 0.0147 1.14 RM2L 0.962 0.922 0.711 0.347
Moderate damage 0.0165 1.19 0.0234 1.10
Extensive damage 0.0411 1.20 0.0587 1.15
Complete damage 0.0960 1.18 0.1600 0.92
Table 10. Damage state probabilities for building Ep-3b-2,
where complete structural damage was observed.

Slight Moderate Extensive Complete

4 RESULTS: DAMAGE STATE PROBABILITIES
URML 0.700 0.465 0.198 0.057
Tables 5–10 give the computed vulnerability of the six RM2L 0.503 0.340 0.116 0.005
different buildings in terms of the probability of reach-
ing certain damage states. For unreinforced masonry
buildings the damage state probabilities exceed those
of the reinforced buildings, which is closer to the actual
damage. Building Ep-6-1 was completely damaged. The
probability of complete damage from the HAZUS
fragility curve for unreinforced masonry buildings
Table 5. Damage state probabilities for building Ep-6-1, (URML) is 59%. Building Ep-6-2 was heavily dam-
where complete structural damage was observed. aged and partially devastated, which is predicted by
Slight Moderate Extensive Complete
the URML curve by a probability of 34%.
Building Lav-2-1’s was completely damaged. The
URML 0.992 0.957 0.826 0.593 computations predict a 66% probability of complete
RM2L 0.970 0.938 0.748 0.400 damage based on the URML fragility curve. Build-
ing Lav-2-2 was, on the other hand, moderately to
lightly damaged. The computations indicate a 53%
probability of moderate damage.
Table 6. Damage state probabilities for building Ep-6-2, Buildings Ep-3b-1 and Ep-3b-2 were both com-
where there was heavy damage and partial devastation. pletely damaged, the computations from the URML
Slight Moderate Extensive Complete
fragility curves give respectively 55% and 6% proba-
bility of this damage state occurring.
URML 0.837 0.639 0.341 0.128
RM2L 0.681 0.527 0.231 0.002

5 DISCUSSION AND LIMITATIONS

Table 7. Damage state probabilities for building Lav-2-1, Assuming un-reinforced structures (URML), the
where complete structural damage was observed. HAZUS fragility curves give the probability of reach-
ing the correct damage level by 34–66%, with the
Slight Moderate Extensive Complete exception of building Ep-3b-2 with 6%. Building
Ep-3b-2 was completely destroyed even though it was
URML 0.995 0.970 0.864 0.656
RM2L 0.980 0.957 0.799 0.484 hit by less than a third of the force which hit building
Ep-3b-1, therefore it is appropriate to believe that this
building is not assigned the correct class and must be
left out of the further discussion.
Table 8. Damage state probabilities for building Lav-2-2, The damage state probabilities for the remaining
where moderate to light structural damage was observed. five buildings must be evaluated, so that one may judge
whether they are applicable or not. The Sarno debris
Slight Moderate Extensive Complete flow used to implement the model is a real case; there-
URML 0.722 0.490 0.216 0.065
fore the actual damage states are known with certainty.
RM2L 0.530 0.365 0.130 0.062 When using the vulnerability model to estimate the
damage from the exact landslides which hit Sarno in

385
1998, one expects the probability for the actual dam- also varies, but assuming Rmax = 2 is conservative
age state to be sufficiently high, so that the model may for any case.
be used.
The question is: What is a sufficiently high prob-
ability? For instance, for the case of the completely 6 CONCLUSION
damaged building Ep-6-1, the probability of complete
damage using the vulnerability model was 59%. Is a The proposed model for assessing the vulnerability of
59% probability of complete damage high enough to structures hit by debris flows has been tested for the
lead to the necessary interventions to prevent damage debris flows in Sarno area, Italy, 5–6 May, 1998. Six
which may happen (as seen in the Sarno case)? different structures with varying damage levels were
One may argue that the model gives imprudently tested. The model gave the probabilities between 34%
low values for the actual damage state arising. If and 66% for reaching the damage levels which actually
so the probabilities should be adjusted, so that, for occurred for five of the six structures. For example for
instance, a 40% probability of a certain damage state is a building which was completely damaged in reality,
equivalent to a higher probability in reality. This may the model gave 66% probability of being completely
also be seen from the values in the fragility curves. damaged for the same debris flow.
Table 11. shows how large the spectral displacement The uncertainties of the model are fairly large,
of an un-reinforced building must be to obtain cer- which makes it difficult to see where the possible dis-
tain probabilities of reaching a given damage state. crepancy lies, whether it is in the input parameters or
The spectral displacements which are required for a the model itself. Although the case had much informa-
two-story building to get 80% or higher probability of tion, it did not give the full information needed to limit
complete damage are unrealistically large. the input uncertainty. To decide whether the method of
The reasons for this discrepancy may either lie in the assessing vulnerability to debris flows is worth using,
model itself or in the input, which leaves the following input from more case studies must be tested, prefer-
options: ably using input more accurate than the Sarno case, to
locate the error sources. A sensitivity analysis should
1. The model gives non-conservative values. The out-
also be done. For a more detailed damage state pre-
put must be correlated to realistic values, if it is
diction the structural model may be extended to more
possible.
degrees of freedom, which would show the effect of
2. The building classes chosen to find the fragility
the loading point and structure height.
curve parameters do not reflect the actual behaviour
The method could be included in a statistical model,
of buildings subject to debris impact, more exact
to give the degree of reliability of the output. To
data must be found on the structures in the area.
increase the applicability, the model may be extended
3. The displacement calculation models are erroneous
to consider punching and rupture of structural ele-
and must be refined.
ments, which will be the case for inhomogeneous
Using the model in practice a sensitivity analysis debris flow (i.e. with embedded rocks) and may be
of the input should be implied to see how the result is more critical. By using GIS software the model could
affected by variations. The impact force, for instance, be used to assess the vulnerability of several structures
may be up to twice the magnitude suggested in in an area exposed to debris flow.
Eq. (5) (Armanini & Scotton, 1992, Lin et al. 1997).
Using double dynamic force for the heavily damaged
building Ep-6-2 gives a 54% probability of being REFERENCES
extensively damaged compared to 34% with the for-
mer force. The impulse force shape (see Figure 2) Armanini, A. & Scotton, P. 1992. Experimental analysis on
the dynamic impact of a debris flow on structures. In Pro-
ceedings, International Symposium, INTRAPREVENT
Table 11. Spectral displacements (in meters) for different 1992, Bern, volume 6, pages 107–116.
damage state probabilities for URML structures. Cardinali, M., Reichenbach, P., Guzzetti, F., Ardizzone, F.,
Antonini, G., Galli, M., Cacciano, M., Castellani, M. &
Damage state probability Salvati 2002. A geomorphological approach to the esti-
for URML structures 50% 80% 95% mation of landslide hazards and risks in Umbria, Central.
Natural Hazard and Earth System Sciences 2(1):57–72.
S.D. giving slight damage 0.008 0.021 0.053 Chopra, A. 2001. Dynamics of Structures, Theory and Appli-
S.D. giving moderate damage 0.016 0.045 0.115 cations to Earthquake Engineering, 2nd edition. New
S.D. giving extensive damage 0.041 0.111 0.292 Jersey: Prentice Hall.
S.D. giving complete damage 0.095 0.255 0.659 Clough, R. & Penzien, J. 1993. Dynamics of Structures, 2nd
edition. Singapore: McGraw-Hill, Inc.

386
European Commission 2003. Lessons Learnt from Landslide Management. www.engmath.dal.ca/tc32/2004Glossary
Disasters in Europe. In Hervás, J. (ed.), Technical report, _Draft1.pdf.
Joint Research, NEDIES PROJECT. Lin, P., Chang, W. & Liu, K. 1997. Retaining Function of
FEMA 1995. Engineering Principles and Practices for Open-Type Sabo Dams. Debris-Flow Hazards Mitigation:
Retrofitting Flood Prone Residential Buildings. FEMA Mechanics, Predication and Assessment, 636–645.
259. Remondo, J., Bonachea, J. & Cendrero, A. 2005. A statistical
Glade, T. 2003. Vulnerability Assessment in Landslide Risk approach to landslide risk modelling at basin scale: from
Analysis. Die Erde (Beitrag zur Erdsystemforschung) landslide susceptibility to quantitative risk assessment.
134:121–138. Landslides, 2:321–328.
HAZUS 2006. Multi-hazard Loss Estimation Methodology, Wong, H.N., Ho, K.K.S. & Chan, Y.C. 1997. Assessment
Earthquake Model, HAZUS MH-MR2, User Manual. of Consequences of Landslides. In: Cruden, D.M. &
FEMA and NIBS. Fell, R. (eds.), Landslide Risk Assessment. Proceedings
HAZUS 2006b. Multi-hazard Loss Estimation Methodol- of the Workshop on Landslide Risk Assessment, Hon-
ogy, Earthquake Model, HAZUS-MH-MR2, Technical olulu, Hawaii, USA, 19–21 February 1997. Rotterdam:
Manual. FEMA and NIBS. Balkema.
Heinimann, H.R. 1999. Risikoanalyse bei gravita- Zanchetta, G., Sulpizio, R., Pareschi, M., Leoni, F. &
tiven Naturgefahren—Fallbeispiele und Daten. Umwelt- Santacroce, R. 2004. Characteristics of May 5–6, 1998
Materialien 107/II. Bern. volcaniclastic debris flows in the Sarno area (Campania,
Hungr, O., Morgan, G. & Kellerhalls, R. 1984. Quanti- southern Italy): Relationships to structural damage and
tative Analysis of debris Torrent Hazard for Design of hazard zonation. Journal of volcanology and geothermal
Remedial Measures. Canadian Geotechnical Journal 21: research 133:377–393.
663–667.
ISSMGE TC32 2004. Glossary of Risk Assessment
Terms, Technical Committee on Risk Assessment and

387
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Engineering geological study on a large-scale toppling deformation

at Xiaowan Hydropower Station

Runqiu Huang, Genlan Yang, Ming Yan & Ming Liu

State Keyl Laboratory of Geohazards Prevention, Chengdu University of Technology, Chengdu, China

ABSTRACT: Toppling failure is one of deformation and failure types of bedded rock mass with high dip
angle. It usually develops near the ground surface. Based on the field investigation, we discovered a large-scale
toppling deformation occurred at the dam site, which developed to a great depth of about 200 m. In this paper,
the deformation phenomena revealed by surface excavation and underground galleries are described in detail
including its dip angle changes and rock mass structure variation. In terms of them, a mechanism model of the
slope deformation is established and its formation process is analyzed.

1 INTRODUCTION 2 GEOLOGICAL CONDITIONS

Toppling is a common kind of deformation and failure Xiaowan Hydropower Station is located in the south-
for rock slopes. Toppling deformation mainly develops western part of Yunnan Province, the position between
in the slopes of platy rock mass formed by steep- Fengqing County and Nanjian County. It lies at the
dip-angle cracks and fissures, and the angle between south of famous Sanjiang Structure Belt in southwest
structure plane strike and slope surface one is less region with complex geological structure. Lantsang
than 30 degrees. Its formation mechanism is that due have been cut into deep vale, and magnificent deep
to bending action caused by the deadweight the steep cut gorge landform was formed (Chen, et al., 2004;
layered rock mass faces are free, much like cantilever Xu, et al., 2004; Cheng, et al., 2005). The dam site of
bending, and gradually develops in the inner slope. Xiaowan Hydropower Station lies at the position that
Toppling deformation finally happens (Huang et al, is 3.85 m far from the joining point between Lantsang
1996, 2000; Wang et al, 1996; Rui et al, 2001; Wang and its branch, Heihui River. Both sides’ valley in
et al, 2001; Sun, 2002). Toppling deformation usually the area of the dam is basically symmetrical, and V
occurs at the after-edge of the slope, especially margin shape gorge is presented. There are grand hills and
of the steep cliff. Limited development depth is a char- steep slope in both sides, and average slope gradient
acteristic, commonly less than 50 m, and vertical depth ranges from 35 degrees to 50 ones. The river elevation
is more than 60 m. The type of toppling deformation is 987.92 m in the region of the dam, and the peak
occurs in the local area of shallow earth surface, and elevation is 2168.8 m. The difference between them is
after-edge part is easily destroyed, and local collapse more than 1000 m.
and landslip are caused. The paper studied the top part of Yinshui Gully
However, in recent years we researched the accu- accumulation before the dam on the left bank of
mulation on the left bank of Xiaowan Hydropower the Xiaowan Hydropower Station, Lantsang (See
Station, and found large-scale deep-layer bending Figure 1). The elevation difference between deforma-
toppling deformation failure phenomena. Its maxi- tion’s lower boundary and the riverbed is about 500 m.
mum horizontal growth depth is near 200 m, and Its north is the gentle-slope flat of Longtai Road, No 2
present most vertical developing depth is 160 m hill girder’s top lie in its south, and there is Malutang
(EL.1460 m∼EL.1600 m). It is meaningful to under- Gully in its northwestern direction. The growth depth
stand deformation failure mechanism of high rock of the large-scale deformable body varies from 150 m
slopes under complex conditions that the characteris- to 200 m, maximum level growth depth across the river
tics of a large-scale toppling failure and its formation direction is up to 180 m, and vertical growth depth is
conditions are disclosed. about 160 m.

389
 The bedrock below the deformation body is mainly body, and extruding face well developed. According
composed of biotite granite gneiss including lamel- to growth scale, the structure planes most belong to
lar lentiform schist. The trend of the granite gneiss is IV-class structure planes. They usually developed
nearly EW direction with vertical dip angle, and its along the schist, thickness and space length of which
depth is about 400 m. It distribute along Longtangan greatly varied. The thickness commonly ranges from
Gully and Gouzi Gully. 20 cm to 50 cm, and the maximum thickness arrives
The most fracture developed in the deformable body at 2 to 3 m. Their space length is usually from 9 m to
is Fault F7, which belongs to second-class structure 14 m, and the maximum length is up to 18 m.
plane with the altitude of N80∼90◦ E/SE∠80∼82◦ . During excavating the rock slopes the different alti-
The width of the crash zone ranges from 3.8 m to 5.0 m, tudes between the bedrock and the top rock mass of
and the effect width varies from 20 m to 30 m. The Yinshui Gully is found, as indicates that there is per-
fault spread out in the center of Yinshui Gully, which haps a special deformation failure mechanism for the
is located in the northern part of the toppling deforma- slope.
tion failure zone. Due to the debris covered the fault In order to investigate deformation failure mech-
crops out in the excavated slope surface. The small anism of the deformable body, we added a series of
fault in the original rock underlies the deformable exploration tunnels in terms of construction of the
drainage tunnels. These tunnels, which is seen in
Figure 2, discloses the multiform deformation and
failure phenomena of the rock mass.

3 GEOLOGICAL PHENOMENA
OF A LARGE-SCALE TOPPLING
DEFORMATION FAILURE

3.1 Deformation failure phenomena revealed

by surface excavation
During excavating the slopes of No. 2 ridge and
Yinshui Gully, altitude of the slope surface between
EL. 1600 and 1460 m is different that of the original
Figure 1. Plan view of Yinshui Gully deformable body
rocks, i.e., there is an evident trend to topple northward
with large-scale toppling deformation: 1. rock layer symbol;
2. accumulation boundary; 3. stratum boundary; 4. contour for the rock layers (See Figure 3).
line; 5. toppling deformable body; 6. ridge number; 7. small The rock stratum’s dip angle in the deformable
fault and its number. body and changing characteristics of the rock mass
structure on the surface are analyzed on the basis of

TNND

0+125.9m
0+105m

0+187m
TMD

0+134m

TA

TB
1 TB

2
TD

Figure 3. Geological sketch of slope surface excavation for

Yinshui Gully and No. 2 Ridge from the altitude 1460 m
Figure 2. Plan view of surface excavation and drainage to the altitude 1600 m above sea level: 1. cracks of original
tunnel distribution: 1. drainage tunnel and its number; rock; 2. schist interlayer; 3. small fault; 4. extruding plane;
2. excavated berm line and its elevation; 3. profile line. 5. boundary of the deformable body.

390
Figure 4. Geological sketch of profile A-A from the altitude 1500 m to the altitude 1520 m above sea level g1∼g9. Schist
interlayer; f1–f4. small fault, extruding belt.

geological investigation and sketch figures from the

most typical excavated profile between elevation 1500
m and 1520 m in the deformable body.

3.1.1 Change of the altitude

It can be seen from Figure 4 that the dip angle of the
rock stratum gradually decreases along the berm from
south to north. Figure 5 shows that the dip angle var-
ied with the distance from south to north from the
initial position of TA drainage tunnel (EL. 1500 m).
In Figure 5 the rock stratum’s dip angles between
0 m to 15 m keep unchanged, and are all more than
80 degrees, which belong to normal rock mass alti-
tude. From 15 m to 55 m the dip angles show great
changes, and the dip angle decreases to 50 degrees
from 85 degrees. In Figure 5 the curve between 55 m
to 150 m is smooth, and the dip angles are about
40 degrees. There is no obvious layer structure to
the north of 150-m point, and the layer structure is
basically even. Figure 5. Curve showing the change of the dip angles along
the berm at the altitude 1500 m.

3.1.2 Rock mass characteristics

Figure 4 indicates that there are normal rock masses to the elevation of 1380 m. The serial number is sep-
the south of TA drainage tunnel exit. In Figure 5 g9 to arately TD, TB, TA, TMD and TNND from bottom
the north of the tunnel exit, which is 4.5 m to 5 m wide, to top, the elevation of which is separately 1420 m,
is the southern boundary of toppling failure. There 1460 m, 1500 m, 1540 m and 1580 m. Toppling defor-
occur normal rock masses between 0 m to 15 m from mation mainly developed in the drainage tunnel TB,
the initial point of the tunnel exit. Toppling-falling TA and TMD above the elevation of 1460 m. In terms
rock is distributed from 15 m to 150 m. There is the of field survey, the rock stratum’s dip angle, rock mass
accumulation of soil including block stone (toppling structure characteristics and crack development of the
failure after-edge) to north of the 150 m point. above drainage tunnels are as follows.
The geological information disclosed by the surface
excavation indicates that large-scale toppling defor- 3.2.1 Dip angle change of the rock strata
mation phenomena mainly occur above the elevation Figure 6 describes the relationship between rock stra-
of 1460 m at Yinshui Gully. No marked toppling phe- tum’s dip angle of different altitude and tunnel depth.
nomena are found in the slope surface between El. For TD drainage tunnel from elevation 1420 m the
1380 m and El. 1460 m. dip angles show no obvious change, and don’t vary
with depth. This indicates that a large-scale toppling
failure hasn’t occurred below elevation 1420 m or TD
3.2 Phenomena disclosed by drainage tunnel
drainage tunnel. According to data statistic the dip
and sub-tunnel construction
angles of the rock strata are all less than 60 degrees in
Five drainage tunnels were constructed for excavated TB drainage tunnel (elevation 1460 m). At the position
slope surface of Yinshui Gully and No. 2 Ridge above of 0+187 m the dip angles begin to well varies with

391
 Table 1. The relationship between the crack aperture and the
horizontal depth.

Distance apart from the slope

surface (m) Opening degree (cm)

100 2∼3
107 0.5∼1
109 1∼2
110 5
113 1∼1.5
118 2∼5
120 80∼100
Figure 6. Graph of dip angle varying with the drainage tun-
nel depths at different levels: 1. TD drainage tunnel (1420 m);
2. TB drainage tunnel (1460 m); 3. TA drainage tunnel The altitude of the rock strata: the altitude is
(1500 m); 4. TMD drainage tunnel (1540 m). dispersed as compared with the original rock. The
strike of the original rock strata is commonly NWW
direction. However, in the local part of collapse accu-
the tunnel depth, and decrease to less than 40 degrees. mulation the strike is a NEE direction, even NE
The boundary between toppling loose rock mass and direction. The altitude of the rock strata is SW direc-
normal one is determined by the cracks and fissures tion, and the dip angles range from 40 degrees to 60
of the drainage tunnels, which is the position of main ones. And there is a trend to gradually decrease from
drainage 0+134 m. The rock stratum’s dip angles in south to north. These show that the local part of the
TA drainage tunnel (elevation 1500 m) investigated rock mass exhibited deflexion to some degrees during
exhibit approximate linear fall, as shows typical char- toppling-falling-covering.
acteristics of toppling failure. In TMD drainage tunnel In the toppling deformable body the joints
(elevation 1540 m) there is a trend to decrease for the mainly include SN-direction steep joints and EW-
dip angles, however, the dip angles are more than those direction bedding joints. Secondly there occur gentle
of another two tunnels, and are more than 60 degrees. SN-direction joints and EW-direction joints. The
SN-direction joints’ dip angles vary from 20 degrees
3.2.2 Crack development characteristics to 35 degrees, and the mean value is 30 degrees. It is
In the most typical TB drainage tunnel (elevation found in terms of joint growth contrast between the
1460 m) the cracks’ width at the position of main original rock and the accumulation that there is dif-
tunnel 0+170 m almost are more than 2 cm (See ference between them. One is that joints’ trace length
Table 1). From Table 1 opening degree of the cracks in the accumulation is greater than that of the orig-
is related with the length apart from the slope surface. inal rock. The other is that the difference from the
As a whole, from 100 m to 125 m far from the slope joint growth frequency is not greater than that from
surface, opening degrees of the cracks increase with the original rock. These indicate that there is an even
depth. At the position 120 m far from the slope surface, trend for toppling accumulation as compared with the
the cavity formed by SN-direction unloading cracks is original rock.
80 to 100 cm long, and show an inverse wedge shape. In the toppling deformable body the physical
There occur rock mass unloading phenomena to the properties are different for different parts. As a whole
south of the point, and a few SN-direction unloading there are greater deflection degrees, smaller blocks,
cracks grow to the north with small opening cracks, and stronger weathering from south to north. At
width of which commonly ranges from 1 mm to 5 mm. the after-edge of the falling-covering accumulation
Distributing characteristics of these cracks shows that the rock blocks usually show full weathering, and the
main tunnel 0+170 m of TB drainage tunnel is the lamination of the accumulation is difficult to identify.
most bending plane of rupture of toppling failure at
the elevation of 1460 m.
4 GROWTH MODEL AND FORMATION
CONDITION ANALYSIS OF THE
3.3 Total characteristic analysis of toppling LARGE-SCALE TOPPLING DEFORMATION
deformation
4.1 Growth model analysis of a large-scale
According to the geological information disclosed by
toppling deformation
surface investigation and tunnels, large-scale toppling
and deformation rock mass researched presents the According to geological information of surface and
characteristics as follows. drainage tunnels deformation failure at the top part of

392
Yinshui Gully, Xiaowan Hydropower Station, can be toppling-falling accumulation; D is the colluvial
generalized into a toppling failure model described in deposit.
Figure 7. In the region of large-scale toppling defor-
mation failure the granite gneiss is cut by the schist
4.2 Forming condition of large-scale toppling
interlayer, and thus stratoid rock mass is presented.
deformation
The rock layer is approximately vertical with near EW
direction of the rock layer surface. During releasing Formation of the large-size toppling deformation fail-
the stress with the river valley incision, ductile bending ure at last has two basic conditions. One is the slope
toppling deformation is generated due to the long-term structure, i.e., stratoid structure and the structure
self-weight stress from the approximate layered rock growth with gentle dip angle, the other is the favorable
mass. The schist interlayer g9 is the boundary of top- landform condition in the failure region, i.e., gently
pling deformation, which is 4.5 m to 5 m wide. It is platform developed at the after-edge of the toppling
difficult to observe the whole failure deformable body deformable rock mass.
due to Occurrence of the gentle slope platform.
In terms of the ground and underground deforma- 4.2.1 The conditions about the slope structure
tion failure we can obtain the deformation failure mod- Underlying rockbed of the Yinshuigou accumulation
els described in Figure 7, i.e., toppling failure models. has a strike of EW-direction with vertical dip angle.
The models have the characteristics as follows The schist in the rockbed well develops. Its width
much varies, and the rock layer is only several cm
1. Great deformation area. In the horizontal range the thin, and even 2 m to 3 m deep. The space between
deformation’s depth ranges from 150 m to 200 m the schist gradually increases from north to south. In
in the slope, and vertical growth depth is about the rockbed slope the average space between the schist
200 m. Therefore, this is a kind of toppling defor- ranges from 9 m to 14 m, and the most one is up to 18 m.
mation that is different from common large-scale However, in Yinshuigou accumulation, the space is
deformable bodies. smaller, and the least is 2 m. Growth of the schist
2. Marked ductile deformation characteristics. The interlayer makes the rock mass in the deformation
rock layer shows big deformation range, and failure zone being layered structure, as provides struc-
toppling-bending is very strong. The dip angle of tural conditions for the large-size toppling deformation
the rock strata becomes approximate level from failure.
being vertical. This is obviously related with weak
schist growth from local parts of the slope, and 4.2.2 Landform conditions
great decrease of the whole rock strength. In the region of the large-size toppling deformable
3. Obvious zoning. In terms of strong degrees of body, underlying rock bed plane is the U-shape trough
deformation, the area can be divided into A, B, C that developed by F7 fault (See Figure 8). In the trough
and D four regions from inner to outside, where, A above the elevation of 1460 m the slope is gentle in the
is the vertical rock mass that no deformation hap- longitudinal direction, and that below the elevation of
pens; B is the toppling loose rock mass; C is the

Figure 7. Yinshuigou toppling deformable body structure

of Xiaowan Hydropower Station in Lantsang: A—Erect
rock mass; B—Toppling loose rock mass; C—Falling accu- Figure 8. The sketch of toppling failure initial form before
mulation; D—Drift bed; ①—Plane of maximum bending toppling failure: 1. EW-direction bedding structure plane;
rupture. 2. EW-direction low-angle structure plane; 3. Lantsang.

393
1460 m is steep. The degree of slope from the origi- Wang Jianfeng, Wilson H Tang & Cui Zhengquan. 2001.
nal rock plane ranges from 25 to 30. Underlying rock Stability analysis of Toppling Failure of block rock slopes.
plane below gently platform of Longtai Road, the top The Chinese Journal of geological hazard and control, 12
part of No. 0 Ridge, is about the elevation of 1460 m (4): 1–8 (in Chinese).
occurrence of the gentle platform from the rockbed Rui Yongqin, He Chunning, Wang Huiyong et al. 2001.
Analysis of Deformation and Failure Development of the
from Yinshuigou to Longtai Road Village provides a Large-scale Toppling-sliding Slope under Mining. Jour-
deformation space for the large-size toppling failure. nal of Changsha Jiao-tong University, 17 (4): 8–12 (in
Chinese).
Wang Xiaogang, Jia Zhixin, Chen Zuyu et al. 1996. The
5 CONCLUSION Research of Stability Analyses of Toppling Failure of
jointed rock slops. Journal of Hydraulic Engineering, (3):
According to geological information about surface 7–15 (in Chinese).
excavation and drainage tunnel construction, knowl- Sun Dongya, Peng Yijiang & Wang Xingzhen. 2002. Appli-
edge on the deformation failure rock mass at the top cation of DDA method in stability analysis of topple rock
slope, Rock Mechanics and Engineering, 21 (1): 39–42
Yinshuigou can be obtained as follows: the dip angle (in Chinese).
of the rock layer plane gradually changes from being Chen Hongqi & Huang Runqiu. 2004. Stress and flexibil-
steep to being gentle. The rock structure shows a trend ity criteria of bending and breaking in a countertendency
to change massive texture into scattered structure from layered slope. Journal of Engineering Geology, 12 (3):
inner to outside. In the deformable rock mass the 243–246 (in Chinese).
unloading cracks well develop neat the inner bound- Xu Peihua, Chen Jianping, Huang Runqiu & Yan Ming.
ary that opening degree of the rock mass is great. The 2004. Deformation mechanism of Jiefanggou high steep
deformable body can be considered as the product of dip slope in Jinping Hydropower Station. Journal of
the large-size toppling deformation in terms of dip Engineering Geology, 12 (3): 247–252 (in Chinese).
Cheng Dongxing, Liu Da’ an, et al. 2005. Three-dimension
angle change of the deformable rock mass, rock mass numerical simulation of deformation characteristics of
structure and unloading crack growth characteristics. toppling rock slope. Journal of Engineering Geology, 13
This large-size toppling deformation usually occurs in (2): 222–226 (in Chinese).
the high slopes that flexible rock layers well develop or
there are many flexible rock layers. And the landform
has accumulating conditions of deformable toppling
bodies with favorable free faces.

REFERENCES

Huang Runqiu, Wang Shitian, Hu Xiewen et al. 1996. Study

on the main Engineering Geological Problems of Xiaowan
high arch dam. Chengdu: Southwest Jiao-Tong University
Press (in Chinese).
Huang Runqiu. 2000. Time-dependent deformation of a
high rock slope and its engineering-geological signifi-
cance. Journal of Engineering Geology, 8 (2): 148–153
(in Chinese).

394
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Characterization of the Avignonet landslide (French Alps)

with seismic techniques

D. Jongmans, F. Renalier, U. Kniess, S. Schwartz, E. Pathier & Y. Orengo

LGIT, Université Joseph Fourier, Grenoble Cedex 9, France

G. Bièvre
LGIT, Université Joseph Fourier, Grenoble Cedex 9, France
CETE de Lyon, Laboratoire Régional d’Autun, Autun cedex, France

T. Villemin
LGCA, Université de Savoie, France

C. Delacourt
Domaines Océaniques, UMR6538, IUEM, Université de Bretagne Occidentale, France

ABSTRACT: The large Avignonet landslide (40 × 106 m3 ) is located in the Trieves area (French Alps) which
is covered by a thick layer of glacio-lacustrine clay. The slide is moving slowly at a rate varying from 1 cm/year
near the upper scarp to over 13 cm/year at the toe. A preliminary geophysical campaign was performed in order
to test the sensitivity of geophysical parameters to the gravitational deformation. In the saturated clays where
the landslide occurs, the electrical resistivity and P-wave velocity are little affected by the slide. On the contrary,
S-wave velocity (Vs) values in the first ten meters were found to be inversely correlated with the measured
displacement rates along the slope. These results highlight the interest of measuring Vs values in the field for
characterising slides in saturated clays and of developing techniques allowing the 2D and 3D imaging of slides.

1 INTRODUCTION Schmutz et al. 2000, Lapenna et al. 2005, Grandjean

et al. 2006, Méric et al. 2007) seismic and electrical
Slope movements in clay formations are world methods were successfully used in clay deposits for
widespread and usually result from complex defor- distinguishing the mass in motion from the unaffected
mation processes, including internal strains in the ground.
landslide body and slipping along rupture surfaces The aim of this study is to test the sensitivity of
(Picarelli et al. 2004). Such mass movements are the main geophysical parameters (with a focus on the
likely to generate changes in the geophysical param- shear wave velocity) to the clay deformation generated
eters characterizing the ground, which can be used by the landslide of Avignonet where geotechnical and
to map the landslide body. Since the pioneering geodetic data are available.
work of Bogoslovsky & Ogilvy (1977), geophysical
techniques have been increasingly but still relatively
little used (or referenced) for landslide investigation 2 THE AVIGNONET LANDSLIDE
purposes (McCann & Forster 1990, Jongmans &
Garambois 2007). The large Avignonet slide (40.106 m3 ) located in the
The recent emergence of 2D and 3D geophysical Trieves region (French Alps, Figure 1) was studied.
imaging techniques, easy to deploy on slopes and This 300 km2 area is covered by a thick Quater-
investigating a large volume in a non-invasive way, nary clay layer (up to 200 m) deposited in a glacially
has made more attractive the geophysical methods for dammed lake during the Würm period (Giraud et al.
landslide applications. One of the key factors con- 1991). These clayey deposits overlay compact old allu-
trolling the success of geophysical techniques is the vial layers and marly limestone of Mesozoic age, and
existence of a contrast differentiating the landslide are covered by thin till deposits. After the glacier
body to be mapped. In the past (Caris & van Asch 1991, melting, rivers have cut deeply into the geological

395
 downstream. South of this slow moving slide, a quick
mudslide (Harmalière, Figure 1) occurred in March
1981, creating a head scarp of 30 m and affecting a
surface of about 450,000 m2 (Moulin & Robert 2004)
in the same material. Between 1981 and 2004, the
head scarp has continuously regressed with an aver-
age of 10 m/year in a north-eastward direction. It
now intersects the limit of the Avignonet landslide.
The source zone displays complex deformation pat-
terns, including rotational slips, cracks, slumps, and
translational failures. In the track and at the toe, the
slide evolves into a flow during heavy rainfalls, con-
tributing to the depletion of the landslide mass and the
southward erosion process.

3 GEOTECHNICAL AND GEODETIC

INVESTIGATION

The Avignonet landslide which affects the hamlet of

Le Mas (Figures 1 and 2) was investigated by four
boreholes (T0 to T3), equipped with inclinometers
(see Figure 4 for location and Table 1 for the main
results). The contact with the alluvial deposits was
Figure 1. Geological map with the location of the Avignonet
found at 14.5 m, 44.5 m and 56 m in T2, T3 and T1
and Harmalière landslides, and of the area investigated by respectively, whereas T0 was still in the clay deposits
Lidar. at 89 m, in agreement with the westwards thickening of
this formation (Figure 2). Inclinometer data revealed
several rupture surfaces, at a few m depth (T0 and T2),
between 10 m and 17 m (T0, T1, T2, T3) and up to
42 to 47 m (T1 and T0) (Table 1, Lorier & Desvar-
reux 2004). Of particular importance is the presence
of a major active slip surface found at 13 m depth in
borehole T2 which is located in the more active area.
Piezometric measurements showed the presence of
a very shallow water table (1 to 3 m below the ground
level). No geotechnical investigation was performed

Figure 2. EW geological cross-section over the hamlet of Table 1. Borehole and inclinometer results.
Le Mas (Figure 1.).
Depth of rupture
Geological formations surface

formations, triggering numerous landslides (Giraud T0 0–5 m: morainic deposits 5m

et al. 1991). Figure 1 shows the simplified geologi- 5–89 m: varved clays 10 m
cal map of the studied area, in the northern part of 47 m
the Trièves region, as well as the two main land- T1 0–5 m: morainic deposits 15 m
slides (Avignonet, Harmalière) occurring in the clay 5–56 m: varved clays 34 m
deposits. An EW synthetic geological section over the 56–59 m: alluvial deposits 42.5 m
Avignonet landslide is presented in Figure 2, showing T2 0–4 m: morainic deposits 1.5 m
the thickness variation of the clay layer, from 0 m to 4–14.5 m: varved clays 4m
more than 200 m. The translational Avignonet slide, 14.5–17 m : alluvial deposits 12 m
whose first signs of instability were noticed between
T3 0–4 m: morainic deposits 16 to 17 m
1976 and 1981 (Lorier & Desvarreux 2004), affects
4–44.5 m: varved clays
a surface of about 1 × 106 m2 , with a global east- 44.5–59 m: alluvial deposits
ward motion to the Drac valley which is dammed

396
on the Harmalière landslide which does not threaten Inside the Avignonet slide, the DEM shows the
any property in the short term. presence of multiple parallel scarps down the slope
with a spacing of about 100 m. Scarp heights are higher
within the slide, suggesting that this latter regresses
3.1 Lidar acquisition toward the plateau at the west and that the motion
could be greater at the toe than at the crown. In the
A Lidar (Light Detection and Ranging) acquisition,
lower part of the hill, the slope increases due to the
covering the zone displayed in Figure 1, was per-
presence of consolidated alluvial layers over which
formed in November 2006 with the handheld airborne
the clay material flows (Fig. 2). The geometry of
mapping system Helimap system® (Vallet & Skaloud
the Harmalière landside is more elongated than the
2004). As regards the Lidar, the measurement unit
Avignonet one, with a the presence of multiple
is composed of 3 sensors: a GPS receiver, provid-
curved scarps in the source area and a funnel shaped
ing the position of the unit, an inertial measurement
track zone through which the material flows to the
unit which provides the orientation of the system, and
lake with a regular slope. To the south-west, the
a laser scanner unit measuring a point cloud of the
Harmalière landslide also intersects another land-
surface. The height of flight was of 500 m above
slide. This difference in the morphology and in the
the ground and allowed to acquire a density of one
mechanical behaviour between the two landslides,
point by square meter in average. The system dis-
developing in the same material, probably results
plays a high accuracy of ∼10 cm both in horizontal
from the disappearance of the crest made of bedrock
and vertical. The interpolated Digital Elevation Model
and alluvial layers to the south, removing the but-
(DEM) is shown in Figure 3 at a resolution of 2 m.
tress that prevents the deep sliding of the Avignonet
The DEM enlightens several landslide indicators. It
landslide.
clearly displays the crescent-shaped front scarp of
the Avignonet landslide which intersects the Harmal-
ière one to the South and another minor one to the 3.2 GPS measurements
North.
The Avignonet slide has been monitored by biannual
GPS measurements at 26 geodetic points since 1995,
while only 6 points were installed around the Harmal-
ière landslide, due to the strong deformation inside the
mass in motion. The locations of the GPS points are
shown on a DEM (Figure 3).
The velocity values at the surface of the Avignonet
landslide, averaged from the GPS measurements avail-
able (Figure 3) increase downhill, varying from 0 to
2 cm/y at the top to more than 13 cm/y in the most
active part of the toe. Most of the area is sliding south-
eastward, parallel to the general slope. In detail, the
deformation pattern is complex and velocity and direc-
tion of the ground movement are influenced by local
geological and morphological features. The concave
shape of the river bank below the landslide clearly
controls the slope orientation and consequently the
slide direction in the lower part of the hill where
the displacement vectors rotate. This morphology
seems to be linked to the presence of old and consoli-
dated alluvial deposits overlying the bedrock, around
which the clay slides (compare Figures 1 and 3).
At lower altitudes, the higher slide velocities mea-
sured with GPS are accommodated by scarps and
bulges spaced by less than 20 m. Strong velocity
contrasts are observed along the landslide toe, which
seem to be linked to slope angles which are higher
Figure 3. Shaded Lidar-derived DEM of the Avignonet and
in the southern part. The GPS measurement on the
Harmalière areas acquired from Helicopter. White circles are ridge between the Harmalière and Avignonet land-
campaign GPS stations. Black lines indicate mean velocity slides shows that the Harmalière lateral head scarp still
values measured by GPS from 1995 to present. Dashed lines actively moves backwards, involving material belong-
show the limits of the two landslide. ing to the Avignonet landslide. On the contrary, GPS

397
measurements northwest of the Harmalière landslide
exhibit little displacement. Finally, the three GPS
points located along the south-western limit shows
the presence of another active slide, south of the
Harmalière one.

4 GEOPHYSICAL PROSPECTING

A preliminary geophysical campaign was performed

in 2006–2007 in order to test the sensitivity of three
geophysical parameters (the electrical resistivity ρ, the
P-wave seismic velocity Vp and the S-wave seismic
velocity Vs) to the deformation resulting from the
slide. It turned out that, in such saturated clays, ρ and
Vp are strongly influenced by the water level and are
little affected by the landslide activity. On the contrary,
Vs showed significant variations both vertically and
laterally. For this reason, we have focused our study
on the Vs measurements.
Shear wave velocity Vs can be measured by a rela-
tively large number of methods including active source
techniques (borehole tests, SH-wave refraction tests,
surface wave inversion) and ambient vibration tech-
niques (Jongmans 1992, Socco & Jongmans 2004).
In the present study we apply the SH seismic refrac-
tion tomography method and the surface wave (SW) Figure 4. Shaded Lidar-derived DEM with location of the
inversion for deriving Vs values. Seismic refraction seismic profiles (P1 to P7) and of the boreholes (T0 to T3).
tomography consists in inverting the first arrival times
picked on all the signals recorded at all geophones for
different shots spread along the profile. In the SH case,
transverse horizontal ground motions are generated
using a sledge hammer hitting laterally a loaded plank
as a source. The picked first arrival times are inverted
using the Simultaneous Iterative Reconstruction Tech-
nique (SIRT, Dines & Lyttle 1979) and provide 2D Vs
images.
In the SW method, Rayleigh waves are generated
by a vertical shock and are recorded by vertical geo-
phones, together with P waves. SW dispersion curves
are computed using the f-k method, which assumes
plane wave propagation and no lateral seismic veloc-
ity variations (Socco & Strobbia 2004). Dispersion
curves are then inverted to get 1D Vs profiles, using
the Geopsy software (http://www.geopsy.org).
For this study, we performed five 115 m long seis-
mic profiles (P1 to P5) and two 470 m long profiles
(P6 and P7). The profile location is given in Figure
4. For the short profiles, we used 24 vertical geo- Figure 5. Seismograms along profile P1 for a shot at 0 m.
a) Vertical motions generated by a sledge hammer. A: P-wave
phones (4.5 Hz) and 24 horizontal geophones (14 Hz) first arrivals, B: Rayleigh waves. b) Horizontal motions for
spaced by 5 m, for recording Rayleigh waves and a SH source. C: SH waves.
SH waves, respectively. Shots were made every 15
m, with two offsets for SW recordings. Figure 5
shows the seismograms generated along profile P1 the Vs structure. The two long profiles (P6 and
by a vertical source and a horizontal SH source. In P7) were conducted with 48 vertical 4.5 Hz geo-
Figure 4a, one can distinguish the P waves from phones, using explosive sources 80 m and 50 m apart,
the Rayleigh waves which are inverted for retrieving respectively.

398
4.1 SH tomography
As both SW and refraction travel-time data inversions

RAYLEIGH WAVE VELOCITY (m/s)

are non unique problems, we performed a joint anal-
ysis of refraction data with surface waves, looking
for a common solution in the 1D part of the pro-
files, with local validation by borehole data (Renalier
et al. 2007). Figure 6 shows the Vs images obtained by
the SH refraction tomography method for profiles P1,
P3 and P5, which are respectively located outside the
landslides, on the Avignonet landslide and on the Har-
malière landside (Fig. 4). Below profile P1, located
on the Sinard plateau, Vs quickly increases to 550 m/s
at 5 m depth. Profile P3 exhibits a low-velocity layer
(Vs around 250 m/s) with a thickness of 14 m to 19 m
from E to W, overlying a more compact layer (Vs >
600 m/s). The depth of this velocity contrast coincides Figure 7. Comparison of the phase velocity dispersion
with the slip surface at 13 m found in borehole T2 curves calculated along profiles P1, P5, P6 and P7.
(Fig. 4). The low Vs values above the slip surface are
probably linked to internal strains in the mass resulting
from the slide. On the Harmalière landslide (P5), the
Vs parameter delineates three distinct layers: a very
slow (Vs between 80 and 200 m/s) 5 m thick layer,
the 115 m long profiles P1 and P5, all geophones were
overlying a 15 m thick layer around 250 m/s, over a
considered together. Dispersion curves were computed
more compact layer (Vs > 500 m/s). These results sug-
for both direct and reverse offset shots in order to check
gest the presence of a slip surface at 20 m deep. SH
the 1D hypothesis. For the 470 m long profiles P6 and
refraction tomography thus enlightens the evolution of
P7 seismogram examination and P-wave tomography
the Vs shallow structure depending on the state of the
images were used to define groups of geophones which
ground—from undisturbed at P1 to highly disturbed
roughly fit the requirement of 1D media: 3 groups
at P5.
along P6 (from West to East: A: 1–15, B: 19–30,
C: 38–48) and 4 groups along P7 (from West to East: A:
4.2 Surface wave interpretation 1–9, B: 9–28, C: 29–35, D: 35–44). For these groups,
dispersion curves were also computed for the two off-
Dispersion curves (Rayleigh fundamental mode) for set shots. All the dispersion curves are gathered on
profiles P1, P5, P6 and P7 are plotted in Figure 7. For Figure 7.
The two different types of profiles can be eas-
ily recognised by their frequency range: dispersion
curves for P6 and P7 (explosive sources) exhibit lower
frequencies than the curves for P1 and P5 (hammer
source). Despite this difference, the Rayleigh wave
velocity at high frequency (over 8 Hz) exhibits a signif-
icant decrease according to the position of the profile
on the two landslides, with the exception of P7C.
Rayleigh wave velocities at 14 Hz are divided by 3,
from 500 m/s out of the landslides (P1) to 150 m/s
on the Harmalière landslide (P5). These results agree
with the Vs images of Figure 6. At lower frequencies,
the curves do not display such decrease because they
are also influenced by the underlying higher velocity
alluvial layers and bedrock, which come up from more
than 200 m deep on the western part of the studied area,
to the surface at the eastern end of the landslides (see
Fig. 2).
In a second step, these dispersion curves
Figure 6. Vs seismic images (SH refraction tests) along were inverted using the Neighbourhood Algorithm
three seismic profiles. a) P1, b) P3 c) P5. RMS values are (Wathelet et al. 2004), giving a two layer Vs model
below 3%. estimate for each group of geophones.

399
 Such variations were not observed on P-wave velocity
and resistivity values. These results highlight the inter-
est of in-situ measuring Vs values for characterising
slides in such saturated clays and of developing tech-
niques allowing the 2D and 3D imaging of landslides.
The relationship between Vs values, deformation and
pore pressure should be investigated through labora-
tory tests in order to allow a quantitative interpretation
of the field results. Combining Vs imaging with mul-
titemporal remote sensing (satellite and aerial) giving
a continuous image of the displacement rates at dif-
ferent times would also allow a deeper insight into
the 3D deformation processes and pattern of the land-
slide. Aerial image archive and new Lidar acquisition
planned in autumn 2008 on the Avignonet and Harmal-
ière landslides will supply a global view on the past
and present day slide velocity.
Figure 8. Evolution of shear wave velocity values at 10 m
depth with slide velocities. Crosses: surface wave data. Dots:
SH refraction data. Vertical error bars on the Vs values ACKNOWLEDGMENTS
are indicated. Horizontal bars indicate the uncertainty range
on the displacement rate when no data are available close This work was financially supported by regional,
to the seismic profiles. An arbitrary low displacement rate national and European funds coming from the Conseil
value of 1 10−2 mm/y has been assumed for profile P1 Général de l’Isère, the region Rhône-Alpes, the Clus-
(Vs = 630 m/s). ter VOR (Vulnérabilité des Ouvrages aux Risques) and
the European project NERIES. The authors thank all
the people who participated to the field investigation,
4.3 Vs value interpretation as well as RTM (Restauration des Terrains en Mon-
All Vs values obtained by the two methods for each tagne) and SAGE (Société Alpine de Géotechnique)
profile or group of geophones are correlated with for providing the geotechnical data.
the displacement rates measured at the GPS points
(Fig. 3). For each SH refraction image, two Vs val-
ues were extracted at the third and the two thirds of the REFERENCES
profile. When no GPS point is close to the seismic pro-
file, the displacement rate is averaged using the two Bogoslovsky, V. & Ogilvy, A. 1977. Geophysycal methods for
closer data. Figure 7 shows the Vs value with error the inverstigation of landslides. Geophysics 42: 562–571.
bars at 10 m depth versus the ground displacement Caris, J.P.T. & van Asch, T.W.J. 1991. Geophysical, geotech-
nical and hydrological investigations of a small landslide
rate in a semi-logarithmic scale. The striking feature in the French Alps. Engineering Geology 31 (3–4):
is the regular decay of Vs values with the displacement 249–276.
rate, from 630 m/s far from the slide, to 225 m/s at the Dines, K. & Lyttle, J. 1979. Computerized geophysical
slide toe, where the slope surface is strongly deformed. tomography. Proceedings of the Institute of Electrical and
These results show that the gravitational deformation Electronics Engineers 67: 1065–1073.
strongly affects the shear wave velocity within the clay, Giraud, A., Antoine, P., van Asch, T.W.J. & Nieyuwenhuis,
which could be used as a parameter for mapping the J.D., 1991, Geotechnical problems caused by glaciolacus-
slide activity. trine clays in the French Alps: Engineering Geology 31,
185–195.
Grandjean, G., Pennetier, C., Bitri, A., Méric, O. & Malet,
J.P. 2006. Caractérisation de la structure interne et de l’état
5 CONCLUSIONS hydrique de glissements argilo-marneux par tomographie
géophysique: l’exemple du glissement-coulée de Super-
On the Avignonet landslide, Vs values at shallow depth Sauze. Comptes Rendus Geosciences 338 (9): 587–595.
(10 m) were found to be inversely correlated with dis- Jongmans, D. 1992. The application of seismic methods
for dynamic characterization of soils in earthquake engi-
placement rates measured by GPS, with a division by a neering. Bulletin of the International Association of
factor of almost 3 between the zones unaffected and the Engineering Geology 46 (1): 63–69.
ones strongly deformed by the landslide. This strong Jongmans, D. & Garambois, S. 2007. Geophysical investiga-
decrease of Vs values is probably linked to internal tion of landslides: A review. Bulletin Société Géologique
strains in the mass above the main rupture surface. de France 178 (2): 101–112.

400
Lapenna, V., Lorenzo, P., Perrone, A., Piscitelli, S., Rizzo, E. Renalier, F., Jongmans, D., Bièvre, G., ‘Schwartz, S. &
& Sdao F. 2005. 2D electrical resistivity imaging of some Orengo, Y. 2007. Characterisation of a landslide in clay
complex landslides in Lucanian Apennine chain, southern deposits using Vs measurements. 13th European Meeting
Italy. Geophysics 70: B11–B18. of Environmental and Enginering Geophysics, Istanbul,
Lorier, L. & Desvarreux, P. 2004. Glissement du Mas 3–4 September 2007.
d’Avignonet, commune d’Avignonet. Proceedings of the Schmutz, M., Albouy, Y., Guerin, R., Maquaire, O.,
workshop Ryskhydrogeo, Program Interreg III, La Mure Vassal, J., Schott, J.J. & Descloitres, M. 2000. Joint
(France). inversion applied to the Super Sauze earthflow (France).
McCann, D.M. & Forster, A. 1990. Reconnaissance geo- Surveys in Geophysics 21: 371–390.
physical methods in landslide investigations. Engineering Socco, L.V. & Jongmans, D. 2004. Special issue on Seismic
Geology 29 (1): 59–78. Surface Waves. Near Surface Geophysics 2: 163–258.
Méric, O., Garambois, S., Malet, J-P, Cadet, H., Guéguen Socco, L.V. & Strobbia, C. 2004. Surface-wave method for
P. & Jongmans, D. 2007. Seismic noise-based methods near-surface characterization: a tutorial. Near Surface
for soft-rock landslide characterization. Bulletin Société Geophysics 2: 165–185.
Géologique de France 178 (2): 137–148. Vallet, J. & Skaloud J., 2004. Development and Experiences
Moulin, C. & Robert, Y. 2004. Le glissement de l’Harmalière with A Fully-Digital Handheld Mapping System Operated
sur la commune de Sinard. Proceedings of the workshop From A Helicopter, The International Archives of the Pho-
Ryskhydrogeo, Program Interreg III, La Mure (France). togrammetry, Remote Sensing and Spatial Information
Park, C.B., Miller, R.D. & Xia, J. 1999. Multi-channel Sciences, Istanbul, Vol. XXXV, Part B, Commission 5.
analysis of surface waves. Geophysics 64 (3): 800–808. Wathelet, M., Jongmans, D. & Ohrnberger, M. 2004. Sur-
Picarelli, L., Urciuoli, G. & Russo, C. 2004. The role face wave inversion using a direct search algorithm and
of groundwater regime on behaviour of clayey slopes. its application to ambient vibration measurements. Near
Canadian Geotechnical Journal 41: 467–484. surface geophysics 2: 211–221.

401
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Deformation characteristics and treatment measures of spillway slope

at a reservoir in China

Nengpan Ju, Jianjun Zhao & Runqiu Huang

State Laboratory of Geo-hazard Prevention and Geo-Environment Protection, Chengdu University of Technology,
Chengdu, China

ABSTRACT: This paper presented the failure model of some oblique slope consisting of silt mudstone,
argillaceous sandstone and carbon shale interlayers, by virtue of analysis of the slope deformation charac-
ter. According to engineering geological conditions and analysis of cracks, rock mass relaxation and bedding
displacement in the slope, a conclusion was drawn that the deformation of the slope was mainly controlled
by weak interlayers and steep unloading cracks dipping out of slope. Owing to tensile stress and shear stress
concentration at the middle and upper of the slope along the bedding displacement belt, and the compressive
deformation of soft layer at the bottom of slope, rock blocks deform out of slope or along dip direction of
bedding planes. And the deformation develops upwards along rock beds and unloading cracks, which results in
step creeping and cracking deformation. Because it is cataclastic rock mass intensely weathered and strongly
off-loaded with many weak interlayers downstream slope, the potential landslide may slide along the weak layers.
So the control of the slope deformation should be concentrated on reduce-loading, together with pre-stressed
cable framework to control the deformation of weak layers.

1 INTRODUCTION (Yang Yonghong & Lv Dawei 2006, Sun Hongyue &

Shang Yuequan 1999, Fan Wen et al. 2000, Huang
Oblique slope means that strike of strata obliquely Zhengjia & Wu Aiqing 2001). Taking slope of spill-
intersects with strike of slope surface. Such slope is way of a hydropower station as an example, under
often high, steep and steady, and its deformation and detailed analysis of geological conditions and phe-
failure is mainly local falling (Yin Yueping 2005). nomena of deformation, this paper studied function
Under rapid basal sapping of river or excavation, of weak structure planes and unloading cracks, which
stress will change inside high and steep slope. Owing plays a controlling role in stability of slope. On the
to release stress, rock mass has to occur unloading basis of analysis of deformation and failure mode of
rebound deformation toward the free face of slope, slope, targeted treatment measures is proposed.
and stress field of valley will be adjusted. With this
process, the shallow slope within a certain range
can form steep and inclined unloading cracks in the
same direction as dip of slope. When slope consists 2 ENGINEERING GEOLOGICAL CONDITIONS
of soft and hard rocks interlayers, along the con- OF SLOPE
tact planes between soft rock and hard rock strata,
shear dislocation will happen due to released stress, The study area lies in the southeastern part of Sichuan
with successional tension cracks (Huang Runqiu et al. basin, steep terrain with asymmetrical ‘‘V’’ valley.
2001, Ding Xiuli et al. 2005). Such weak structure Slope is on the right bank of Taoxi River and the nat-
planes and unloading cracks often become controlling ural angle of slope is 35◦ to 55◦ . There are two gullies
structure planes which can control stability of high in excavated slope, which are located at 0+000 and
steep slope. When slope accrues deformation, such 0-045 respectively. Excavated slope is discount linear
cracks having the same strike as controlling structure and dip direction is 39◦ to 80◦ ; Along dip of slope,
planes commonly are formed on the surface of slope. the slope is divided into A, B, C, of which dip is 74◦ ,
So study of relation between slope deformation and 39◦ , 40◦ to 80◦ respectively. Height of slope is 88 m,
structure planes will be helpful to know deformation and 6 steps are designed in the process of excavation.
and failure mode of slope, which is very important The first level slope is sidewall slope of lock chamber
to ensure reasonable and effective slope treatment which is a vertical slope. Gradient of the second and

403
Figure 1. Plane sketch map of distribution of cracks.

the third level slopes is 1:0.5, and gradient of the rest

slopes above the fourth level is 1:0.75, see Figure 1.
The inclined slope consists of soft and hard interlay-
ers. Exposed bedrock is the middle Jurassic Qianfuyan
Formation silt mudstone, pelitic siltstone, carbona-
ceous shale, muddy siltstone, and carbonaceous shale.
They has a characteristic of inter-layered distribution,
occurrence of strata N 60◦ to 80◦ E/NW ∠ 20◦ to 40◦ .
The study area lies in northwest wall of Wen-
quanjing anticline and southeast wall of Pingloushan
syncline, with simple geological structure. F5 reversed
fault is found in a gully, on downstream side of Area A,
and its occurrence is N30◦ ∼70◦ E/NW 37◦ ∼45◦ , and
it tends to taper. Shale interlayers mostly develop into
bedding displacement zone, generally about 10 cm
thick. The total amount is more than 10; these weak
interlayers often control stability of slope.
Weak interlayers exist in the inclined slope which
consists of soft and hard interlayers. Aclinic depth
of unloading rock mass is about 22 m; aclinic depth
of strong weathering is about 10∼25 m, and aclinic Figure 2. Stereographic polar projection of dominant struc-
depth of weak weathered is 33 m. The original ter- tural planes and slope surface.
rain is small ridges at 0+00∼0+35 in Area A. Aclinic
depth of excavation is less than 15 m, so rock mass
of excavation plane is in the strong weathering and weathering, and rock mass structure is the sub-massive
offloading zone. Coupled with the impact of F5 faults, structure, where joint planes are partial opening with-
rock mass was broken into cataclastic structure, and out filling. Groundwater of slope develops, which is
most of structure planes are open and filled with sec- locally found in this slope, and it seeps along bedrock
ondary mud. The fourth grade slope of Area B, C is cracks.
also in strong weathering and unloading zone. Slope Main three groups of dominant structure planes
from the first to the third grade in Area B, C is in weak developed in this slope: (1) contact planes along

404
the carbonaceous shale, mudstone and sandstone can
develop into weak interlayers, 0.1 m to 1 m thick,
occurrence N60◦ E/NW ∠ 21◦ , which can form poten-
tial sliding surface of block; (2) Unloading cracks
along river, occurrence N20◦ W/NE ∠ 86◦ , can form
cutting plane of rear edge of block; (3) occurrence of
joints is N58◦ E/SE ∠ 56◦ , which is nearly vertical to
strike of slope and dips downstream. They can form
lateral edge of block. From Figure 2, the above three
groups of structure planes made up block in Area B,
and it has conditions of deformation and failure. It may
generate deformation along weak interlayers which is
dip out of slope.

3 DEFORMATION CHARACTERISTICS
AND MECHANISM ANALYSIS OF SLOPE

3.1 Deformation and failure characteristics of slope

Figure 1 shows that deformation of the slope occurs
mainly in Area A and B, and there are no obvious signs Figure 3. Relaxation of rock mass on the slope surface.
of deformation and failure in Area C. Successional and
big cracks form the upstream border of the block in mainly distribute in the excavation plane and gullies,
Area B. Gully forms the downstream border of it in the amount more than 10. Moreover, two main cracks
Area. Cracks of house at the back of the fifth grade exist in Area B. Extending length of them is more than
slope in Area I form the rear boundary edge, and the 30 m, and their opening width is 10 to 30 cm.
exposed carbonaceous shale layer at the sidewall of
lock chamber forms the bottom boundary. Confined
block generates deformation. Deformation of the slope 3.1.2 Relaxation of rock mass in the slope
mainly includes the following three aspects: Relaxation of rock mass shows that rock mass becomes
loose, so these original filled structure planes can fur-
3.1.1 A large number of cracks in the slope ther open, and local rock mass generates deformation
According to strike of cracks, they can be divided (Figure 3). Relaxation of rock mass can generally
into three groups. LF1 nearly is parallel to excavation attribute to two reasons: one type of relaxation is
plane, occurrence N30◦ to 50◦ W. It is a big cracks caused by rebounding and unloading function (Huang
which develops along the river, and its opening width Runqiu 2000). The relaxation shows that it is strong
2 cm to 10 cm. Unloading cracks control formation in the shallow part of slope, and gets gradually weak
of the group of cracks which are N20◦ W/NE ∠ 86◦ ; inwards slope so as to disappear. The other relax-
LF2 cuts excavation plane with a large angle, occur- ation is caused by deformation of slope. Potential
rence N50◦ to 70◦ E, while length of extension is not slip surface becomes relaxation boundary. Inside it,
long, and it has close relation with joints which is N58◦ relaxation phenomenon is not existence and rock mass
E/SE ∠ to 56◦ ; LF3 also cuts excavation plane with is integrity; Outside it, repeated unloading function
a large angle. As a whole it inclines upstream. The makes relaxation phenomenon of mantle rock mass
group of cracks has a large quantity and short exten- become more strong, and the most strong relaxation
sion. When block creeps along the layer, these cracks phenomenon of rock mass is close to potential shear
are formed. edge, and then local position of slope causes bulging
Cracks intensively distribute in two zones: below deformation, perhaps resulting in collapse. The slope
488 platform in Area A and above Altitude 494. Cracks below 488 platform in Area A has a smaller excava-
of Area A are mainly big LF1 at 488 platform and step tion depth. It is in a strong weathering and unloading
cracks which are formed by LF2 and LF3 below this zone. Open structure planes are filled with plenty of
platform. Its opening width is 5 to 10 cm. There are mud and debris. Hole can be found in local posi-
3 big cracks LF1 above Altitude 494. Their extending tion. Instability of local block can be found near the
length is 15∼30 m. They cross houses at the top of the road which leads to dam. This explains that relax-
slope, and their opening width is 5 to 10 cm. Extending ation of rock mass below 488 platform in Area A
length of LF2 and LF3 is smaller, about 0.5∼1 m. in the slope is mainly caused by repeated unloading
The opening width is less than 2 cm. These cracks function.

405
3.1.3 Interbedded shearing deformation along 3.2 Analysis of deformation and failure
the carbonaceous shale layer mode of slope
From Figures 4 and 5, when the first grade slope was
A number of weak interlayers and steep dip outwards
excavated in Area B, weak weathered sandstone con-
off-load cracks develop in the slope. The combination
taining thin silt mudstone was exposed, thickness of
of the two structure planes can form stepladder-like
which ranges from 30 to 50 cm; Furthermore, there is
sliding plane, as shown in Figure 6. It is extensive
black carbonaceous shale zone, thickness of which is
that the stepladder-like creep and crack deformation
about 1 m. Deformation is always along plane and car-
distributes in hard and thick strata, and the develop-
bonaceous shale seam, which moves towards outside
ment of this deformation is often from up to down.
of the slope and upstream. It shows that displace-
It creeps along structure planes dipping out of slope
ment deformation of rock mass developed towards
with middle-low angles to the dip direction of the slope
outside of the slope. Displacement distance upstream
(Huang Runqiu 2004, Yang Xuetang et al. 2004). The
is 3∼6 cm (dip direction), and displacement distance
deformation is gradually transferred by steep cracks.
towards outside of the slope is 5 cm. Meanwhile, fresh
The stepladder-like creep and crack develops from
tension cracks are found at the back of the second
down to up, namely: exposed mudstone and car-
grade berm. Its extension length is about 30 m, and
bonaceous shale in the first grade of the slope form
its width is 10∼20 cm. LF2 and LF3 constitute step
unloading rebound and compressive deformation, and
plane, which separates first grade slope from the upper
block consists of 3 dominant structure planes can form
rock mass. It shows that it is possible for the block to
creep deformation along weak interlayers dipping out
get instable along potential sliding surface, which is
of slope and strata dip direction. In addition fracture
formed by carbonaceous shale layer and cracks.
happens along off-load cracks upwards.
Clearly, deformation of the slope can be divided into
From Figure 6, it is clear that deformation of the
three levels: (a) width of cracks above Altitude 494
slope controlled by weak interlayers is the defor-
is smaller, and extending length is also short. Detri-
mation of block comprised by dominant structure
tus soil is exposed near houses. Cracks running along
planes. If supporting intensity in not enough, it can
slope have long length, which damage foundation of
develop upwards along stepladder-like sliding plane
houses, so that wall cracks; (2) below 488 platform in
comprised by weak layers and steep dip offload cracks.
Area A, the slope has strong unloading and weathering
Under certain conditions, Area A and B may form
rock mass. Deformation shows that a larger longitudi-
stepladder-like creep and crack, resulting in massive
nal crake is formed at the 488 platform, and relaxation
landslide.
of rock mass is in front of the slope. Local instability
In Area C slope was curved. Combination of struc-
can be found; (3) Excavation depth of the first grade
ture planes can not form potential instable block
slope in Area B is about 60 m rock mass is weak weath-
controlling stability of slope, but higher step and
ering, and it forms creep sliding deformation along
steeper slope may form local falling. So overall sta-
carbonaceous shale intercalation. Under certain con-
bility of slope in Area C is better, but protection of
ditions the three levels of deformation above can cause
slope surface has to be done.
overall instability of slope.

4 STUDY ON TREATMENT

Deformation of the slope is mainly caused by surfi-

cial excavation in Area A. In certain range of depth,
the excavated surface is filled with cataclastic rock in
which joints are extremely developed, and inner weak
interlayers of the slope are also developed. Combina-
tion of structure planes is harmful for the stability of
the slope. It can easily make deformation of block,
resulting in a large-scale instability of slope. There-
fore the emphasis of slope treatment should be put on
following aspects:
1. A majority of rock mass in Area A is located in the
strong weathering and off-load zone, and excavated
depth is not enough to form relative protuberant
terrain between 0+00 and 0+35. 488 platform is
20 m wide, and the slope has a very good con-
Figure 4. Slipping along carbonaceous shale. dition to be cut below it. The slope above road

406
Figure 5. Geological sketch map of the first slope in Zone B.

3. For slope of the Area A and B, the support project

of combination of the pre-stressed anchor cable and
framework beams is chosen. Length of anchor cable
is 20 m to 25 m. After cutting and offloading slope
above the third step, there is still a certain depth
of rock mass which is strong weathering and off-
load. Direction of anchor should be perpendicular
to slope. The first and the second step are weak
weathered rock mass. So preventing rock mass
from creeping along the weak layers is key task.
It is why direction of anchor wire should be per-
pendicular to the slope. Anchor framework is used
to slope protection in Area C and length of anchor
Figure 6. Distribution sketch map of deformation and ranges from 8 m∼10 m.
controlling measurements of 0+35 profile. 4. Rock mass which is incised by structural planes
is fragmentation. The groundwater of slope is also
developed. So a row of acclivitous drainage holes
leading to dam in Area A, is cut by ratio of 1:1, of 15 m deep are set at the bottom.
and it’s beneficial to improving the stress condi-
tion of the slope. But after cutting, there still exist
some parts of strong weathering and off-load rock 5 CONCLUSIONS
mass. Height of the step is 15 m and a lot of weak
interlayers of slope are well developed, so cutting 1. The oblique slope is an inclined slope which con-
and offloading can’t satisfy the request of keeping sists of interbedded mudstone, siltstone, carbona-
stability of the slope in a long run. ceous shale. And interbedded displacement zone is
2. The slope has many weak interlayers, and all of also well-developed. The deformation of slope is
them can form bottom sliding plane of creep and showed with offloading rebound and compressive
crack deformation. Through the deformation and deformation of soft rock. Blocks transfer defor-
failure model of slope, it can be concluded that mation towards top, which are formed by weak
holistic deformation of the slope may be caused by layers and off-load cracks. The shallow surficial
creep deformation of carbonaceous shale interlayer. rock mass of the slope is quite crashed, and joints
And along combination of steep dip off-loading are well developed. The slope has a large number
cracks and weak layers, it may gradually develop of cracks, of which strikes are in the same direction
upwards. Because above the tear line, slope is very with these dominant structural planes.
tall and the weak layers are extremely developed, 2. Combination of structural planes forms a large
after cutting, there is the possibility of large-scale amount of potential instable blocks. Deformation
creep deformation in Area A. So the emphasis and failure model of slope is stepladder-like creep
of controlling slope deformation is to prevent the and crack, which develops from bottom to top.
deformation of weak layers. Controlling of the slope deformation should be set

407
 on cutting and offloading, cleaning out cataclastic Huang Runqiu, Lin, Feng & Chen Deji. 2001. Formation
rock in Area A, and in order to control the defor- mechanism of unloading fracture zone of high slopes and
mation of weak layers, anchor measures should be its engineering behaviors. Journal of engineering geology
taken. 9(3):227–232.
3. Because of spatial difference of rock structure, Huang Zhengjia & Wu Aiqing, 2001. Usage of Block The-
ory in Three Gorges Projects. Chinese journal of rock
deformation of slope in different parts is usually mechanics and engineering 20(5):648–652.
different, therefore, study of deformation and fail- Huang Runqiu. 2004. Mechanism of large scale landslides in
ure model of slope should analyze the mutual western China. Advances in earth science 19(3):443–450.
relationship of rock structure and deformation in Yang Xuetang, Ha Qiuling, Gao Xizhang. 2004. Research
details. on unloading deformation and support of high slope
rock interlaced with hard rock. Chinese journal of rock
mechanics and engineering, 23(16):2681–2686.
REFERENCES Sun Hongyue & Shang Yuequan. 1999. Study on deformation
and failure characteristics of inclined slope. Journal of
engineering geology 9, 7(2):141–146.
Ding Xiuli, Fu Jing & Liu Jian. 2005. Study on creep behavior
Yang Yonghong & Lv Dawei. 2006. Study on treatment of
of alternatively distributed soft and hard rock layers and
high-cut carbonaceous shale slope in expressway. Chi-
slope stability analysis. Chinese journal of rock mechanics
nese journal of rock mechanics and engineering, 25
and engineering. 24(19):3410–3418.
(2):392–398.
Fan Wen, Yu Maohong & Li Tonglu. 2000. Failure pattern
Yin Yueping. 2005. Human-cutting slope structure and fail-
and numerical simulation of landslide stability of starti-
ure pattern at the three gorges reservoir. Journal of
fied rock. Chinese journal of rock mechanics and eng.
engineering geology 13(2):145–154.
19(supplement):983–986.
Huang Runqiu. 2000. Time-dependent deformation of a high
rock slope and its engineering-geological significance.
Journal of engineering geology. 8(2):148–153.

408
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Sliding in weathered banded gneiss due to gullying in southern Brazil

W.A. Lacerda
COPPE/UFRJ, Rio de Janeiro, Brazil

A.P. Fonseca
CEFET-RJ, Rio de Janeiro, Brazil

A.L. Coelho Netto

IGEO/UFRJ, Rio de Janeiro, Brazil

ABSTRACT: This geotechnical study of a case of sliding associated with gullying is of great importance
for understanding the various mechanisms of this type that are frequently found in the Middle Paraiba do Sul
Valley in the southeast of Brazil. The evolution of erosive canals has been widely studied in this region (Coelho
Netto, 2003). Once an erosive canal reaches a drainage basin with a steeper topography, it can trigger landslides.
These landslides frequently intensify erosion (the soil is left more exposed to surface drainage) and significantly
increases the volume of soil carried to rivers. In this paper landslides caused by gullying in Três Barras were
studied. In this concavity two independent mechanisms for the movement of masses exist: the erosion of the
more friable material (recent residual material) and the sliding of overlying soil masses.

1 INTRODUCTION

In the area of study it was noted that the propaga-

tion of gullying was associated with the expansion
of the canal network, which can rupture and force Y
back topographical divides. The advance of the gully
undermines the stability of the two masses with dif-
ferent slide directions, one along the erosion axis and
the other approximately perpendicular, which is dislo- X
cated as if it were a plate. These masses are called A
and B respectively, and can be seen in Figure 1.
The two unstable masses studied (A and B) were
measured with inclinometers and piezometers.
Figure 1. Photo of gullying in study. Mass B moves along
a planar surface, with zero movement at point X and
maximum movement at point Y. Mass A moves along a
2 AREA LOCATION conchoidal surface.

The area of study is the Três Barras concavity, located

in the southeastern plateau of Brazil, on the northern
watershed of Serra da Bocaina, in the middle Paraiba
do Sul valley, close to the frontier between the states 3 GEOTECHNICAL PROFILE OF THE AREA
of Sã Paul an Ri de Janeiro.
In the area of gullying there was a vein of pegmatite 14 holes were drill using percussion drilling and
at the latitude N 55◦ E/35◦ NW, in other words parallel 3 using mixed drilling techniques in the area of study.
to the strike of the Pre-Cambrian rock foliation. Piezometers and inclinometers were placed in these
The actions of a temporary aquifer in contact with holes. Figures 2, 3 and 4 show the geotechnical
the altered rock on top of the pegmatite vein were sections. The topographic plan with the instrumenta-
also observed. This aquifer is solely supplied by local tion points and the section sketches are shown in
rainfall. Figure 6. The subsoil essentially consists of layers of

409
Figure 2. Geotechnical section SP6-SP12 of the area of
study.

Figure 5. Topographical plan with the location of the

drilling, the instrumentation and the indication of the subsoil
sections (elevations in meters).

Figure 3. Geotechnical section SP6-SP3 of the area of

study.

Figure 6. Average piezometric elevations.

4.1 Piezometers
Figure 6 shows the Piezometric elevations for all the
piezometers.
Piezometers 2a, 2b, 6, 7 and 13 remained continu-
ally dry. Some instruments remained dry for sus-
tained periods of time, which is represented by the
Figure 4. Geotechnical section SM3-SP13 of the area of discontinuities in the chart.
study. The piezometric level of Pz1 declined continually,
as can be seen in Figure 5. In relation to this, it should
be noted that the nearest piezometers, Pz2 a and b,
sandy clay in a process of laterization over the residual remained continually dry and that Pz1 is at the top
sandy-silty gneissic soil. of the slope, on a elevation 15 meters above Pz2.
Therefore, it can be assumed that the instrument in
question had some sort of defect. As a result Pz1 was
4 INSTRUMENT READINGS not considered in the analysis.
It can also be noted that the piezometers that con-
14 Casagrande type piezometers and 3 inclinometers tained water during the monitoring period are located
were set up in the area. The location of these instru- in the direction of the erosion axis, favored by the
ments is shown in Figure 5. It should be noted that probable geological fractures in the area of study. Ave-
piezometers (Pz) were placed in the holes dug by per- lar and Coelho Netto (1992) obtained similar data
cussion drilling (SP) and the inclinometers (I) in the when monitoring the Bom Jardim Concavity, also in
holes dug by mixed drilling (SM). Bananal.

410
 Figure 7 shows the accumulated rain over a 25 day and pz14) did not present any relation with the rainfall
period measured in the Bananal pluviometric station that occurred, though pz10 and 14, which are located
(SIGRH/SP). at the bottom of the canal, along the erosion axis,
The variation in the piezometric level of instruments presented artesianism.
Pz3a and b, Pz4, Pz5 and Pz9 bore some relation to the
rainfall measured. Looking at Figures 7 and 8, it can be
noted that elevations in the piezometric levels occurred 4.2 Inclinometers
approximately two months after the respective accu- Table 1 shows the depths at which the inclinometers
mulated rainfall over a 25 day period measured in the were installed and the date of the initial reading taken
Bananal station. two days after installation.
This variation in the piezometric levels following Figures 8, 9 and 10 show the inclinometer readings.
a long delay in relation to the measured rainfall is Figure 11 shows the topographic plan with the vector
common in the region. Rocha Leão (2005) found an direction of the displacements measured at the depth of
interval of time similar to those found in this paper the rupture surface. Figure 12 shows the geotechnical
when he compared the piezometric levels and the profiles in the direction of movement of each incli-
local rainfall in an area of gullying in Bela Vista, nometer with the maximum displacement measured
approximately 7 km from the area of study. (resultant displacement).
It can also be seen that the piezometric levels in Table 2 shows the maximum horizontal displace-
instruments placed in the lowest elevations (pz8, pz10 ments measured and the depth of the rupture surface.
It is evident that part of the rupture surface is prac-
tically situated on the contact between the soil and
the rock.

Table 1. Depth of installation and date of initial reading of

inclinometers.

Date of initial Installation

Inclinometer reading depth (m)

I 01 05/05/2004 15.0
I 02 01/05/2004 24.5
I 03 29/04/2004 21.0
Figure 7. Accumulated rainfall over a 25 day period.

Figure 8. Readings of inclinometer I1. Figure 9. Readings of inclinometer I2.

411
Figure 10. Readings of inclinometer I3.

Figure 12. Geotechnical profiles in the direction of move-

ment of each inclinometer with the maximum displacement
measured (resulting) and the indication of the probable
surface slide.

Figure 11. Displacement vectors measured at rupture

surface. Table 2. Maximum horizontal displacement and the depth
of the rupture surface between June 2004 and October 2005.

Maximum
5 INSTABILITY MECHANISMS Inclinometer displacement (mm) Depth (m)

I 01 7.1 14.0
A hypothesis that can explain the phenomenon is that I 02 13.2 21.5
soil mass A becomes periodically unstable due to the I 03 17.7 20.0
gradual erosion that occurs around its base. When this
takes place mass A moves, which in turn also permits
the movement of mass B.
Stability analyses have shown that the shear strength
along the slide surface can be classified as residual. 6 CONCLUSIONS
Skempton (1964), Lupini et al. (1981) and Leroueil
(2001) have shown the importance of studying the It was noted that the underground water in the area is
residual condition in the case of slides that may have fed by the local regional water table and not by local
already occurred. rain. Some points of artesianism were measured in

412
the area, all of which were at the bottom of the ero- Avelar, A.S. & Coelho Netto, A.L. (1992a) ‘‘Fraturas e desen-
sive channel. The flow of water has been eroding the volvimento de unidades côncavas no médio vale do rio
saprolitic soil, loosening the underlying levels which Paraíba do Sul’’. Rev. Bras. De Geociências, vol. 22
then rupture and slide. In this way a continuous pro- no. 2.
cess of erosion and sliding was started throughout the Avelar, A.S. & Coelho Netto, A.L. (1992b) Fluxos d’água
subsuperficiais associados a origem das formas côncavas
entire area. do relevo. Anais da 1a conferência Brasileira de Estabil-
In the large majority of cases the development idade de Encostas/COBRAE, ABMS and SBGE, Rio de
of concave units is associated with rock fractur- Janeiro; vol. 2: 709–719.
ing, which, according to Avelar and Coelho Netto Coelho Netto, A.L. (2003) ‘‘Evolução de cabeceira de
(1992), allows the exfiltration of ascendant under- drenagem no Médio Vale do Rio Paraíba do Sul (SP/RJ):
ground water flows (artesianism). Due to the excess A formação e o crescimento da rede de canais sob controle
pore-pressure on the exfiltration rock faces located estrutural’’. Revista Brasileira de Geomorfologia, vol. 4,
on the gully walls, these flows can unleash erosive no. 2, pp. 118–167.
mechanisms. Fonseca, A.P. (2006) Analise de mecanismos de voçoro-
camento associados a escorregamentos na Bacia do rio
It was also noted through the inclinometer readings Bananal (SP/RJ). Doctoral Dissertation, COPPE/UFRJ.
and the field observations that the shearing surface of Heilbron, M. (1995) O Segmento Central da Faixa Ribeira:
mass A is conchoidal while that of mass B is planar, síntese geológica e ensaio de evolução geotectônico. Livre
fan-shaped in plan. Docência. DGeoUERJ, 110p.
Leroueil, S. (2001) ‘‘Natural slopes and cuts: movement and
failure mechanisms’’. Géotechnique, vol. 51, no. 3, pp.
ACKNOWLEDGEMENTS 197–243.
Lupini, J.F., Skiner, A.E. & Vaughan, P.R. (1981) ‘‘The
The authors would like to register their gratitude to drained residual strength of cohesive soils’’. Géotech-
nique, 31(2), pp. 181–213.
CNPq and FAPERJ, through the PRONEX Program, Rocha Leão, O.M. (2005) Evolução regressiva da rede de
for the financial support they provided for this research canais por fluxos de água subterrânea em cabeceiras
project and to MSc students Raquel Maciel, Vitor de drenagem: bases geo-hidroecológicas para recuperação
Aguiar and Marina Duarte for support during the de áreas degradadas com controle de erosão. Doctoral
field work. Thesis, IGEO-UFRJ, Rio de Janeiro.
SIGRH/SP—Homepage: http: www.sigrh.sp.gov.br/cgi-bin/
bdhm.exe/plu
REFERENCES Skempton, A.W. (1964) ‘‘Long-Term stability of clay slopes.’’
Fourth Rankine Lecture, Géotechnique, vol. 14, no. 2
pp. 77–101.
Almeida, J.C.H., Eirado Silva, L.G. & Avelar, A.S. (1991)
‘‘Coluna tectono-estratigráfica da parte do Complexo
Paraíba do Sul na região de Bananal, SP’’ Anais do Simp.
Geol. Do Sudeste/SBG, São Paulo.

413
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Experimental and three-dimensional numerical investigations of the impact

of dry granular flow on a barrier

R.P.H. Law, G.D. Zhou, C.W.W. Ng & W.H. Tang

Department of Civil Engineering, Hong Kong University of Science and Technology, HKSAR

ABSTRACT: In this study, the impact behaviour of dry granular flows on a barrier was investigated both
experimentally and numerically. Physical model tests were carried out using a 3.8 m long flume whereas numerical
simulations were conducted using a three-dimensional particulate code (PFC3D ). Uniform clean sands (Leighton
Buzzard sands fraction C and E) were used for the flume tests. During each test, impact profiles were recorded
using a high speed camera and impact force was measured by recording the deformation of a barrier with known
stiffness. Results obtained from experimental and numerical tests are compared and analyzed.

1 INTRODUCTION According to Wendeler (2006), flexible barriers

were originally designed to protect villages, highways
Many landslides with long travel distance are flow- and railway lines from rockfalls. Their main loadbear-
like in character. Some, such as debris flows, have ing principle is to restrain the falling rocks using a
distributed velocity profiles resembling the flow of long braking distance and therefore producing a soft
fluids (Hungr 1995). Debris flow poses a costly stop, reducing the peak loads in the barrier compo-
risk for lives and properties in mountainous and nents and the anchors. The same principle also works
rainy areas. Understanding of the flow and impact for a variety of other problems such as snow slides,
characteristics of debris flow would facilitate deci- tree falls, floating woody debris during flooding and
sion and implementation of debris flow mitigation debris flows. According to Rorem (2004), several nat-
strategy. ural debris flow events in the USA and around the
In previous work by Law et al (2007), a series of world have demonstrated that flexible barriers can be
flume model tests and numerical simulations were con- effective at stopping debris flows. However, as far as
ducted to investigate the flow characteristics of dry the authors are aware, impact behaviour of debris on a
granular material. Pore fluid was not considered in the barrier is not fully understood yet.
tests to simplify the experiments and numerical anal- This study therefore aims at investigating the impact
ysis. Physical model test results revealed that small characteristics of dry granular flow through flume
plugs were observed in all dry flow tests. Plugs did model tests (see Figure 1) and three-dimensional
not grow in size significantly during the flow pro- numerical simulations by PFC3D . Dry Leighton Buz-
cess. Reverse segregation was clearly observed in all zard sands (LB sands) fraction C and E (see Figure 2)
flow tests using non-uniform silty sand (completely were used for the experimental tests. Granular flow
decomposed granite). Coarse particles were observed simulations started with the release of dry granular
at the front of the flow, whereas fine particles were materials from the straight flume. Impact process was
located at the rear and at the bottom of the flow. observed through high speed image and the force
Reverse segregation was also simulated in 3D numeri- exerted on the barrier was recorded. Emphasis was
cal codes (PFC3D ). In the previous publication, impact made to study the effect of grain size on the mea-
characteristics of granular flow on a barrier are not sured force. Numerical simulation of dry granular
studied. flow was conducted by using Particle Flow Code in
In the field, a flexible barrier may be composed of Three Dimensions (PFC3D ) (Itasca 2005). Elastic balls
rings, ropes, nets, springs and dashpots. Anchors may were used to model granular materials and the barrier.
be installed to fix the barrier in position. Anchorage Emphasis was made on the influence of static load of
may break under heavy impact load, leading to the deposited material on impact force. Computed results
failure of the barrier. Impact force analysis is therefore are compared and analyzed with experimental results
important for barrier design. in this paper.

415
 by laser sensor. The measured force is calculated by
Holding area
Inner volume: 0.1m3
the measured deformations of the springs times the
stiffness of the springs. The installation of springs
0.5m
serves to provide flexibility of the aluminium plate
0.4m
upon impact. The shafts and the springs are supported
0.6 m by an aluminium structure. The structure restricts the
Barrier with
adjustable movement of the shaft such that the aluminium plate
stiffness could move parallel to the channel only. The barrier
Channel was calibrated by static load test, in which loads were
Length: 3.8m added successively onto the barrier and the displace-
ment was recorded. The calibrated stiffness of the
40o springs was 29.3 N/mm. Materials to be tested were
placed in a holding area at the highest point of the
slope. The inner volume of the holding area is 0.1 m3 .
Figure 1. The experimental setup for study of granular flow.
The prepared dry granular materials were placed in
the holding area just before tests started to minimize
possible consolidation and segregation.
Two impact tests were conducted. In each test, 30 kg
of dry LB sands (either fraction C or E) was released
on the channel. LB sand (fraction C) composed of
fairly uniform grains with grain diameter ranging from
300 μm to 600 μm whereas LB sand (fraction E)
composed of fairly uniform grains with grain dia-
meter ranging from 150 μm to 90 μm. Figure 2
shows the particle size distribution of the LB sands
(fraction C and E) used. The two tests aim to examine
the effect of grain size on measured impact forces.

3 INTERPRETATIONS OF MODEL TEST

Figure 2. Particle size distributions of dry LB sands (frac- RESULTS
tion C and E).
3.1 Effects of grain size on measured impact force
Figure 3 shows the observed time history of mea-
2 DETAILS OF PHYSICAL MODEL TESTS sured impact forces by LB sands (fraction C and E).
The measured forces were deduced by using the aver-
Figure 1 shows the setup of flume model. Two tests age calibration constants obtained from the best fit
were carried out using the flume of 3.8 m long, 2.8 m
high, 0.4 m wide and 0.6 m in depth. The slope angle
was adjusted to be 40◦ . The base of the slope was
smoothened with polythene to ensure uniform rough-
ness along the slope. The flume was mainly made of
plywood, except that one side of channel was made
of transparent Perspex. The channel was held in posi-
tion by a heavy duty crane system. A MotionScope®
PCI Series high speed imaging system was installed to
view the impact behaviour. The recording speed was
125 frames per second and the resolution was 480 ×
420 pixels.
For modelling a barrier, an aluminium plate was
installed to stop the flow at the lower end of the chan-
nel. The plate was 0.5 m in depth, 0.4 m in width
and 0.01 m in thickness. The aluminium plate was
supported by a set of shafts and springs of known
stiffness. Load was transferred to the springs during an Figure 3. The time history of measured impact forces using
impact. The deformations of the springs were recorded LB sands (fraction C and E).

416
line. The moment at which the fraction C sand first 3.2 The influence of deposition process
impacted the barrier is taken as time zero. As shown on measured force
in the figure, the measured force rises continuously
Figure 4 shows the images recorded during the depo-
in two stages with a similar rate of increase until the
sition processes in the two tests. Each dotted line
time reaches approximately 0.35 s for the test using
represents the surface of deposited granular material
fraction C sand. After 0.35 s, the measured force does
at each given time interval. The time increment (t)
not change noticeably as time elapses. The measured
between each dotted line is 0.08 s. Since the shutter
maximum force is 165 N. Regarding the test using
speed of the camera is 0.008 s, distances between two
fraction E sand, the measured force rises continuously
dotted lines therefore represent the observed surfaces
in multi stages with different rates of increase until
obtained between every ten camera shots. In each of
the time reaches approximately 0.95 s. After 0.95 s,
the Figures 4a and 4b, the rightmost dotted line is the
the measured force does not change significantly as
deposited surface 0.08 s after the granular material
time elapses. The measured maximum force by the
first impacted on the barrier in each test. It can be
static load of deposit is 171 N for fraction E sand. It
seen in both figures that that the horizontal distribu-
is observed that no peak force was measured for both
tion between every two dotted lines is fairly constant
tests using fraction C and E sands. The duration of
and uniform, thus suggesting a progressive accumula-
impact was much longer for fraction E sand than that
tion of the deposit of the granular material. Progressive
for fraction C sand.
accumulation of the deposit involves a relatively longer
According to the research work done by Shinohara
(2000), the angle of internal friction of dry powders
increases with a smaller porosity. As the mean grain
size of uniform fraction C sand is larger than that of
fraction E sand, it is expected that fraction E sand
should have a smaller porosity and hence a higher
dynamic angle of internal friction to resist to flow.
Thus, it may be reasonable to deduce that fraction E
should flow slower because of its larger internal fric-
tional resistance and so it take a longer duration to
complete the impact process. The time history of the
measured forces suggests that the duration of impact
decreases with an increase in grain size of uniform
granular materials.
In Figure 3, the measured forces are compared with
a predicted static force. The predicted force represents
the total static load exerted on the barrier by deposited
granular material. The predicted force is calculated by
the following simple equation:

F = mg sin θc sin θi (1)

where F is the predicted impact force, m is mass of the

granular material, g is the gravitational acceleration,
θc is an angle of the slope, θi is an angle between the
barrier and the channel. Based on the set up of the
flume, m, θc and θi are taken to be 30 kg, 40◦ and
90◦ respectively. It should be noted that the rate of
change of momentum of the granular material and the
basal frictional resistance of the channel are not con-
sidered in the equation. The predicted force therefore
does not vary with time. The dashed line represents the
predicted static force by Equation 1 and the predicted
value is 190 N, which is greater than the measured
static load acting on the barrier in both tests. It could
be explained by the reduction of the measured static
load by the basal frictional resistance, which is not Figure 4. Recorded images of deposited surfaces in tests
considered in the equation. using a) LB sand (fraction C) and b) LB sand (fraction E).

417
impact process than impact which is caused by rigid
object, such as a boulder impact. A longer impact pro-
cess reduces the rate of change of momentum of the
granular material, leading to smaller dynamic impact.
The smaller dynamic impact force helps to explain
why no peaks of the measured force were observed.
According to Wendeler (2006), the debris deposit
on a ring and net barrier (one form of flexible barrier)
serves as a dam. The dam is strong enough to retain
the remaining debris. In the process of the progressive
accumulation of the deposit, the deposit is expected
to act like a dam to reduce the dynamic impact of the
upcoming granular material. As a result, the existence
of accumulated deposit during the impact process is
expected to reduce the impact load.
Figure 5. A typical setup of numerical model.

4 NUMERICAL MODELING OF THE IMPACTS

OF GRANULAR MATERIAL ON A BARRIER
spring matrix. The aluminium plate of the barrier
Particle Flow Code in Three Dimensions (PFC3D ) is modelled by the plate matrix. The plate matrix
(Itasca 2005) is a numerical simulation tool that mod- composes of 2400 spheres. Each of the spheres is
els the motion and interaction of spherical particles 0.005 m in radius and 106 N/m in stiffness. Based
by the distinct element method (DEM) (Cundall & on the density of the aluminium plate used in exper-
Strack 1979). In this study, it is used to study the imental tests, the density of the spheres is set to be
impact behaviour of dry granular materials on a bar- 2700 kg/m3 . The springs of the barrier is modelled
rier. It should be pointed out that the main purpose of by the spring matrix. The plate matrix composes of
the numerical simulations is to investigate the impact 48 spheres. Each of the spheres is 0.05 m in radius.
behaviour qualitatively, not quantitatively due to the Based on the barrier stiffness used in experimental
complexity of dry granular flows, uncertainty in model tests, the stiffness of the simulated barrier is set to be
parameters and limitations of the numerical code such 29.3 N/mm. Correspondingly, the normal stiffness of
as handling the presence of pore fluid. the spheres that compose the spring matrix is set to
Figure 5 shows the setup of the numerical model. be 3960 N/m. Since a ball with zero density is not
The two side walls and the bottom floor of the flume (as allowed in the code, the density of spheres that com-
shown in Figure 1) are idealized as elastic ‘‘walls’’ with pose the spring matrix is set to be 100 kg/m3 . Based
normal and tangential (shear) stiffness of 108 N/m. on the barrier design in experimental tests, each of
For simplicity and reduction of computational time, the spheres in the plate matrix and the spring matrix
LB sand is modelled by a cluster of 21 mm-diameter could move parallel to the channel only. The local
identical spheres. Obviously, the diameter of each damping ratio of the barrier is roughly taken as 0.7.
ball is much larger than that of soil grains used in All spheres that constitute the barrier are frictionless.
the flume tests. Hence, computed results should be The normal and shearing bonding in the barrier is set
treated with caution. In order to reduce the computa- to be 107 N/m. The arrangement of the spheres in
tion time, stiffness of the spheres is limited. Therefore, the plate matrix and the spring matrix is shown in
the stiffness of the spheres is set to be 106 N/m. The Figure 5.
density of spheres is chosen as 1480 kg/m3 , which During impact, the spring matrix is compressed by
is the measured bulk density of fraction C sand used approaching granular material. As the stiffness of the
in laboratory tests. According to Brown (2000), the plate matrix is 250 times higher than the spring matrix,
internal friction angle of LB sand fraction C is 35.4◦ . the compression of the barrier is mostly contributed
The internal friction angle of the granular material by the spring matrix. The aluminium structure in the
is therefore set to be 35.4◦ . The interface friction barrier is modelled by a rigid wall. A rigid wall is
angle between the granular material and the channel installed at the back of the barrier to support the bar-
are set to be 22.5◦ . The value was obtained by test- rier and record the impact force. The computed impact
ing a hollow paper cylinder sliding down the channel. force is calculated by the force acting on the rigid
The hollow paper cylinder was filled with LB sand wall minus the static load by the weight of the bar-
fraction C. rier. Table 1 summarizes the parameters used for the
Two matrices of spherical particles were used to numerical analysis. Details of the contact behaviour of
simulate the barrier, namely the plate matrix and the the model could be found in Law et al. (2007).

418
Table 1. Parameters adopted for numerical simulations.

Parameter Magnitude

Slope angle 40◦

Length of the channel 3.8 m
Stiffness of walls Normal: 108 N/m
Tangential: 108 N/m
Number of balls Granular material: 4000
Plate matrix: 2400
Spring matrix: 48
Ball radius Granular material: 0.01 N/m
Plate matrix: 0.05 m
Spring matrix: 0.03 m
Density of each ball Granular material: 1450 kg/m3
Plate matrix: 2700 kg/m3
Spring matrix: 100 kg/m3
Figure 6. Comparisons of measured and computed time
Mass of granular material 30 kg
history of impact forces.
Normal and tangential ball Granular material: 106 N/m
stiffness Plate matrix: 106 N/m
Spring matrix: 3960 N/m
Internal friction angle 35.4◦ mass of balls from 1480 kg/m3 to 100 kg/m3 when-
Interface friction angle 22.5◦ ever they rest on the barrier. This is represented by the
Contact bond Normal: 107 N/m dashed line in Figure 6. Thus, the difference in mag-
Shear: 107 N/m nitude between the dotted and the dashed lines is the
Local damping Granular material: 0.05
contribution of static load of the granular material. It
Barrier: 0.7
Viscous normal damping Granular material: 0.2 can be observed that the computed and the reduced
Barrier: 0 force, in general, rise in magnitude to 101 N at about
Viscous shear damping Granular material: 0.2 0.35 s. While the computed force continues to rise to
Barrier: 0 a maximum of 198 N at the elapsed time is 0.9 s, the
reduced force, however, falls to 15 N at time equal to
1.05 s. This ‘‘residual’’ 15 N force is due to the static
load of the deposited granular material specified with
the reduced density of 100 kg/m3 in the PFC analysis.
The ratio between the maximum computed reduced
5 THE ROLE OF STATIC LOAD ON IMPACT force of 101 N and the maximum residual reduced
FORCE force (i.e., 15 N) is around 15%. Thus, the static effect
of the granular material is considered to be partially
Figure 6 compares the measured and computed time eliminated from the reduced force.
histories of impact forces acting on the barrier. The By comparing the difference in magnitude between
solid line represents the measured force in the test the dotted and the dashed lines, the contribution of
using fraction C sands. The time zero represents the static load is observed to increase during the impact
moment at which the granular material first hit the bar- process. The computed force at the end of impact is pri-
rier, whereas the dotted line represents the computed marily composed of static load. It is explained by the
force time history. It can be seen that the computed accumulation of deposited granular material during
time history is generally consistent with the measured the impact process. On the other hand, it is observed
one, although the duration of computed impact (about that the maximum measured and computed forces were
0.9 s) is almost twice as the measured value (about recorded at the end of impact. The results suggest
0.35 s). In addition, it takes numerous stages in the that the maximum force for dry granular material is
numerical prediction, rather than two stages in the substantially contributed by static load.
experiment, to reach the maximum computed impact
force of 198 N, which is about 17% larger than the mea-
sured value of 171 N. Considering the approximations
involved and simplifications made in the numerical 6 CONCLUSIONS
simulations and the uncertainties encountered during
the experiment, the fairly consistent results obtained In order to investigate the impact characteristics of
are quite fortuitous. dry granular materials, flume model tests and numer-
In order to understand the components of an impact ical simulations were conducted. Physical model test
force, a reduced force is calculated by reducing the results reveal that the duration of impact was longer

419
for fraction E sand (smaller particles) than that of Chau, K.T., Wong, R.H.C. & Wu, J.J. 2002. Coefficient
fraction C sand. It may be explained by the differ- of restitution and rotational motions of rockfall impacts.
ence in grain size between the fraction C and E sands. International Journal of Rock Mechanics & Mining Sci-
No peak force was measured for both fraction C and ence 39: 69–77.
E sands. Progressive deposition was observed in tests Cundall, P.A. & Strack, O.D.L. 1979. A Discrete Numeri-
cal Model for Granular Assemblies. Geotechnique, 29, 1:
using fraction C and E sands. 47–65.
Numerical simulations indicate that the contribu- Hungr, O. 2000. A model for the runout analysis of rapid
tion of static load increases during the impact process. flow slides, debris flows, and avalanches. Canadian
Static load is illustrated and verified to contribute the Geotechnical Journal 32: 610–623.
maximum impact force substantially. Itasca, 2005. Itasca Consulting Group Inc. PFC3D (Particle
Flow Code in 3 Dimensions), Version 3.1. Minneapolis:
ICG.
ACKNOWLEDGEMENTS Law, R.P.H., Zhou, G.D., Chan, Y.M. & Ng, C.W.W. 2007.
Investigations of fundamental mechanisms of dry granular
debris flow. Proc. 16th Southeast Asia Geot. Conf. 8–11
The authors would like to acknowledge the financial May, Malaysia: 781–786.
support from the Research Grants Council of HKSAR Rorem, J. 2004. Debris Flow Remediation: Case study San
(grant no. HKUST6294/04E) and research grants Bernardino Mountains near Crestline and Lake Arrow-
DAG05/06.EG39 and CA-MG07/08.EG01 provided head, CA/USA. Geobrugg North America, LLC.
by Hong Kong University of Science and Technology Shinohara, K., Golman, B. & Oida, M. 2000. Effect of
(HKUST) and GEL05/06.EG01 by Geotech Engineer- particle shape on angle of internal friction by triaxial
ing Ltd. The assistance and contribution by Messrs compression test. Powder Technology 107: 131–136.
T.T.L. Yeung, Y. Kwok & T.H.N. Ho and technicians Wendeler, C., McArdell, B.W., Rickenmann, D., Volk-
at the Geotechnical Centrifuge Facility, HKUST are wein, A., Roth, A. & Denk, M. 2006. Field testing and
numerical modeling of flexible debris flow barriers. Proc.
also gratefully acknowledged. 6th International Conference on Physical Modelling in
Geotechnics, Hong Kong. Vol. 2: 1573–1578. ISBN:
0-0415-41586-1.
REFERENCES Wu, S.S. 1985. Rockfall evaluation by computer simulation.
Transportation Research Record., volume: issue: 1031:
Brown, C.J., Lahlouh, E.H. & Rotter, J.M. 2000. Experi- 1–5.
ments on a square planform silo. Chem. Eng. Sci. 55(20):
4399–4413.

420
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Temporal survey of fluids by 2D electrical tomography: The ‘‘Vence’’

landslide observatory site (Alpes-Maritimes, SE France)

T. Lebourg, S. El Bedoui, M. Hernandez & H. Jomard

UMR Géosciences Azur, CNRS-UNSA-IRD-UPMC, Sophia-Antipolis, France

ABSTRACT: A best knowledge about landslide processes requires the characterisation of the triggered factors
and their impact on the process cinematic. These factors are often time dependent and that a reason why it is
very complex to have a quantitative approach. We propose a temporal imagery of water circulation in a moving
mass by Electrical Resistivity Tomography (ERT). The purpose is to be able to quantify the coupling between
the rainfalls and the water inflows in a sliding mass, and the evolution of resistivity values acquired by ERT. Our
work was based on a multi-scale and multi-survey device approach on the ‘‘Vence’’ landslide (South-eastern
France). It is considered like a sandy mass sliding on a sedimentary rock substratum (limestone and calcareous)
and largely controlled by rainfall.

1 INTRODUCTION

In most of failure landslide processes fluids are con-

sidered as the most important triggered factor. In order
to constrain the risk associated to the landslide, it
appears necessary to localise and understand the flu-
ids draining network inside the soil mass. Since ten
years, the geophysical methods were developed and
applied to the study of instabilities of ground. These
methods are characterized by their non-destructive
aspect and the important investigation depth. They
allow obtaining some quantitative data about the slip
surfaces, the faults networks and the water flows in
depth.
Indeed, combination of fractures, breccia, gouge
and the fluids can produce large contrasts in observ-
able geophysical properties (Lebourg et al., 2003). The Figure 1. Localisation of the field study.
electrical resistivity is one of them and it is particularly
sensitive to the water content of investigated areas.
Precedent authors (Eberhart-Philipps and al., 1995;
Saarenketo, 1998; Fukue and al., 1999, and Friedel in various areas (Ritz et al., 1999; Jongmans et al.,
and al, 2006) shown that the clay resistivity values 2000; Demanet et al., 2001; Lebourg et al., 1999;
are strongly dependent with the water content. They 2001; 2005; Wise et al., 2003; Lapenna et al., 2003).
conclude that there is a critical water content (depend- However, very few studies have been realised in order
ing with the pore distribution) producing a strong to obtain data on the water spatiotemporal distribution
resistivity value modification (by a factor 1000 in their and its role in the sliding dynamic. Recently a first ten-
case). Consequently a strong resistivity gradient can tative has been successfully realised on the La Clapière
be associated with the presence of water in the case of landslide (Alpes-Maritimes, SE France) (Lebourg
a clay ground. et al., 2001, Jomard et al. 2005). Their preliminary
During the last few years, a large number of results allowed the identification of the sliding surface
electrical surveys have been conducted to acquire and the network water drainage strongly dependent
information about fault geometry or sliding surface with the inherited tectonic structures.

421
 possibility to prove a high and fast correlation between
an external water injection and resistivity changes.
Finally, we performed an original survey device based
on resistivities, pluviometric and piezometric data
acquisitions (Figure 3).

2 HISTORICAL, GEOLOGICAL AND

HYDROGEOLOGICAL SETTINGS

The ‘‘Pra of Julian’’ landslide, more commonly called

the ‘‘Vence’’ landslide, is regarded like a rotational
one including 1.2 × 106 m3 of material. It affects an
area about 250 m by 350 m, with a slope around 12◦
to 20◦ . The current landslide activity is underlined by
some surface expressions appearing in the landscape
morphology: tension cracks, scarps, disorders affect-
ing constructions and particularly the deviation of the
‘‘Lubiane’’ river at the foot of the slope. The histori-
cal activity of the landslide is related since 1950/1960,
with the first constructions, but was restricted to minor
disorders (little cracks in the soil). During the 70’s
years, urbanization increased (9 houses in 1970 to 17
in 1980) and more disorders appeared a few days after
strong rainfalls (250 mm in 2 days). The acceleration
of the landslide kinematic induced the obstruction of
the ‘‘Lubiane’’ river bed. In October 1981, an intense
rainfall episode (250 mm of 24 hours) generated mud
flows. In 1989, the number of houses was about 26.
From 1990 to present, the area was more and more
urbanized and the sliding phases occur always with

Figure 2. Photography of the landslide (A) and the mains

scarps (C) of the studied area.

This study is based on an approach using resistiv-

ity method, with a temporal follow-up, on a landslide
located in Vence (Alpes Maritimes, SE France), The
aim is to characterized the drainage channel system
and its evolution in time and space by the resistivity
variations.
The choice of this landslide was motivated by:
1) its size is well adapted to such dense geophysics
researches, 2) its geological framework is well known
and present the advantage to illustrated the possible
role of folded and fractured structure in the draining
channel network, and, 3) raining conditions can be
concentrated on a short time with an abundant volume
of water and clearly control the landslide activity.
In a first time, our work has been focused on
geotechnical, geological and hydrogeological inves-
tigations. After that, we include the fluids propagation Figure 3. Localisation of the electrical tomography pro-
survey by ERT in order to demonstrate: 1) the precise files (ERT1, ERT2 and ERT3) and the temporal tomography
geometry of the potential sliding surface and, 2) the profile (ERT4 & PERM_ERT).

422
intense rainfall episodes. At the end of 2000, the land-
slide activity became more intense with the opening
of important tension cracks in the upper part and the
destabilisation of the front with a scarp about several
meters.
Geology and geomorphology: The studied area is
localized within the NW-SE trending the Oligocene
‘‘Vence’’ fold related to the alpine tectonic deforma-
tions (Laurent, 1998).
The hydrogeology of the sector shows the presence
of a free water table in the sand and clay layer and a cap-
tive water table in marly limestones. The presence of
faults in the limestones creates a connection between
the Cretaceous calcareous permeable reservoir and the
overlying sandy-clay layer.
F

3 A CONCEPTUAL MODEL OF THE ‘‘VENCE’’ F

LANDSLIDE

The conceptual model consisted in the definition of the

geometrical limits observed on the field and also limits
in depth of the sliding surface. In order to describe it,
we realised several ERT in two dimensions (not devel- F
opped in this short paper): three profiles perpendicular
to the slope. They enable us to approach the morphol-
ogy of the active slide surface, and to highlight drained F

faults in the landslide.

To understand these complex relations between
water and structure of a landslide we carry out and Figure 4. ERT 1, 2, 3 (resistivity in Ohm.m), vertical
compare the three types of measurements: faults (F) and landslide surface in dotted points).

– Large scale profiles to complete the geometri-

cal/conceptual model of the Vence landslide, and the sliding mass and the rock substratum by draining
to calibrate the resistivity values with the boreholes structures. In this study, we want analyze the water cir-
and field observations (not developed in this paper, culation during the time in the sliding mass (ERT4 by
ERT1, 2, 3), injection) and the rate between the higher watertable
– temporal measurements (ERT4), based on a con- and the drained faults system. This correlation can
trolled/artificial water injection (controlled quan- be obtained by the coupled analysis between temporal
tity) coupled with an ERT survey. ERT4 was resistivity variations (Perm_ERT), the elevation of the
performed on the top of the basal scarp, during piezometric level and the analysis of the rainfall.
2 days with the geophysical survey of the evolution We use two profiles at different scales. The first one
of the resistivities at a regular temporal follow-up is parallel to the middle active scarp (ERT4), during
(short period of about 1 hour at the beginning and a 2 days period and coupled with a controlled water
3 times 24 h after the first day), but not developed injection. The second one is longer and parallel to
in this paper. the active scarp of the slip and crossing the landslide
– permanent measurement device (Perm_ERT) to (Perm_ERT). It is performed during three months and
analyse the correlation between daily piezometric coupled with the rainfall measurement.
and rainfall measurements. The first phase of measurement was realized during
summer 2005. We imposed a controlled quantity of
water close to the electrical profile (ERT4, Figure 3,
4 RESISTIVITY TEMPORAL FOLLOW UP
(Jomard et al, 2006)). In order to explore the resistivity
variations assign to the injected water, we carried out a
4.1 ERT 4: to analyse the resistivity soil evolution
measurement before (reference), during and after the
versus water injection
injection of water (Figure 9).
The precedent approach shows that there are several The water was injected between the electrodes 4
origins of water supply in the solid mass: the rain and 5. The interpretation of the first profile (T0) cor-
fall (external factor) and the interconnection between related with the field information (sliding surface at

423
 from external conditions. As we have just shown it
(4.1), a temporary water contribution on the surface
involves a modification of the resistivity, but what can
be the response resistivity measurement at a natural
scale?
The acquisition period presents here (resistivity,
rainfall and piezometric level) is from July 2006 to
November 2006.
The first treatments (calibration of the repeatability
of measurement), which are not the aim of this paper,
enabled us to show that there was a daily repetitivity
of the measurement: a maximum average variation of
+/−0.2 Ohm.m/day without climatic changes.
For each day we consider:
– The average of the 574 points of measurement
Rho_day (Rd_n),
– The standard deviation of the 574 points (SdDay_n).
After different studies of the data distribution, we
choose to represent one acquisition day by the data
average: Rd_n (n from day 1 st July the first to the last
day November the first).
To highlight the differences from a day to another,
we calculate the difference between the resistiv-
ity value of a day (Rd_n) and the average of the
resistivity day considering a three month period
(Mean_Rho_day: Rd3 months).

4.3 Discussion and interpretation of data

Figure 5. Temporal evolution of resistivity (Ohm.m)/ The observations carried out over the 3 months period
rainfall (mm)/piezometric level (cm), (Rho_n average of the show that there is a very fast response of the field
574 points of measurement Rd_n) and A, B C and D are four to the rainy episodes 4.3, as described with the
rainfalls events.
ERT4. We observe a several centimetres increasing
of the piezometric level and a period of 24/48 h of
12 m) confirm us the presence of the sliding surface stabilization.
towards 12 m of depth in the first right half part of the According to the resistivity survey, we measured:
tomography and an increase of the sliding surface on – in a drilled hole of 20 m of depth near ERT_perm the
the level of the electrode 24, which follows the cal- evolution of the piezometric level. Measure carried
careous substratum until its appearance on the field to out manually with a centimetric resolution,
the level of electrode 32. – near the site (200 m from ERT_perm) we daily quan-
tify the rainfall with a pluviometre characterized by
4.2 Perm_ERT : 3 month of follow up to a millimetric resolution.

We propose to focus on 3 months of monitoring with When we observe the evolution of the curves
a daily measurement: resistivities/rainfalls/piezometry we can analyse:

– ERT with 574 points of measurement, – that important precipitations (events A, B, C and D)
– Measurement of the piezometric level (with a make decreasing resistivity,
centimetre resolution). – at the same periods we could observe a strong
– Measurement of rainfall events with a pluviometre increase in the piezometric levels (an increase of
(with a millimetre resolution). 5 cm after the event C (2006/09/25) measured the
2006/09/27 due to strong precipitations). This event
On the basis of the ERT measurement, we try to involved a very important reduction in the daily
extract a signal representing the resistivity variations resistivity average. Thus, the daily average falls of

424
 39.19 Ohm.m to 38.35 Ohm.m, which represents a water content. The results of hydrogeological investi-
signal/noise ratio of 3. At the same time we obs- gations and the influence of weather conditions will
erved an increase in a factor 5 of the piezometric allow establishing an evolutive slope stability calcu-
level. lus. Finally, it could possible to related an electrical
signal with a slope stability value.
These observations on the event C (September 25
2006) are the same for event A, B and D, but to a lesser
extent.
Several remarks which will be the subject of a new
REFERENCES
work are to be brought:
Notice 1: The traditional analysis by inversion of the Demanet Donat, Eric Pirard, François Renardy & Denis
data (RES2Dinv, Loke, 1997) and the subtraction of Jongmans, 2001. Application and processing of geo-
the matrices of inversion (Jomard and al., 2007) do physical images for mapping faults. Computers & Geo-
not allow highlighting an effective signal. sciences, Volume 27, Issue 9, 1 November 2001, Pages
1031–1037.
Notice 2: The ERT survey seems to be a powerful/ Eberhart-Phillips D., William D. Stanley, Brian D. Rodriguez
adapted tool in the study of the rainfall impact on the & William J. Lutter, 1995. Surface seismic and electri-
moving mass. . .. But, it appears necessary to develop cal methods to detects fluids related to faulting. Journal
an adapted data treatment related to the temporal of geophysical research, vol. 100, NO. B7, pages 12,
aspect of the survey. 919–12, 936.
Notice 3: The distribution of the 574 points of Friedel S., Thielen A. & Springman S.M. 2006. Investigation
measurement is carried out according to a bimodal of a slope endangered by rainfall-induced landslides using
distribution. Indeed we can observe that the data 3D resistivity tomography and geotechnical testing. Jour-
are distributed around a first peak located between nal of Applied Geophysics, Volume 60, Issue 2, Pages
100–114.
20 and 25 Ohm.m, then that the other data are dis- Fukue M., T Minato., Horibe H. & Taya N. 1999. The
tributed according to another mode from 50 Ohm.m. micro-structures of clay given by resistivity measure-
The average of the resistivity is thus not inevitably the ments. Engineering Geology, Volume 54, Issues 1–2,
tool most representative of the variation of the elec- September, pp 43–53.
trical measurements, nevertheless this mathematical Griffiths, D. H. & Barker, R.D. 1993. Two-dimensional
variable is for the moment the most representative of resistivity imaging and modelling in areas of com-
the daily group of 574 measurements. plex geology: Journal of applied geophysics, v. 29, p.
211–226.
Jomard, H., Lebourg T., Binet S., Tric E. & Hernandez
M. 2006. Characterization of an internal slope move-
ment structure by hydrogeophysical surveying: Terra
5 CONCLUSION Nova.
Jongmans, D., Hemroulle P., Demanet D., Renardy F.
The electrical resistivity tomography is a powerful tool & Vanbrabant Y., 2000. Application of 2D electri-
to constraint geometrical and geological boundaries cal and seismic tomography techniques for investigating
in landslide areas. A complex problem in landslide landslides: Eur. J. Environ. Eng. Geophys., v. 5, p.
survey is related to the temporal survey of triggered 75–89.
factors as the water. Lapenna, V., Lorenzo, P., Perrone, A. & Piscitelli,
S. 2003. High resolution geoelectrical tomogra-
On the basis of the electrical sensitivity toward phies in the study of Giarrossa Landslide (South-
the soil water content, we show the possibility to ern Italy): Bull. Eng. Geol. Env., v. 62, p.
have a quantitative approach of soils response from 259–268.
external water solicitations, using electrical resistivity Laurent O., Stephan J-F. & Popoff. M. 2000. Modalités de la
measurement. structuration miocène de la branche sud de l’arc de Castel-
It seems to be accessible to clearly associate a rain- lane (Chaînes subalpines méridionales). Géol. France, 3,
fall measurement with a piezometric variation and p.33–65.
with a electrical signal. In this case, another mathe- Lebourg, T., Frappa M. & Sirieix C. 1999. Reconnaissance
matical data treatment must be considered in order to des surfaces de rupture dans les formations superficielles
instables par mesures éléctriques: PANGEA, v. 31, p.
have higher signal accuracy and to locate the ground 69–72.
areas where the variations are concentrated. Lebourg, T., Binet, S., Tric, E., Jomard, H. & El Bedoui,
In the case of the ‘‘Vence’’ landslide, two aspects S. 2005. Geophysical survey to estimate the 3D sliding
must be more improved: 1) the impact of the drained surface and the 4D evolution of the water pressure on
fault in the water flow inside the slide mass, 2) part of a deep seated landslide: Terra Nova, v. 17, p.
the mechanical parameters evolution toward the soil 399–407.

425
Lebourg, T. & Frappa, M. 2001. Mesures géophysiques pour weathering mantles over granitic and metamorphic
l’analyse des glissements de terrain: Revue Française de basement of eastern Senegal, West Africa. Journal of
Géotechnique, v. 96, p. 33–40. Applied Geophysics, Volume 41, Issue 4, June 1999,
Loke, M.H. 1997. Res2Dinv software user’s manual. Pages 335–344
Loke, M.H.& Barker, R.D., 1996, Rapid least square inver- Rizzo E., Colella A., Lapenna V. & Piscitelli, S. 2004.
sion of apparent resistivity pseudosection by a quasi High-resolution images of the fault.
Newton method: Geophysical research letter, v. 44, p. Saarenketo Timo. 1998. Electrical properties of water in clay
131–152. and silty soils. Journal of Applied Geophysics, Volume
Ritz Michel, Parisot Jean-Claude, Diouf S., Beauvais A., 40, Issues 1–3, pages 73–88.
Dione F. & Niang M. 1999. Electrical imaging of lateritic

426
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Characteristics of landslides related to various rock types in Korea

S.G. Lee, K.S. Lee & D.C. Park

Department of Civil Engineering, The University of Seoul, Seoul, Korea

S. Hencher
Halcrow China Ltd, Department of Earth Sciences University of Leeds, UK

ABSTRACT: Korea experiences numerous landslides every summer due to intense rainfall and typhoons
which result in considerable damage to property, loss of life and disruption to traffic. This paper concerns the
nature of landsliding related to rock type in Korea and is based on empirical studies of landslides together with
sensitivity analysis of slopes with respect to groundwater levels. The vast majority of natural terrain landslides
are very shallow and associated with the sliding of thin weathered soil horizons on underlying stronger rock,
irrespective of rock type.

1 INSTRUCTION Table 1. Landslide survey areas and numbers of landslides.

Many landslides occur each summer in Korea due to Rock types Location 2001 2002 2003 Total
heavy rainfall. These result in 60 deaths on average,
Igneous Hongcheon (I-1) 36 − − 36
hundred of millions US dollars damage to property
rocks Gangneung (I-2) − 46 44 90
and considerable disruption to traffic flow (Lee & Donghae (I-3) − 46 − 46
Hencher, 2007). Gimcheon (I-4) − 25 − 25
More than 70% of Korea is mountainous. Typically Samcheok (I-5) − − 8 8
in these hilly areas, only a thin layer of soil overlies
rock and the majority of landslides occur at the bound- Metamor- Hamyang (M-1) − 26 − 26
ary between rock and the overlying soil. In terms of phic rocks Geochang (M-2) − − 4 4
Samcheok (M-3) − − 7 7
scale, most natural terrain landslides are less than 20
to 30 m in length, less than 10 m wide and less than Sedimen- Pohang (S-1) − 2 − 2
1 m deep (Lee, 1987, 1988, 1995; Lee et al., 2008). tary rocks Hapcheon (S-2) − 19 4 23
Rainfall is an important trigger of landslides but
Total 36 164 67 267
this is not the only factor. Other important susceptibil-
ity conditions are underlying geology and topographic
situation. This paper reports on a study to investi-
gate the relationships between landslide occurrence
and landslide characteristics and geological factors. of those areas finally selected areas finally selected
Furthermore, back analysis has been conducted for study are shown in Table 1 and Figure 1 (Um &
to look at the sensitivity of different geological and Reedman, 1975).
topographic settings to changes in groundwater con-
ditions.
3 METHOD OF FIELD STUDY

2 SELECTION OF SURVEY AREAS At each landslide location various aspects were

recorded. Overall slope angle was measured using a
Areas were selected for study that had numerous land- clinometer; depths, lengths, and widths of each fail-
slides triggered by heavy rainfall in the summers of ure were measured physically using tapes. The nature
2001–2003 and where the landslide damage is still evi- of the slip surface was recorded using Schmidt Ham-
dent and amenable to measurement and analysis. Areas mers (SHV value) to provide quantitative data on the
were also selected for study with the aim of discerning nature of underlying rock, where appropriate (ISRM,
the influence of geological factors. The distribution 1978).

427
 1.Resident Soil:0.5

2.SHV:30-35
2.SHV:45-50
3.Sheeting joint

Width : 21m
Length : 32-33m

Figure 3. View and geological feature of igneous rock area

landslides (Site I-2).

Figure 1. Locations and geological feature of survey areas.

1.Complete
weathered:0.4-0.5m
2.Weekrock layer
SHV:38-52

3.Joint surface Foliation direction

Width : 15m
Length : 35-40m

Figure 4. View and geological features of typical landslide

in metamorphic rock area (Site M-2).

Figure 2. Schematic representation of surveyed features.

1.Complete SHV: 27-29 SHV: 26-28
weathered:0.5m
4 GROUND CHARACTERISTICS
OF LANDSLIDES 3.Weathered shale

4.1 Igneous rock areas

Most of the surveyed areas underlain by igneous rock SHV: 34-38
have a thin surficial layer of about 0.5 m of weath-
ered granite soil (residual and completely weathered). Width : 16m
2.Sandstone layer SHV: 35-37 Length : 25m
The underlying rock is typically highly to moderately
weathered (‘‘soft rock’’ in Korean terminology) with
Schmidt Hammer rebound values of the exposed sur- Figure 5. View and geological feature of sedimentary rock
face in the range of 30 to 50 (Hencher & Martin, 1982) area landslides (Site S-2).
(Figure 3).

4.3 Sedimentary rock areas

4.2 Metamorphic rock areas
In areas underlain by sedimentary rock there is, as in
Most of the metamorphic rock areas are underlain by the igneous rock and metamorphic rock areas, a thin
gneiss and, as in the granitic areas, there is a thin layer soil horizon of perhaps 1 m developed, underlain
of soil with a thickness of 1–2 m underlain by highly, by relatively strong sandstone and weak shale. The
moderately and slightly weathered rock with Schmidt Schmidt Hammer rebound values of the relatively
Hammer rebound values on exposed failure surfaces strong sandstone were in the range of 25–35 which is
typically of the order of 35–55. indicative of a moderately weathered state for a rock

428
that is moderately strong (unconfined compressive In case of igneous rock areas and metamorphic rock
strength >12.5 MPa) or stronger in its fresh state areas, the frequency of occurrences of landslides at 8th
(Anon, 1995). The shale was much weaker with no or lower ridges was low; however, in case of sedimen-
rebound value from a Schmidt Hammer. tary rock areas, the landslides were occurred at 5th or
lower ridges at fairly high frequencies.

5 GEOMETRICAL CHARACTERISTICS
OF LANDSLIDES 5.3 Longitudinal section types of slopes
Collapse type was linear in terms of longitudinal
5.1 Original slope angles section type regardless of rock types (Figure 8). Espe-
The slope angles of the surveyed areas were measured cially, 99% of the landslides on sedimentary rock areas
using a clinometer. The average angle of those slopes were occurred on parallel slopes and the cause was that
in which landslides occurred was 39◦ . By rock type, the longitudinal direction slopes in sedimentary rock
landslides were generally developed in areas underlain areas were almost linear.
by igneous rock in relatively steeper terrain with an
average angle of 43◦ ; for metamorphic rock areas the 5.4 Cross section types of slopes
average was 39◦ ; for sedimentary rock the average was
35◦ with 15% occurring in terrain with an original As shown in Figure 9, the cross section types at the
slope angle of less than 30 degrees (Figure 6). landslide points were classified into three types; the
majority of landslides were developed in concave,
depressed topography in all rock types although some
5.2 Location of landslides relative to mountain landslides occurred in flat terrain. The occurrence of
crest landslides in depressed terrain is probably related to
The locations of collapse starting points were classi- surface flow concentration allowing saturation of the
fied into 10 kinds between the peak points and lowest
points starting from mountain tops or mountain ridges.
In terms of the location where the collapse started,
most collapses frequently started at the 7th–8th ridges.

Figure 8. Longitudinal section types by rock types.

Figure 6. Original slope angles by rock types.

Figure 7. Location of collapse starting points by rock types. Figure 9. Cross section types by rock types.

429
Figure 10. Slope angles by rock types after collapse. Figure 11. Collapse lengths by rock types.

surficial soil layer, the development of flows along the

boundaries between rocks and soil and elevated ground
water pressures above those boundaries resulting in
reduced shear strength. Surficial and internal erosion
(piping) may also be important factors (Hencher et al.
2006).

5.5 Slope angles after collapse

The average angle of all slopes after occurrence of
landslides was 37◦ and in terms of the average values
by rock type, igneous rock areas showed 41◦ , meta-
Figure 12. Collapse widths by rock types.
morphic rock areas showed 36◦ , and sedimentary rock
areas showed 32◦ (Figure 10).
The average slope angle after collapse was 37◦ and
was smaller than the original average slope angle, 39◦ ,
by 2◦ . In terms of the change in slope angles by rock
type, the angle was changed from 43◦ to 41◦ in case of
igneous rocks, from 39◦ to 35◦ in case of metamorphic
rocks, and from 35◦ to 33◦ in case of sedimentary
rocks.

5.6 Length of collapse

The lengths of collapse were physically measured in
the filed with tape measures. The average length was
30 m: in terms of rock types, igneous rock landslides Figure 13. Collapse depths by rock types.
were on average 29 m, metamorphic rock landslides
41 m, and sedimentary rock landslides 13 m. Most of
the igneous rock and sedimentary rock landslides were
less than 25 m of but landslides in metamorphic terrain was 10 m; igneous rocks showed 11 m, metamorphic
were 40 m or longer in 46% of cases (Figure 11). rocks showed 12 m, and sedimentary rocks showed
8 m (Figure 12).
Although the collapse widths differed slightly
5.7 Width of collapse according to rock type, most of the landslides were
Generally the width of landslides were constant from 10 m or smaller regardless of rock type.
the upper part to the lower part as might be expected for
such shallow detachments of soil above rock, typical
5.8 Collapse depth
of most of the studied landslides. In some cases how-
ever, where the width was different in upper, central The average depth of landslides was 1.0 m: igneous
and lower parts, average values were recorded as the rocks showed 0.8 m, metamorphic rocks showed 1.1 m,
collapse widths. The average of widths of landslides and sedimentary rocks showed 0.9 m (Figure 13).

430
 Figure 16. Collapse volume by rock types (Back analysis).

Figure 14. Original slope height versus original slope angle.

5.11 Collapse volumes
Collapse volumes were compared by rock types as
shown in Figure 16. Landslides in sedimentary rock
and igneous rock areas were typically 100 m3 or less
in volume in many cases. However, in metamorphic
rock areas landslide volumes were sometimes rather
larger.

6 CRITICAL GROUNDWATER PRESSURES

FOR FAILURE BY BACK ANALYSIS

Figure 15. Collapse areas by rock types. In Korea, it is recommended cut-slope design stan-
dards that ground water is assumed up to ground
surface, i.e. 100% (KEC, 2001).
Landslides in areas underlain by sedimentary and The Slope/W program was used to carry out
igneous rocks were typically shallower than 1 m but in back analysis of trial landslides using the Janbu
metamorphic rock areas they were deeper than 1.5 m (1954) method for interpretation. The shear strength
in 37% of cases which is indicative of a greater devel- of boundaries between soil and rocks for different
oped thickness of weathered soil in areas underlain by rock types were measured under saturated condi-
metamorphic rock. tions using laboratory direct shear tests (Lee et al.
2008).
When back analysis was performed in consider-
5.9 Angles and heights of original Slopes ation of the properties of each rock type and with
changing the ground water table, the time of col-
It is shown in Figure 14 that most landslides occurred lapse occurrences was regarded as Fs = 1; and then,
in slopes with original angles of 30–50◦ of angle and this assumption was applied to 4 areas to which the
heights of 10–30 m regardless of rock type although back analysis was to be applied and the result was
the distribution was different depending upon rock arranged as shown in Table 3 and Figure 17. The
type. Especially, sedimentary rocks showed the land- underground level on the upper part of rocks at cre-
slides of which the slope angles and heights were same ation of landslides was assumed by back analysis;
with the original angles and heights, compared to other the distribution at such level was 25–39% in case of
rock types. igneous rock areas, 20–44% in case of metamorphic
rock areas, and 35–49% in case of sedimentary rock
areas.
5.10 Collapse areas
The reason for such difference in underground
Surface areas of landslides were 500 m2 or smaller in water levels by rock types in back analysis was
97% of cases in igneous rock areas, 61% in meta- deemed that the original slope angle influenced in
morphic rock areas, and 91% in sedimentary rock formation of underground water levels in natural
areas. slopes.

431
Table 2. Percentages of underground water levels by rock overlying stronger, less weathered rock. The boundary
types when the factor of safety by back analysis is 1. is often well defined and acts as an aquiclude allowing
the development of perched water tables and through
Site no. flow. The thickness of the weathered soil layer is some-
times greater in the case of metamorphic rocks but
Rock types 1 2 3 4 this may be partly related to the location of landslides
Igneous rocks 39% 39% 25% 36% in metamorphic terrain in valley bottoms rather than
Metamorphic rocks 44% 20% 36% 20% side slopes (this may be a study-specific conclusion
Sedimentary rocks 35% 40% 44% 49% rather than generally applicable). Failed slopes typ-
ically occurred in terrain with original slope angle
of 30–50◦ . Landslides do however occur in gen-
tler terrain in areas underlain by metamorphic and
Table 3. Result of back analysis of ground wuawater level, sedimentary rock.
case of igneous rock area no.1 in Table 2. Landslides were typically 20–30 m long, and less
than 10 m wide; 85% of collapses occurred at 7th–8th
Ground water Factor of Ground water Factor of ridges. In terms of longitudinal section type, 80–90%
level (%) safety (FS) level (%) safety (FS) of landslides were created at parallel slopes; in terms
100 0.738 50 1.056
of cross section type, 70–80% of landslides occurred
90 0.819 40 1.109 in depressed areas, probably associated with surface
80 0.878 33 1.163 water concentration.
70 0.937 30 1.163 Also, the underground water level when the fac-
60 0.994 20 1.216 tor of safety by back analysis for each rock types is 1
10 1.269 was obtained; igneous rocks took 25–39%, metamor-
phic rocks took 20–44%, and sedimentary rocks took
35–49% in whole soil layers; such small difference
was deemed to be caused by the difference in outflow
of underground water depended upon the difference in
slope angles of original slopes.

ACKNOWLEDGEMENT

This research was partially supported by a grant

(NEMA-06-NH-05) from the Natural Hazard Mitiga-
tion Research Group, National Emergency Manage-
ment Agency.

REFERENCES

Anon 1995. The description and classification of weathered

rocks for engineering purposes. Geological Society Engi-
Figure 17. Estimated result of ground water level by back neering Group Working Party Report. Quarterly Journal
analysis (Slope/W Interpretation cross section). of Engineering Geology, 28, pp 207–242.
Korea Expressway Corporation (KEC). 2001. Road design
handbook (II). 406–410.
Hencher, S.R. & Martin, R.P. 1982. Description and clas-
7 CONCLUSION sification of weathered rocks in Hong Kong for engi-
neering purposes. Proceedings of the 7th Southeast Asian
Landslides triggered by rainfall in 2001 to 2003 in Geotechnical Conference, Hong Kong, 1, pp 125–142.
Korea have been studied to examine their typical Hencher, S.R., Anderson, M.G. & Martin, R.P. 2006.
Hydrogeology of landslides. Proceedings of International
characteristics and to consider any variation with Conference on Slopes, Malaysia, pp 463–474.
fundamental geology. International Society for Rock Mechanics (ISRM). 1978.
The vast majority of natural terrain landslides stud- Suggested methods for the quantitative description of dis-
ied, irrespective of rock type, are associated with the continuities in rock masses. Int. J. Rock Mech. Mining Sci.
sliding on thin, 1–2 m thick, weathered soil horizons Geomech. Abstr., 15: 319–368.

432
Janbu, N. 1954. Application of composite slip circles for Lee, S.G., Hencher & Kim, B.S. 2008. A study on the shear
stability analysis. Proc. European Conference on Stability strength of boundaries between soils and rocks in Korea.
of Earth Slopes, Stockholm, 3: 43–49. Proc. 10th Int. Symp. on Landslides and Engineered
Lee, S.G. 1987. Weathering and Geotechnical characteriza- Slopes, Xi’an, China (in press).
tion of Korean Granites. PhD thesis. Imperial Collage. Lee, S.G. & Hencher, S.R. 2007. Slope safety and landslide
University of London. risk management practice in Korea. Proc. of 2007 Inter.
Lee, S.G. 1988. A study on landslide in Korea. Researches Forum on Landslide Disaster Management, Hong Kong
on geological hazards, Research report of Korea Institute (in press).
of Geoscience and Mineral Resources (KIGAM), KR-88- Um, S.H. & Reedman, A.J. 1975. Geology of Korea. Korea
(B)-7: 145–148. Institute of Geoscience and Mineral Resources (KIGAM),
Lee, S.G. 1995. Natural hazard in Korea. Proc. of the int. Seoul.
Forum on Natural hazard mapping, Geological Survey of
Japan Report (281): 145–148.

433
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Two approaches to identifying the slip zones of loess

landslides and related issues

Tonglu Li & Xiaoyan Lin

Department of Geological Engineering, Chang’an University, Xi’an, China

ABSTRACT: With the rapid development of highways on the loess area in recent decades, a large number of
loess landslides are encountered and some of them have to be controlled. Unfortunately, the invalid technical
methods for landslide exploration generally employed often led to misjudgment of their geological conditions
and stability that make stabilizing construction either failure or extra investment. For landslide exploration, it
is a key problem to locate the slip zone, so we are going to discuss some practical skills in loess landslide
exploration after the research project we have finished which funded by Communication Bureau of Shaanxi
Province, China. The skills include drilling technique in loess landslides, identification of loess slip zone and
use of electric resistivity for inspecting the slip zone. From in situ and laboratory tests and observation, we have
suggested a series of easy operating and effective ways for loess landslide exploration.

1 INTRODUCTION bit disturb the soil seriously, so the soil from borehole
can seldom be used to identify the slip zone. Therefore,
Loess covers about 631 thousands square kilometers we have researched on the aspects of drilling skills of
of Chinese land surface, occupies about 4.4 percent loess landslides, identification of the loess slip zone,
of the total land area of China (Sun, 2005). It mainly and application of electric resistivity, which consists
distributes in the central area of China, the middle of composite methods for loess landslide exploration.
reach of the Yellow River in where the source of Chi-
nese civilization is located. The loess region has been
rising up by wind blown deposit while cutting down 2 THE DRILLING SKILLS OF LOESS
by the tree-like river system. The two reverse pro- LANDSLIDE
cesses consequently produce a dense distributed loess
slopes in the sides of the river valleys. As they are It is common concerned to determine the slip zone of
going on, the slopes are becoming higher and steeper, landslides in their exploration, because the features of
finally to form a large number of landslides. Statis- loess slip zone is changeable. It may be thin, few mil-
tics of the investigation data suggests that there are limeters to even a contact plane and may be thick, from
16616 landslides and soil falls developed in the north several centimeters to meters, may has typical shearing
of Shaan’xi, the density exceeds 5 ones in each square traces and may has not, may be single layer or several
kilometer. In the eastern and western Gansu Province, layers. Therefore, we need continuous undisturbed soil
14109 loess landslides had developed since the end core from boreholes to identify the slip zone, but as
of 1950s to 1992, the density exceeds 6 ones each mentioned above, the present used drilling skills is not
square kilometer (Lei, 2001). Loess landslides have satisfactory for the purpose. The general used method
been the main geological hazards in the traffic and of continuous flight auger or percussion hammer in
civil engineering construction. Therefore, some of the foundation exploration disturbs the soil completely in
landslides which the engineering encountered have to the process of driving down, so the sampler is used to
be controlled. In-situ Investigation and exploration of collect undisturbed soils in certain depth in this case.
the landslides should be taken before design of the While for landslide drilling, the depth of slip zones
controlling construction. But the engineers now still is difficult to predict and it is not practical to sam-
use some unreasonable methods to give a unreliable ple in a long range continuously, so the methods is
geological information, which led to the design being not available. Generally, the direct-rotary method is
either high risk or high invest, such as the general used to drill rock mass and sometimes to drill soil
used continuous-flight auger or direct rotary drilling under water level. This method usually is carried out

435
with circulation-liquid for lubricating and cooling the easily taking the soil core out and the outer one has
bit. In the most codes and specifications of China, it a continuous flight auger round its out surface. The
is prohibit to use circulation-liquid above the water tip of the outer tube attaches three hard steel blades as
level in landslide drilling because the liquid will soak cutter heads whose width is in the range from the out
the soil core and make it dispersed easily. As drilling side of the auger to the inner wall of the inner tube.
without circulation-liquid by this method, the sample The out side of the auger attaches vertical steel rids on
is distorted seriously as showing in Figure 1, so it is each side which plays a role in smoothing the borehole
not available too. wall and keeping vertical of the hole. Fig. 2A shows
In ASTM standard (2003), an hollow stem auger is the feature of the designed auger.
introduced to drilling in soils as an conventional meth-
ods in American, but it is seldom applied in China.
Referring this technique, we design a similar auger
which has dual tubes with the inner tube static, the
outer one rotated. The inner tube can be spited for

Figure 1. Loess samples taken by direct rotary method

without circulation fluid. (A) Brown-yellow Loess and red Figure 2. A dual tube auger (A) A sketch of the auger. (B)
paleosol mixed together, completely disturbed; (B) The Drilling with the auger, the soil cohered on the flight of the
white concretions in brown-red paleosol was crushed, also outer tube was taken off, then the soil core in the inner tube
completely disturbed. was taken out while.

436
 The drilling test with the auger was carried out in a
large loess landslide, the Taiping landslide to the north
of Xi’an City. With rotation of the drilling rod, the cut-
ter head would cut down the outer ring part of the soil
first and keep the inner rounded part without disturb-
ing; meanwhile the cuttings of the soil outside would
go along the flight auger and the soil core inside goes
into the inner wall up. The soil cohered on the flight of
the outer tube was taken off (Figure 2B) while the soil
core in the inner wall was taken out. The soil core keeps
dry and intact. Figure 3A is the sample taken by the
auger, which has really perfect original structure of the
loess. The samples could be used to not only identify
the slip zone, but to do the laboratory tests for phys-
ical and mechanical properties. Figures 3B and 3C
show the core soils taken from the slip zone of which
the original material is loess and paleosol respectively;
they are schistose and compacted apparently.

3 IDENTIFICATION OF THE LOESS

SLIP ZONE

The above technique provides us a continuous intact

soil core, but how to recognize the slip zone from the
core is also an essential work in landslide exploration.
Megascopically, as observing the structures of loess
from slip mass through slip zone to slip bed in the exca-
vated investigation pits or the soil core from boreholes,
the differences among them are apparent. The slip
mass has loose structure and the red paleosol generally
is enclosed by yellow-brown loess or inter-enclosed
each other. The root holes have been damaged by com-
pression and extension in most portions, but locally
they can be seen in the centre of some soil blocks.
These appearances are not difficult to be observed.
The soil of slip zone shows compact and homogonous
structure, the root holes are crushed completely by
extensive compressing and shearing. The foliations
and striates within the zone could be seen times but not
always. The soil of slip bed has the typical structure of
loess that is homogonous with vertical root holes and
inter-bedded loess-paleosol sequences. Both in exca-
vated pit and soil cores, looking for the slip zone should
be from lower to upper. Because the slip bed is easy
to be recognized, as the soil structure going to com-
plex, it implies that the soil was disturbed by slipping.
Even we could not found the typical features of slip-
ping movement as foliations and striates, variation of
the soil structures is helpful to judge the position of
slip zone.
Microscopically, the soil of slip zone shows appar-
Figure 3. Soil cores taken by dual tube auger (A) From
ent compacting and distorting effect relative to the
slip bed, the loess the perfect original structure of root holes natural soil. Comparing the SEM photos under same
being clearly seen; (B) From slip zone, loess become schis- amplified times between the loess undisturbed and
tose and dense; (B) From slip zone, paleosol become hard that disturbed in slip zone, it is found that the undis-
and compacted. turbed loess has typical hollow-skeletons structure,

437
 Plot these parameters and their relevant results verse
to the position relative to slip zone. It reveals that the
water content of slip zone is higher than that of slip
mass and slip bed, but there is one exception that the
water content above the slip zone is higher in Taip-
ing landslides (Figure 5A); plastic limit is lower in
slip zone without exception (Figure 5B); liquid limit
is higher in slip zone with one exception that the
Dongfeng landslide has a lower liquid limit in slip zone
(Figure 5C), plastic index is higher in slip zone, but in
which Dongfeng landslide is not so apparent (Figure
5D), and liquid index is higher in slip zone, in which
Taiping landslide is not so apparent (Figure 5E) and
average particle diameter is higher in slip zone in all
the four landslides (Figure 5F). It implies that abra-
(a) sion between slip mass and slip bed has made the soil
finer and compact, which correspondingly changed its
water content and plasticity. Even so the indexes have
exceptions in some of the landslides which may be
caused by the processes of sampling and testing, they
are helpful to determine the slip zones of the loess
landslides.

4 APPLICATION OF ELECTRIC RESISTIVITY

IN LOESS LANDSLIDES

Geophysical methods are often used to prospect the

slip zone of landslide, such as seismic method, elec-
tric method, geological radar and twinkling alteration
electric-magnetic waves etc. However, the methods
(b) based on wave transition may be suitable to inspect
the large scale and deep sited geological body. In land-
Figure 4. Comparing of the microstructures between the slide prospecting, the slip zone is generally no more
undisturbed loess and the loess in slip zone. (A) Microstruc- than 100 meters depth, so the error of these methods
ture of intact loess (500 times); (B) Microstructure of the is not suitable to locate the exact position of slip zone.
loess in slip zones (500 times). Electric method has the similar difficulties, so it should
not be available alone. While combining with in-situ
exploration, an effective result may be produced.
while the loess in slip zone is extensively compacted Therefore we apply the electric method in investi-
and the scales of clay minerals are distorted. However, gation pits and boreholes for further confirming the
the directed features megascopically do not reveal in slip zone. First, the electric resistivity was measured
microscopic structures (Figure 4). on the section crossing slip zone in the pits. They
Quantitatively, to find differences of the physical were measured in three states of natural, freshwater
indexes among the soils from slip mass to slip bed, saturated and salt liquid saturated to the sections. The
we sampled with short intervals of 5 cm on the four results shown in Fig. 6 clearly reflect that resistivity of
investigation pits in which the slip zone outcropped both natural and freshwater saturated decline sharply
out. The pits was excavated near the back of the four as they cross the slip zone; the resistivity of slip mass is
large loess landslides of Dongfeng, Taiping, Shutang- much higher than that of slip bed. That of the freshwa-
wang and Xiuchidu which located to the north side of ter saturated soil is a little lower than natural soil. As
Jinghe Rive and in the south side of a loess platform, the soil is saturated by salt liquid, the resistivity of slip
to the north of Xi’an City. The sampling section is mass and slip bed has no much difference, the elec-
nearly perpendicular to the slip zone and at least one tric property mainly controls by dielectric other than
sample in slip zone was collected. Then the samples soil properties. The high resistivity of slip mass may
were tested in lab for the parameters of water con- be because of its loose structure. The results suggest
tent, plastic limit, liquid limit, particle components. that we can directly measure the resistivity to confirm

438
 Water content(%) Plastic limit(%)
10 15 20 25 24 22 20 18 16 14
60 60

Slip Slip
mass
40 mass 40

20 20

Slip Slip
0 zone 0
zone

-20 -20
Slip Slip
bed bed
-40 -40

-60 -60
Depth(m) Depth(m)
(A) (B)

Liquid limit(%) Plastic index(%)

20 25 30 35 5 10 15 20
60 60

Slip Slip
40 40
mass mass

20 20

Slip Slip
0 0
zone zone

-20 -20
Slip Slip
bed bed
-40 -40

-60 -60
Depth(m) Depth(m)
(C) (D)

Liquid index Mean particle diameter(mm)

-1.5 -1 -0.5 0 0.5 1.0 0.025 0.020 0.015 0.010
60 60
Slip Slip
mass 40 mass
40

20 20

Slip Slip
zone 0 0
zone

-20 -20
Slip Slip
bed bed
-40 -40

-60 -60
Depth(m) Depth(m)
(E) (F)

Figure 5. Variation of physical indexes near slip zone. Of the figures, --denotes the samples collected in the investigation
pit on the back of Shutangwang landslide; -♦-denotes that of Xiuchidu landslide; --denotes that of Dongfeng Landslide;
-O-Denotes that of Taiping landslide.

439
 Resistivity( .m) Resistivity( .m) Resistivity( .m)
0 50 100 150 0 50 100 150 0 50 100 150
1.2 1.2 1.2

Slip mass 0.6 0.6 0.6

Slip zone 0.0 0.0 0.0

Slip bed -0.6 -0.6 -0.6

-1.2 -1.2 -1.2

Depth(m) Depth(m) Depth(m)
Measured on Xushidu landslide Measured on Dongfeng landslide Measured on Taiping landslide

Figure 6. Variation of eclectic resistivity near slip zone in the exploration pits. Of the figures, -♦-denotes measuring on
natural state; --denotes measuring after water saturated; --Denotes measuring after salt liquid saturated.

Borehole Zk2 Borehole Zk9

Depth Resistivity Depth Resistivity
Strata Column Strata Column
(m) (m)
30 60 90 120 30 60 90 120
2 2

4 4

6 Slip mass 6

8 8

10 10 Slip mass
Slip zone
12 11.3 12

14 14

16 16

18 18

Slip bed
20 20
Slip zone
22 22 22.3

24 24

26 26

Slip bed
28 28

30 30

Figure 7. Eclectic resistivity logging in boreholes.

440
the slip zone of loess landslides. This method could diameters is lower in it, therefore, the slip zone of
further be used in electric logging to determine the loess landslide can be easily identified.
slip zone in boreholes. Figure 7 are the electric logging 3. The resistivity has great discrepancy between slip
curves of the two boreholes in the Wohushan landslide mass and slip bed in loess landslide, the value of
near Yan’an City Shanxi Province. The whole holes are slip mass is much higher than that of slip bed. It
loess and the slip zone is not so confident by observing implies that the steepest point on the curve of resis-
the soil core. From the logging curve, we clearly see tivity verse depth is corresponding to the position
the discrepancy between slip mass and slip bed, the of slip zone. The method applied to borehole elec-
slip zone could be confidently confirmed. tric logging is simple and effective to confirm the
depth of the slip zone

5 CONCLUSION

In Chinese loess area, some of old landslides may be ACKNOWLEDGEMENTS

reactivated by natural agents or human engineering
disturbing, but many engineers are always confusable The research work was funded by the projects
to determine the slip zone. The main reasons are the from National Natural Science Foundation (Project
invalid drilling skills and unclear of the criteria for No.40772181) and Communication Bureau of Shaanxi
slip zone determination. Here suggests some practical Province. The postgraduate students, Long Jianhui, Fu
and easy operated methods in loess landslide explo- Yukai and Lei Xiaofeng, attended the field investiga-
ration. The above motioned is concluded as the follow tions and measurements. Here our thanks are extended
aspects: to the ones contributing this work.
1. A dual tube auger was designed for the loess
landslide drilling, which can collect continuous
undisturbed soil core for slip zone determination REFERENCES
and laboratory test of soil properties.
2. Determination of slip zone is the most important Lei Xiangyi. 2001. Geological hazards and human activity
work in landslide exploration. As observing in on the Loess Plateau. Beijing: Geological Press.
megascopic and microscopic scales carefully, the ASTM 2003. Annual Book of ASTM Standards, Vol.04.09,
West Conshohochen.
structural differences of the soils among slip mass,
Li Xingsheng & Li Tonglu. 1997. The characteristics and
slip zone and slip bed are apparent. the physi- analysis of loess landslide in China. International Sympo-
cal indexes as water content, liquid limit, plastic sium on Landslide Hazard Assessment. 361–366.
indexes and liquid index shows higher values in Sun Janzhong. 2005. Loessology. Hong Kong Archaeologi-
the slip zone while plastic limit and mean particle cal Society Press, Hong Kong.

441
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Testing study on the strength and deformation characteristics of soil

in loess landslides

H.J. Liao
Department of Civil Engineering, Xi’an Jiaotong University, Xi’an, China

L.J. Su
School of Civil Engineering, Xi’an University of Architecture and Technology, Xi’an, China

Z.D. Li & Y.B. Pan

Department of Civil Engineering, Xi’an Jiaotong University, Xi’an, China

H. Fukuoka
Disaster Prevention Research Institute, Kyoto University, Uji, Kyoto, Japan

ABSTRACT: Landslides groups in southern highland area of Jingyang are investigated and analyzed. Char-
acteristics of and reasons for the landslides were statistically analyzed. A case history of a large-scale landslide
in loess in Dongfeng village of Gaozhuang town in Jingyang is studied. The displacement, maximum width,
circumference and volume of the slipping mass were obtained through in situ measurement. Laboratory tests
were carried out on soil samples taken from a depth of 2 m below the front edge of the slipping mass. Basic
property and shear strength parameters of the loess soil were obtained. These results will benefit further studies
on the kinetic mechanism of loess landslide in southern highland area of Jingyang.

1 INTRODUCTION geological disasters, especially the loess landslides,

will be encountered inevitably with the Development
There are frequently occurred geological disasters in of the West Regions. The loess landslide disasters
the vast territory of China. Every year, landslides, are one of the major landslides in special soil (Wu
debris flows, ground collapses and etc. bring enor- et al. 2002). According to incomplete statistics from
mous economic loss and casualties, especially the 1983∼1993, about 200 loess landslides had collapsed
landslides (Fig. 1) (Ai, Z.X. 2005). In China, loess only in Guanzhong area of Shaanxi, the direct eco-
is widely distributed, especially in the northwest area, nomic loss caused by which is more than 100,000,000
where the loess is the most representative. Various yuan (Lei, X.Y. 1996). There are over 1000 active
landslides at present, which bring serious threats to
human’s property and life.
The loess is widely distributed in China and land-
slides in loess are very typical. It is of great importance
for preventing landslide to study its strength deforma-
tion characteristics and the mechanism of its formation
and movement by both theoretical and testing methods.
In the southern highland area in Jingyang, Shaanxi,
landslide and collapse develop seriously and bring
severe threats to human’s life and property. Based
on a loess landslide in the southern highland area in
Ground fissures Ground collapses Jingyang, the causes of landslide were analyzed and
Debris flows Dilapidation Landslides physical and mechanical tests were performed and ana-
lyzed aiming at providing some information for further
Figure 1. Statistics of geological dissasters occurred in study on the mechanism of loess landslide occurred in
2004. this area.

443
2 IN SITU INVESTIGATION AND ANALYSIS 100

Slope angle (degree)

OF LANDSLIDE GROUP IN SOUTHERN
HIGHLAND AREA IN JINGYANG 80
60
The southern highland area in Jingyang is situated to
the south of Jingyang County and on the south bank 40
of Jinghe River. The area of the southern highland 20
is 180 km2 which is 23.1% of the whole county’s.
The elevation of the highland is 430∼500 m. Steep 0
Location
slopes eroded by Jinghe River connect the floodplain Gaozhuang Jiangliu Zhaitou
at the edge of the southern highland area in Jingyang.
The steep slopes on the edge of the highland provide 100
gravity conditions for landslides. Therefore landslides

Slope height (m)

with different scale are often evoked by external fac- 80
tors such as frequent irrigating activities, rainfall and 60
earthquake, etc.
After Weihe River was channelled towards large 40
areas in the southern highland area in Jingyang for
20
irrigation in 1976, landslides occurred more and more
frequently. More than 40 landslides had been recorded 0
by 2004, including 7 huge landslides and several col- Location
lapses. More than 1.60 × 107 m3 of soil and rock Gaozhuang Jiangliu Zhaitou
cascaded down, 29 were killed and 27 injured, 140 hm2
250
of farmland was damaged, over 10 houses and nearly
200 cave-houses were demolished, and 100 cattle were 200
killed. The total direct economic loss reached about
150
3.00 × 106 yuan (Wang et al. 2004).
The surface area of the southern highland in 100
Jingyang is quite broad. It is higher in the northwest 50
lower in the southeast. The surface of the highland is
30∼90 m higher than the floodplain of Jinghe River. 0
The angles of the slopes lie between 45◦ ∼80◦ (Lei,
Location
Gaozhuang Jiangliu Zhaitou
X.Y. 1995).
Taking Gaozhuang town as the starting point, varia-
Figure 2. Statistics of landslides from east to west along the
tions of the angle, height and volume of some previous southern highland area in Jingyang.
occurred landslides were statistically analyzed along
the southern highland area from east to west. The
results are shown in Figure 2. Lines shown in the figure
are exponential trend lines of the measured data. It is of the landslides group in southern highland area of
obvious that the angle, height and volume of the land- Jingyang are as following:
slides tend to increase from east to west. Especially
the volume increases very obviously, which is related 1. Along the boundary of the highland, the landslides
to the local geological and topographical conditions. group is distributed like a belt due to the effect of
According to the literature and in situ investiga- topographical characteristics and erosion of Jinghe
tion, the frequency occurrence of landslides is closely River.
related to the structure of loess layer. The loess in 2. The landslide blocks moved fast and could reach
southern highland in Jingyang is loose in which verti- large displacements. According to the author’s
cal joints developed. It is with small cohesion and large inspection, the largest displacement of the slope
permeability, and is easy to be softened by water. The in Dongfeng village of Gaozhuang town may reach
main component of the slopes is paleosol or loess in 300 m after it slipped from its toe.
late and middle Pleistocene. When the soil is dry, it 3. In southern highland area of Jingyang, the slope
is dense and with a large cohesion and high shearing stability is low. Old landslides might revive and
strength. When it is infiltrated by water, the cohesion new landslides might mobilize behind the wall of
and shear strength will decrease sharply and the loess the old landslides.
will become soft or flowing plastic form which forms 4. The sliding mass often cuts the loess layer deeply
slip planes of landslides. The overall characteristics and slides out from paleosol at the toe of the slope.

444
5. Rainfall could evoke landslides. Landslides in 3 LABORATORY TESTING STUDY
southern highland area of Jingyang often occurred
during or after rainfalls. 3.1 Basic properties and collapsibility analysis
According to the in situ inspection, the typical char- Basic property tests were conducted on soil samples
acteristics of the development of loess landslides in taken from a depth of 2 m below the front edge of the
southern highland area of Jingyang are as follows: slipping mass of the landslide in Dongfeng village.
most of them maintain a shape like a round-backed The results are shown in Table 1.
armchair; the wall of the steep slope is almost ver- Figure 4 shows the results from the particle size dis-
tical; the sliding mass moves along the floodplain tribution test. The percent of particles with diameters
after it slides out from the slope toe. The sliding mass larger than 0.075 mm is 27.8%, which is smaller than
looks like a tongue and undulant terrain is formed by 50%. The percent of particles with diameters smaller
alternately arranged drumlins and depressions. than 0.005 mm is larger than 10%. The plasticity Index
A case history of a large-scale loess landslide in Ip is between 3 and 10. Therefore the soil can be
Dongfeng village of Gaozhuang town in Jingyang is classified as loess-like clayey silt. The characteristic
studied in this paper (Fig. 3). The landslide occurred of thixotropy and liquefaction of silt is an important
at 4:40 AM on July 23, 2003. A sliding mass of factor that causes the landslide group.
53,360 m2 moved along the floodplain of Jinghe River. Collapsibility is one of the typical characteristics
In situ inspection of the landslide was conducted in of loess. Double-line collapsibility test was carried
April 2006. Undisturbed soil sample was taken in order out on soil samples taken from the southern highland
to perform laboratory test. area in Jingyang and the results are shown in Figure 5.
Based on the measured data, characteristic of the Based on the initial stress level at the depth where the
landslide can be obtained. The displacement of the soil samples were taken, the calculated coefficient of
slipping mass along the floodplain of Jinghe River is collapsibility is 0.03 and the soil can be classified as
more than 300 m. The maximum width is 400 m. slightly collapsible loess. It is observed that most of
The surface area covered by the slipping mass is the cracks are due to collapse of the loess. The cracks
about 106,421 m2 . The circumference and volume are located at edge of the depression area caused by
of the sliding mass are 1192 m and 1.28×106 m3 collapse of the loess. Depression area was found on
respectively. farmland a few days before the landslide occurred in
Dongfeng village in Jingyang.

3.2 Results from direct shear tests

Both unconsolidated and consolidated direct shear
tests were performed under four different vertical pres-
sures of 100 kPa, 200 kPa, 300 kPa and 400 kPa.
These two types of tests are suitable for clay whose
coefficient of permeability is smaller than 10−6 cm/s.
Results of the direct shear tests are shown in Figure 6
and Figure 7. The cohesion and internal friction angle
from unconsolidated direct shear tests are 35.74 kPa
and 33.8◦ respectively while those from consolidated
direct shear tests are 85.77 kPa and 24.3◦ respec-
tively. The shear strength parameters obtained from
consolidated tests are larger than those obtained from
unconsolidated tests, which is in accord with the
Figure 3. A photo of the landslide in Dongfeng village. fact.

Table 1. Basic property test results.

Coefficient Coefficient
Water content Density Liquid limit Plastic limit Specific gravity of collapsibility of compression
wn (%) ρ (g/cm3 ) wL wp ds δs a1−2 (MPa−1 )

6.40 1.80 22 13 2.72 0.03 0.155

445
 100 300
Percentage finner (%)

Shear strength (kPa)

80 250
60 200
40 150
20 100
0 50
0.001 0.010 0.100 1.000 10.000
Particle size (mm) 0
0 100 200 300 400 500
Figure 4. Particle size distribution results. Vertical stress (kPa)

Figure 7. Results of consolidated direct shear test.

20.50
20.00 After soaking
200 =50kPa

Deviator stress (kPa)

19.50 Before soaking 3
h (mm)

3
=100kPa
19.00 160 =150kPa
3
18.50 120 3
=200kPa
18.00 80
17.50
40
17.00
0 200 400 600 800 1000 0
p (kPa) 0 5 10 15 20
Axial strain (%)
Figure 5. Curves of the collapsibility test.
Figure 8. Deviator stress-axial strain curves.

400
Shear strength (kPa)

160
Porewater pressure (kPa)

300 140 3

120 3

200 100 3
3
80
100 60
40
0
0 200 400 600 20
Vertical stress (kPa) 0
0 5 10 15 20
Axial strain (%)
Figure 6. Results of unconsolidated direct shear test.
Figure 9. Porewater pressure-axial strain curves.

3.3 Results from triaxial tests

Consolidated undrained (CU) tests were carried out on
undisturbed soil samples under confining pressures of
50 kPa, 100 kPa, 150 kPa and 200 kPa. The results are
shown in Figures 8 to 10.
During shearing, strain softening was observed
after the deviator stresses reached their peaks at the
axial strain of 2%∼4%. The cohesion and internal
friction angle obtained from the failure envelope are
39.54 kPa and 10.4◦ respectively. Figure 10. Failure envelope of the CU tests.

446
 The obtained shear strength parameters are smaller ACKNOWLEDGEMENT
than those obtained from consolidated undrained
direct shear tests. The reason for this is that the shear- The research is financially supported by Nature Sci-
ing plane is fixed between the lower and upper box ence Foundation of China (50379043). The authors
in direct shear tests but the weakest plane may not be would like to acknowledge the ‘‘FY2006 JSPS Invi-
this fixed plane. The shearing plane in triaxial tests is tation Fellowship Program for Research in Japan’’ for
not fixed and the water drainage is strictly controlled. its support.
Therefore the results from triaxial tests are more close
to the theoretical value (Yan, L.F. 2004). The shear
strength parameters of loess are quite scattered so that REFERENCES
different results would be obtained from different type
of tests. The shear strength parameters obtained from Ai, Z.X. 2005. Analysis of landslide evoked by rainfall.
the above tests are close to the actual value of loess Disaster and Prevention Engineering (2): 9–11.
(Wang, J.D. 1999). Lei, X.Y. 1995. The hazards of loess landslides in the
southern tableland of Jingyang county, Shaanxi and their
relationship with the channel water into fields. Journal of
Engineering Geology 3(1): 56–64.
4 CONCLUSIONS Lei, X.Y. 1996. The study of loess landslide type caused
by human activities in Guanzhong of Shaan’xi province.
The landslide groups in southern highland area of Hydrology Geology and Engineering Geology (3): 36–39.
Jingyang were evoked by both internal and exter- Wang, D.Y., Du, Z.C. & Zhang, M.S. 2004. Geological haz-
nal factors. The internal factors include new tectonic ard of cliff collapse, landslide and their occurrence in
movement, topography and geography, lithology and southern Jingyang county of Shaanxi province. Bulletin
underground hydrology conditions etc. The external of Soil and Water Conservation 24(4): 34–37.
factors include irrigation, excavation, and rainfall etc. Wang, J.D. 1999. The systematic geological research of typi-
cal high-speed loess landslide groups. Chengdu: Sichuan
Based on the large-scale loess landslide groups in Science Press.
Dongfeng village of Gaozhuang town, in situ observa- Wu, W.J. & Wang, N.Q. 2002. Basic types and active features
tion and a series of laboratory tests have been carried of loess landslide. The Chinese Journal of Geological
out. Stress-strain curves and shear strength parameters Hazard and Control 13(2): 36–40.
of the loess were obtained. These results will bene- Yan, L.F. 2004. Comparison of contrast between direct box
fit further studies on the kinetic mechanism of loess shear test and triaxial test. Shanxi Architecture 30(24):
landslide in southern highland area of Jingyang. 64–65.

447
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Failure mechanism of slipping zone soil of the Qiangjiangping landslide

in the Three Gorges reservoir area: A study based on Dead Load test

Xianqi Luo & Ailan Che

School of Naval Architecture, Ocean and Civil Engineering, Shanghai Jiao Tong University, China

Ling Cao & Yuhua Lang

China Three Gorges University Key Laboratory of Geological Hazards on Three Gorges Reservoir Area,
Ministry of Education, China

ABSTRACT: The Qiangjiangping landslide occurred during the rainy season after the first impoundment of
the Three Gorges Reservoir up to 135 m. Rainfall and reservoir water are considered as two main inducements
for its failure. The failure mechanism of the Qiangjiangping landslide induced by these two factors was explored
through Dead Load test (DL test). In the test, consolidation pressures σ1 and σ3 were kept unaltered to simulate
the deformation behavior and strength characteristics of the landside slipping zone soil while the matrix suction
decreasing.

1 INTRODUCTION And the attitude of stratum was chiefly stable; its vari-
able scope was 110◦ 1̃50◦ ∠15◦ 3̃0◦ . There were many
The Qianjiangping landslide took place in Shazhenxi faultages, fissures and soft interlayer in this region,
Town, Zigui County, Hubei Province, PRC at 0:20 which worked as the intrinsic condition and material
a.m. on July 13, 2003. 129 shelters had been dam- basis for forming the landslide.
aged; 1200 men had left homeless; 14 died and 10 The impoundment of Three Gorges Reservoir has
were missing till the morning of July 20. This land- been planed to carry through in three stages of time: the
slide caused a traffic halt, collapse of shelters and first stage starts from June 1, 2003, when the reservoir
workshops, block of the Qinggan River and enormous water level is 67 m, to June 15, 2003, the water level
economic loss. Four enterprises, a silicon company,
a brick-fabricating factory, a transportation company,
and an architecture and construction company, were
destroyed. The rush of the landslide fiercely struck
the lower part of Shazhenxi Town with huge waves.
Some anchors anchored on the dock were overturned;
the vegetation partially demolished; the use value of
the land debased.
The Qianjiangping Landslide locates on a slope
with southeast trend along the left bank of Qinggan
River, which is in Qiangjiangping Village, Shazhenxi
Town, Zigui County. The angle of the slope is 35◦ 1̃5◦ ,
from the top to the bottom. The landslide develops
in clastic sedimentary rocks of lower-middle Jurassic
Niejiashan formation. In terms of lithology, it consists
of medium and thick layer siltstone, silty mudstone
and shale. The incline of the terrane is roughly iden-
tical with that of the slope, its angle is 35◦ 1̃9◦ , from
the top to the bottom, i.e., it is a dip slope. Owing
to the influences by regional structure, soft interlayer
in the dislocation-mudding zone was more developed. Figure 1. Location of the Qianjiangping landslide.

449
then rises up to 135 m; the second stage lasts from
September 22, 2006 to October 27, the reservoir water
level rises from 135 m up to 156 m; the last stage will
begin in the year 2009, the water level when finished
will go up to 175 m.
The Qiangjiangping landslide is a dip rocky-slope;
its geomorphic features, geologic constitution and
structure, and physical composition are quite typical in
the Three Gorges reservoir area. Features of dip slope
and that of stratum structure with soft and hard rock
layer spaced in-between are seen on the location of the
Qiangjiangping landslide. They provide physical basis
for new failures of reservoir landslide. The landslide
occurred during the rainy season after the first stage
of impoundment of the Three Gorges reservoir, it is
natural to associate its potential inducement with the Figure 2b. View of the Qianjiangping landslide from the
rainfall and reservoir water. It is definite that these two upstream side of the Qingganhe River.
factors are quite common in the Three Gorges Reser-
voir area, so that it is significant important for later
landslide controlling and forecasting in the area.

2 COMPOSITION OF THE QIANGJIANGPING

LANDSLIDE SLIPPING ZONE SOIL

The Qiangjiangping landslide is a dip slope with soft

and hard stratum layers spaced in-between. Carbona-
ceous shale is merely appeared in one of the layers
within the extent of the landslide, which thickness is
20–30 cm. It is the primary intercalation of the land-
slide. Interlamination stick slip during the course of
anaphase tectogenesis rebuilds the layer containing
carbonaceous shale, and then forms an intercalated
crushed dislocation-mudding zone, which worked as
the slipping zone for the Qiangjiangping landslide
during the rainy season and after the first stage of
impoundment of Three Gorges Reservoir. Figure 3. Picture for the intercalated crushed dislocation-
Such typical slipping zone soil with quite clear dis- mudding zone.
tinctions are found in the Qiangjiangping landslide

Figure 4. Particle-size distribution for the soil used in DL

Figure 2a. Overlook of the Qiangjiangping landslide. test.

450
Table 1. The average index value of its physical property.

Wet density/dry Natural water Specific Void Saturation Liquid Plastic

density (g/cm)3 content (%) gravity ratio (%) limit (%) limit (%)

2.03/1.79 13.5 2.70 0.52 70.2 34.7 18.3

in prospecting drift wall mapping, borehole core and Table 2. Physic parameters of the Qiangjiangping Land-
field surface outcrop. Since the slipping zone soil mass slide slipping zone soil.
may be affected, more or less, by water and slide
movement, samples for test were collected from No.1 Unit Water Dry
weight Specific content density Void
prospecting drift wall mapping in the middle part of
(kN/m3 ) gravity (%) (g/cm3 ) ratio
hauling area of the landside mass, where most typical
slipping zone soil were found (Figure 2). 19.6 2.70 19.6 1.67 0.57
The intercalated crushed dislocation-mudding zone
resulted from partial mylonitization of broken
porphyry. Its protolithic composition is described: car- 3.2 Test instrument
bonaceous shale and limestone banding/conglome-
ration spaced in-between. Cataclasite distributed on The instrument used for this test was GDS. Unsat-
the dislocation surface. Its terrane was extruded into urated Triaxial Testing System (GDS UNSAT) pro-
flakiness. Agglutination was found between calcare- duced by GDS in UK. The biggest strongpoint of this
ousness and mudding, and weathering was noticeable. instrument stays with its preciseness in measuring the
Steps and longitudinal scratches were often found. The volume of samples; and also their volume changing
granulometric composition obtained from particle size can be read at any moment during the test by survey-
analysis on slipping zone soil sample was as follows: ing, continuously and automatically, the water-level
particles with its diameter >2 mm count 49 %∼53 %; discrepancy between that in the inner cell and that
particles with its diameter 2∼0.075 mm, 43 %∼47 %. in the reference tube through a sensitive differential
Powder cosmids with its diameter <0.075 mm, is only pressure transducer (Zhang and NG 2006).
contained about 4%. And Figure 4 is the particle-size
distribution for the soil used in DL test. The aver-
3.3 Testing program
age index value of its physical property is shown as
table 1. The test is conducted through 4 individual groups
in accordance with different suctions; they are 0 kPa,
50 kPa, 100 kPa and 200 kPa, respectively. In each
group, there are 3 samples, which net confining pres-
3 CONSOLIDATED-UNDRAINED TEST
sure (σ3 − ua ) are, 50 kPa, 100 kPa, and 150 kPa,
(CU TEST) OF THE QIANGJIANGPING
respectively.
LANDSLIDE SLIPPING ZONE SOIL
Mohr circles under the same suction are put in the
same group; the common tangent of these circles is
Generally, the Qiangjiangping landslide slipping zone
its failure envelope. And the intercept and slope of
soil is unsaturated, which shear strength usually get
the failure envelope are respectively c and ϕ under
from CU test where suction and net confining pressure
different suction (Table 3).
are under control simultaneously.
From Table 3, it is noticed that the value of φ has
no significant changes within the range of suction
(0∼200 kPa) given in the test. When the given value of
3.1 Sample preparation
suction is 0, it was actually a triaxial test of saturated
Before the test, the soils collected were dried and soil, therefore it is φ  = 26.6◦ .
scrunched, and then screened through with a 2 mm Table 2 demonstrates principally a linear relation-
screen. Then it was made into wet soil with reference ship between gross cohesion C and suction ua − uw .
to the liquid limit water content. After 24 hours’ moist Its linear equation is: c = c + (ua − uw )tgφ b . Where,
curing, which served to homogenize water content in c is the intercept of Mohr-Coulomb envelope surface
the soil, it was chopped into remodeling clay soil sam- and shear stress axis, when both matrix suction and
ples, with 100 mm height, and 50 mm diameter. The net normal stress are 0, that is, the intercept of gross
proportion, weight, natural water content, dry density cohesion. From the equation, we get: c = 15.7 kPa,
and void ratio were measured respectively (Table 2). φ b = 22◦ .

451
Table 3. c and ϕ of the Qiangjiangping landslide slipping
zone soil under different pressures.

Suction (kPa) 0 50 100 200

c (kPa) 10 40.3 59.1 94

φ (◦) 26.6 27.2 27.5 28.2

4 DEAD LOAD TEST (DL TEST)

OF THE QIANGJIANGPING LANDSLIDE
SLIPPING ZONE SOIL
Figure 5. Stress path of the DL test.
During the action of rainfall and reservoir water,
the shear breakdown to soil mass owes chiefly to the
increase in pore water pressure or, in other words, the
to results in the CU test for the value of ϕ  , namely,
decrease in matrix suction. According to Brand (1981)
ϕ  = 26.6◦ k0 = 0.55). After the stabilization of
and Chen et al (2004), the decrease in matrix suction
anisotropic consolidation, measure the gross volume
will cause debasing effective stress on the potential
deformation, axial deformation, lateral deformation
slip zone of the slope, where the shear strength will
and volume change of the specimen, and calculate the
be weaken down. Anisotropic triaxial test can simu-
water content of the soil mass. The increasing rate of
late the process in which the pore water pressure rises
pressure and the criterion to stop the application of
from a negative value till the failure of soil mass. If
suction is referenced (Zhan 2003; Sivakumar 1993;
σ1 andσ3 remain unaltered, unsaturated soil test can
Yang 2005; Fredlund and Rahardjo 1997).
be a relatively accurate simulation to the failure stress
From Geotechnical Test Regulations (SL237–1999)
path of the unsaturated-soil slope when couple affected
for the criterion to stop the consolidation stage during
by the rainfall and reservoir water, and therefore the
triaxial test, the tonnage is less than 1 mm3 within 2
deformation features of the slope soil mass could be
hours.
examined.
After the stabilization of consolidation, ua remains
Dead Load test of unsaturated soil is also able to
unaltered, uw increases at each phase, in other words,
simulate the influences upon the strength and volu-
the matrix suction gradually decreases. During such
metric strain behavior of the soil mass while the matrix
a process, the pressure of anisotropic consolidation
suction is decreasing during the process of rainfall and
keeps unaltered. It is meaning that gross stress σ1 , σ3
rising level of reservoir water.
and deviatoric stress q = σ1 − σ3 , as Dead Load, con-
stantly acts on the samples. Shear deformation occurs
along with the decreasing of matrix suction. While the
4.1 Sample preparation, test instrument samples are reaching shear, use the strain controlling
and testing program mode within ‘‘4D UNSATURATED’’module of GDS
Sample preparation and test instrument are identical till axial strain εa reaches its maximum value, which
with that in the CU test. And the testing program is as can be regarded as the sheer failure of the samples.
follows:
Figure 5 shows the stress path of the DL test. The
4.2 Results and discussions
test consists of four stages: saturation and consolida-
tion, moisture dehydration, anisotropic consolidation (1) The stage of moisture dehydration
and suction decrease. The next is our test operations. The backpressure and consolidation is given as the
At first, put samples into the pressure chamber, initial state of this stage, which is when the tonnage
and give counter pressures to saturate and consolidate and gross volume deformation is measured after the
them. Then take p−ua = p0 under control to make sure consolidation; the water content was calculated to
its value keeps unaltered. When the pressures given 18.50%. The suction load was increased at each indi-
from each direction are identical, gradually rises the vidual phase, which target value is 20, 60, 100 kPa,
matrix suction from 0 to a certain value. Anisotropic respectively.
consolidation begins after the suction becomes stable. Figure 6a shows the variation in the water content
The top of each sample are made connected with the during the stage of moisture dehydration. It divided
probe of the axial pressure sensor while anisotropic itself into three phases noticeably. With target suc-
consolidation σ1 = σ3 /k0 take place. Here coeffi- tion increasing and test time lasting, the water content
cient of at-rest earth pressure k0 = 1 − sin ϕ  (refer decreased more and more sharply. At the first phase,

452
when the suction ua − uw = 20 kPa is stable, the Where, vd is the sample’s volume when the suction
water content is get asω = 17.66%. At the second is stable in the end of the stage of moisture dehydration.
phase, when the suction is stable, the water con- The volume shrinkage in this test is 3.35%.
tent is get as ω = 16.02%, and at the third phase, Moreover, the relationship between specific volume
ua − uw = 100 kPa, ω = 14.95%. (v = 1 + e) and its corresponding suction value is also
Figure 6b shows the volume variations of the sample found, while the suction getting stable in each of the
in the process of moisture dehydration. Similar to the three phases during the stage of moisture dehydration.
variations in water content, the volume variation curve Figure 7 shows the relationship between the specific
may also be divided into three phases. The rate of volume and ln [(s + pat ) /pat ] (where s is the value of
volume change in each phase presents the same trend: matrix suction pat is referential pressure, namely, a
gradually getting slower till down to a certain value. standard atmospheric pressure). It is shown that the
The volume decreases in the whole stage. Here its specific volume and suction (within the range from
absolute shrinkage ηd is defined as: 0 to 100 kPa) is basically of a liner relation, which can
be described as the following formula (Alonso 1998):
 
|vd − v0 | s + pat
ηd = × 100% v = vs − ks ln
v0 pat

where vs and ks are respectively the intercept and the

slope of the relation curve v − ln [(s + pat ) /pat ] when
the value of suction is 0, where ks = 0.07, vs = 1.64.
This result does not testify a suction increase sug-
gested by Gens (1992) & Alonso (1990) for unsat-
urated soil. The reason may be considered as that
the maximum value of suction we discuss here, i.e.,
100 kPa is not big enough to reach its yield point to
suction. The suction in this test has not gone beyond
the elastic stage of soil mass deformation.
(2) The stage of anisotropic consolidation
When the suction becomes stable, (ua − uw ) =
100 kPa, ua = 150 kPa and uw = 50 kPa; then con-
duct anisotropic consolidation ( p = (σ1 + 2σ3 )/3,
σ3 = k0 σ1 ). The stable state after the first stage is taken
as the initial state for the stage of anisotropic consoli-
Figure 6a. Variation in water content during the stage of dation. At the end of axial loading, σ1 −ua = 270 kPa,
moisture absorption. σ3 − ua = 80 kPa, axial deformation of the sample
reaches 5.19%. Figure 8 shows the relation curve of
εa − p (axial deformation-average net normal stress).
A distinct yield point can be found in the compres-
sion curve when the suction is 100 kPa. The average
net normal stress at the yield point is 114.6 kPa, and the
void ratio e is 0.561. After the yield point, the compres-
sion coefficient is αv = 0.009 kPa−1 ; this data enables
to calculate slope deformation in the latter stages.

Figure 6b. Variation in volume during the stage of moisture

absorption. Figure 7. The volume-suction relation curve.

453
 Figure 9a. Axial strain-matrix suction curve.
Figure 8. Compression curve during anisotropic consolida-
tion.

(3) Suction down to the failure stage

After the stabilization of anisotropic consolidation,
DL test is conducted to the stress routine of decreasing
matrix suction to the testing slope receiving actions
of rainfall and rising reservoir water. That means
anisotropic consolidation σ1 , σ3 remains unaltered,
σ1 = 420 kPa, σ3 = 230 kPa, and deviatoric stress
q = σ1 − σ3 , as Dead Load, continuously acts on the
sample. When decrease gradually the matrix suction
(100 kPa–80 kPa–60 kPa–40 kPa–30 kPa–20 kPa–
5 kPa), the deformation of the sample increase grad- Figure 9b. Volume change-matrix suction curve.
ually with it. Similarly, the stable state of anisotropic
consolidation is taken as the initial state for this stage.
Figures 9a, 9b demonstrate the relationship between
axial strain and matrix suction, between volume τf = c + (σ − ua )f tgφ  + (ua − uw )f tgφ b and
change and matrix suction. From Figure 10b it shows φ  = 26.6◦ , φ b = 22◦ , c = 15.7 kPa,
that unsaturated soil in the Qiangjiangping Landslide
demonstrates dilatability in the course of moisture
absorption. It can be explained as the influences upon suppose stress path would not affect shear strength
deformation of unsaturated soil by pore water and of the soil mass, to substitute the known values of
pore gas. For triaxial test to the route of moisture φ  , φ b and c in the above formula, then the shearing
hydroscopic, during the process of unsaturated soil stress on the shear failure surface of the soil mass is
consolidation, i.e., the process when ambient pressure, τ = 86.96 kPa, shear strength is τf = 84.05 kPa, when
back pressure and atmospheric pressure in various σ1 −ua = 270 kPa, σ3 −ua = 80 kPa, ua −uw = 5 kPa.
phase are reaching its balanced, macroscopic pores It is suggested that moisture hydroscopic of the soil
crush and die away; and in the shear process, water in mass caused by the rainfall and rising reservoir water
microscopic pores redistribute and thus cause the soil (increasing pore water pressure and decreasing matrix
mass to dilate. suction ua − uw ) would lead to the failure of soil mass.
When matrix suction ua − uw finally down to the
value of 5 kPa and the axial strain εa = 12.6%, it is
shown as Figure 10a that the axial strain-matrix suction
curve εa −(ua − uw ) get to its peak value and becomes 5 CONCLUSION
even. It is meaning that the soil mass has come into the
state of failure. Then the stress condition is as follow: The main results are as follows:

ua − uw = 5 kPa σ1 − ua = 270 kPa 1. From the DL test, the following data of the
Qiangjiangping landslide slipping zone soil are
σ3 − ua = 80 kPa obtained: in case of the net ambient pressure
p − ua = 80 kPa, water content variations and
volumetric behavior during the stage of mois-
Compared these results with ones obtained from the ture dehydration; through its volumetric behavior,
unsaturated soil CU test concerning strength parame- liner relationship between specific volume and
ters, in the CU test, suction is:

454
  
s + pat REFERENCES
v = vs − ks ln
pat
  Alonso E.E., 1998. Modeling expansive soil behavior,
s + pat
= 1.6448–0.0751 Proc of the 2nd International Conference of Unsaturated
pat Soils.Beijing.
Alonso E.E., GENS A. & JOSA A., 1990. A constitutive
2. A distinct yield point has been found in the com- model for partially saturated soils. Geotechnique, 40 (3):
pression curve when the suction is set as100 405–430.
kPa during the stage of anisotropic consolidation. Brand, E.W., 1981. Some thoughts on rain-induced slope
failures, Proc., 10th Int. Conf. on Soil and Mechani-
Across the yield point, the compression coefficient
cal Foundation Engineering, Stockholm, Sweden, Vol. 3,
is as follow: αv = 0.009 kPa−1 . 373–376.
3. Dilatancy has been demonstrated in the shear pro- Chen H., Lee C.F. & Law K.T., 2004. Causative Mecha-
cess of the Qiangjiangping landslide unsaturated nisms of Rainfall-Induced Fill Slope Failures, Journal of
slipping zone soil. The pore water and pore gas have Geotechnical and Geoenvironmrntal Engineering. 130(6):
influences upon the deformation behavior of the 593–602.
soil mass. And during the stage of unsaturated soil Fredlund D.G. & Rahardjo H., 1997. Unsaturated Soil
consolidation, i.e., the process when ambient pres- Mechanics.
sure, back pressure and atmospheric pressure in Gens A. & Alonso E.E., 1992. A framework for the behavior
of unsaturated expansive clays, Canadian Geotechnical
various phase are reached its balance, macroscopic
Journal, 29: 1013–1032.
pores crush and die away; and in the shear process, Sivakumar V., 1993. A critical state framework for unsatu-
water in microscopic pores redistribute and thus rated soils, Sheffield: University of Sheffield.
cause the soil mass to dilate. Yang Heping & Xiao Duo, 2005. The Influence of Alter-
4. After anisotropic consolidation, keep its net stress nate Dry-wet Effect on the Strength Characteristic of
p−ua and deviatoric stress q unaltered, when matrix Expansive Soils, Journal of Hunan Light Industry College
suction ua − uw finally down to 5 kPa and the axial (natural science), 2 (2): 1–5.
strain εa = 12.6%, the axial strain-matrix suc- Zhan Liang tong, 2003. Field and laboratory study of an
tion curve εa − (ua − uw ) get to its peak value and unsaturated expansive soil associated with rain-induced
slope instability. Hong Kong: The Hong Kong University
becomes even, the soil mass has come into the state
of Science and Technology.
of failure. To substitute the known data from the Zhang Liang tong & Ng, C. W.W., 2006. The Triaxial
CU test, φ  = 26.6◦ , φ b = 22◦ , c = 15.7 kPa, the Test Study on Strength and Deformation of Unsatu-
following parameters are gotten: τ = 86.96 kPa, rated Expansive Soils, Chinese Jounal of Geotechnical
τf = 84.05 kPa. Engineering, 28 (2): 196–201.

455
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Post-failure movements of a large slow rock slide in schist near Pos Selim,
Malaysia

A.W. Malone
Dept of Earth Sciences, University of Hong Kong, China

A. Hansen
Dept of Spatial Sciences, Curtin University of Technology, Australia

S.R. Hencher
Halcrow China Ltd., Hong Kong; School of Earth & Environment, University of Leeds, UK

C.J.N. Fletcher
Dept of Building & Construction, Hong Kong City University, China

ABSTRACT: This paper describes the results of the monitoring, by total station and photogrammetric surveys,
of the movements of a slow compound rock slide from failure in 2003 to December 2006. During this period
the head moved downwards more than 21 m. Whilst the rate of displacement is declining slightly year on year,
for much of the time the landslide mass is accelerating and then decelerating in surges. Evidence is presented
of some correspondence between the timing of the surges and the seasonal rainfall pattern. It is inferred from
surface observations that the failure involves sliding at the head and in the upper main body of the landslide on
joints roughly orthogonal to the foliation, which dips at a shallow angle into the slope. In the central toe zone the
landslide slides up and out on the foliation. The failure, which occupies an area of about 8.5 ha, has reactivated
major pre-existing faults which run obliquely through the landslide mass.

1 INTRODUCTION 2 TOPOGRAPHY, GEOLOGY AND THE

LANDSLIDE
A landslide occurred in September 2003 during hill-
side excavation for a new strategic road in mountain- The site is on the western hillside of the Gunung Pass
ous terrain near the Cameron Highlands hill resort ridge which reaches an elevation of 1587 m above sea
in northern Peninsular Malaysia. The site is on the level. Prior to cutting, the valley sides were densely
Simpang Pulai—Lojing Highway, close to Longitude forested and generally steeper than 30◦ , with ravines
101◦ 20 43 Latitude 4◦ 35 27 (Figure 1). Road- leading down to the deeply-incised River Penoh some
works commenced in 1997 and movements occurred 600 m below the ridge.
in roadside cut in the vicinity of chainage 23+900. The cuts were formed largely in nominally 12 m
The slope was cut back to a flatter angle but insta- high 1:1 batters and 2 m berms to produce an exca-
bility persisted. Progressively more extensive slope vation up to 260 m high and inclined about 33◦ –35◦
flattening was undertaken in response to continu- overall.
ing failure until the works reached the ridgeline, The geology of the Gunung Pass area consists
200 m to 260 m above the road. Gross movements of a sequence of sedimentary rocks, probably of
occurred in the cut in September 2003 with the for- Paleozoic age, which have undergone low- to medium-
mation of a main scarp and associated disruption grade dynamic metamorphism. The metasedimentary
and the displaced mass has since moved continu- rocks outcrop in a 4 km-wide shear zone contained
ously. A study of the landslide was carried out by within Mesozoic granites. The landslide has taken
the authors in 2005 and 2006 (Andrew Malone Ltd, place in quartz mica schists (Figures 1 & 2) which
2007). at the base contain impersistent graphite schist layers

457
 joints
X
face foliation
35/245
Y
toe zone
GS
-6/246

QM S
joints
head
63/238.5

100m natio nalg rid

equal angle, lower hemisphere, 541 poles
Shading represents concentration of poles (5% maximum).
QMS Quartz Mica Schist Unit GS Graphite Schist Unit The circles X & Y are centred on the mean disposition of vec-
ABC geological faults; T1 T2 T3 transverse cracks
1P1, 2A etc total station survey markers tors within 100 m wide blocks of ground (on the centreline) at
points used in the photogrammetric adjustment the highest part of the head (X) and at the toe zone
+ survey check points survey targets (Y) and enclose >80% of the vector data points.

Figure 1. Outline of the landslide superimposed on a Figure 3. Stereographic projection of poles to joint planes
simplified geological map of the site; and survey points. and foliation and surface displacement vectors.

show signs of recent but pre-landslide movement and

have been reactivated during landslide movements.
The schists at outcrop are highly jointed with typical
joint spacing less than 0.5 m (Figure 2). The poles of
the joints form a girdle that is roughly orthogonal to the
low-dipping foliation (Figure 3). Where unweathered,
the schist is generally strong to very strong and most of
the rock material currently exposed across the site can
be classified as ‘slightly’ to ‘moderately weathered’
(British Standards Institution, 1999: Figure 19).
300mm The main surface features of the landslide are
the main scarp, the head graben, the north and
south flanks, the counterscarps of the oblique faults
(A, B & C etc) and a low-angle push-out structure at
Figure 2. Several joint sets cutting the low-dipping folia- the toe zone (Figure 1). Neither the northern flank,
tion within the Quartz Mica Schist Unit. which is partly concealed beneath an earthflow, nor
the toe of the basal slip surface can yet be fully delin-
eated. The term ‘toe zone’ is therefore used here in
less than 30 cm thick. The foliation strikes generally preference to the word ‘toe’.
north and the orientation of the excavated face of the The head graben is crossed by multiple high-angle
hillside is NNW-SSE (Figure 3). The foliation dips at internal shears with counterscarps at outcrop (T1,
shallow angles towards the east, i.e. into the slope. T2 & T3 etc).
The rock sequence at site is cut by sets of pre-
existing faults. The most prominent fault set dips
steeply towards the E to ESE and three of these faults 3 DEFORMATION MONITORING BY TOTAL
can be traced across the landslide (Figure 1, faults A, STATION
B & C). The fault planes form counterscarps at outcrop
on the landslide and have oblique and vertical stria- Monitoring of the landslide has been carried out by
tions, suggesting distinct phases of slip. The faults also the road contractor since October 2003. The work

458
 Plane of foliation
Displacement

Figure 4. Displacements and velocities from total

station monitoring; and rainfall at Stesen Kajicuaca
Cameron High-lands.

Figure 5. 3D visualisation of the excavated hillside showing

involves nominally weekly measurement, by total sta- the surface displacement vectors (2003–5) to scale. Inset:
tion (Sokkia SET5E), of distances and horizontal and surface movement of the central toe zone block relative to
vertical angles from base stations west of the road foliation.
to reflective markers installed on the landslide. The
plan co-ordinates and reduced levels of the markers
are computed from these data. The magnitude, dip for photogrammetric adjustment and errors in the digi-
and dip direction of the displacement of each survey tal camera system. The photogrammetric survey was
marker have been calculated and velocities of move- compared to the November 2003 survey at 56 sur-
ment have been determined. Some of the monitoring vey check points in areas thought not to have moved
data are presented in Figure 4. between 2003 and 2005 (see Figure 1). The error stan-
Uncertainty in the data may be assessed by examin- dard deviation is 0.2 m. Uncertainty in the dip and dip
ing the reported movements of a marker located above direction of the vectors is less than 1◦ for the longest
the crown of the landslide (Marker 2D). In contrast to vector (24.3 m), a possible maximum of 13◦ for the
the markers within the landslide, the reported changes shortest vector (1.6 m), and an average of 2.7◦ for the
in the position of Marker 2D are very small (save mean vector (8.0 m).
for an unexplained excursion in September and Octo-
ber 2006) and no systematic pattern is evident. The
variance in horizontal position data to August 2006 5 POST-FAILURE LANDSLIDE MOVEMENTS
(standard deviation of data = 45 mm) is about three
times that expected from equipment error alone. The The surface displacement vector data (2003–5)
variance in height determination is as expected from advance our understanding of landslide behaviour.
equipment error alone. Viewed in plan the vectors are seen to be normal to the
slope face contours with lateral extension revealed by
radial divergence (north-south spreading), conform-
4 TOPOGRAPHIC SURVEY AND ing to topography, which takes the form of a subdued
PHOTOGRAMMETRIC MEASUREMENTS ridge. Movements are greater at the head than in the
toe zone (compression) and, on any slope face contour,
A digital elevation model had been created from topo- displacements are greater in the north than in the south
graphic survey in November 2003 and another was (rotation). Viewed in cross-section the vectors are seen
made photogrammetrically from aerial photographs to plunge at the head of the landslide, to generally lie
taken from a helicopter in September 2005. Displace- sub-parallel to the slope in the upper main body and
ment vectors were constructed from the differences to emerge in the toe zone.
between the two digital elevation models at identifi- It is instructive to examine the disposition of vectors
able features such as the ends of drainage channels and along the centreline of the displaced mass by means
berm edges. of stereographic projection (Figure 3). The directions
Many of the 150 displacement vectors are shown in of vectors within 100 m wide blocks of ground at the
Figure 5. highest part of the head (mean 238.5◦ ) and at the toe
Uncertainty in these measurements is associated zone (mean 246◦ ) closely correspond to the dip direc-
with the coordinates of the ground control points used tion of the face (mean 245◦ ). The disposition of vectors

459
at the highest part of the head coincides with a con- It appears that the volume of the landslide is about
centration of joint planes (Figure 3 – X) and at the 2 million m3 .
central toe zone block corresponds to the attitude of After failure the landslide decelerated until March
the foliation (Figure 3 – Y and Figure 5 inset). 2004 and it has since continued to move, for much of
The vectors reveal significant downslope compres- the time accelerating and then decelerating in surges.
sion. Compressive strain (defined as the displacement The timing of the surges generally coincides with
normalized against downslope length) measured on peaking in the 30-day rolling rainfall (Figure 3), rain-
centreline between upper main body (at the eleva- fall being measured at the Stesen Kajicuaca Cameron
tion of Markers 2A-2C) and the toe zone is about 5% Highlands raingauge of the Malaysian Meteorological
(2003–5). Such compression is evident in small-scale Service, 13 km SSE of the site. The bimodal rain-
sliding on foliation seen as shear offsetting (‘kicking fall pattern shown in Figure 4 is characteristic of an
out’), especially in the southern part of the landslide, inland climatic regime in peninsular Malaysia. It may
and by slip on the reactivated faults A, B & C etc. be that the landslide is responding to rainfall-induced
Observed fault slip movements are dextral, increase seasonal rise and fall of groundwater levels. Such fluc-
to the south and are greatest on fault B, where slip at tuation is manifest by intermittent seepage from the
the centreline is 3.5 m. southern toe zone. Other causal factors may have con-
The total station data give further insights into tributed to surges: a surge in late 2004 concurred with
landslide behaviour. Whilst the overall rate of dis- the removal of 100,000 m3 of ground from the northern
placement is declining slightly year on year, for much toe zone of the landslide.
of the time the displaced mass appears to be either
accelerating or decelerating. Five surges are apparent
(Figure 4) and comprise an accelerating phase (six to 7 CONCLUSIONS
eight weeks) and a decelerating phase (two to three
months). The velocity reached during surges at mark- The landslide is a slow rock slide in schist. Failure
ers 2A-C is generally about 20 mm/day (greater in late occurred in September 2003 and by December 2006
2004). the head had moved downwards more than 21 m. The
rate of displacement is declining slightly year on year,
but for much of the time the landslide mass is accel-
6 DISCUSSION erating and then decelerating in surges. There is some
correspondence between the timing of the surges and
The nature of the basal sliding surface(s) is of inter- the seasonal rainfall pattern. It is likely that the surges
est. Evidence is given above of movement at surface are induced by groundwater fluctuations. It may be
stations which is parallel to joint planes at the high- inferred from surface observations that the failure
est part of the head and to foliation in the central toe involves sliding at the head and in the upper main body
zone; slip on foliation is also visible on the ground. It of the landslide on joints roughly orthogonal to the
may be inferred, if the effects of non-parallel internal foliation, which dips at a shallow angle into the slope;
shear and change in landslide thickness are assumed in the central toe zone the landslide is sliding up and
insignificant, that the landslide is sliding on joint out on the foliation. The failure, which is probably
planes at the highest part of the head (i.e. at the main a compound slide of volume about 2 million m3 , has
scarp ‘normal fault’) and sliding upwards on foliation reactivated major pre-existing faults that run obliquely
in the central toe zone (but oblique to dip, Figure 5 through the landslide mass.
inset).
The vectors plunge steeply at the head and emerge
sharply in the toe zone, the profile suggesting a ACKNOWLEDGEMENTS
non-circular basal slip surface (Figure 5). The pres-
ence of multiple counterscarps in the head graben The study of the Pos Selim landslide was carried out on
(T1, T2, T3 etc. Figure 1) may signify curvature behalf of the Slope Engineering Branch of the Public
of the basal slip surface (Hutchinson, 1988). There Works Department of Malaysia and facilitated by the
are joints disposed to facilitate slip on such a curved road contractor, MTD Construction Sdn Bhd., who
surface (Figure 3 – dashed oval). The landslide is supplied the total station data. Prof YQ Chen of the
probably a compound slide. An educated guess was Department of Land Surveying and Geo-informatics,
made about the geometry of the basal slip surface, Hong Kong Polytechnic University checked the dis-
using the surface station movements and crack pat- placement and velocity calculations and assessed the
terns, and estimates were made of landslide volume. errors in the total station surveying.

460
REFERENCES Department of Spatial Sciences, Curtin University of
Technology, 281p.
Andrew Malone Ltd 2007. Landslide study at Ch 23+800 Hutchinson, J.N. 1988. General Report: Morphological
Simpang Pulai-Lojing Highway, Malaysia. Report to and geotechnical parameters of landslides in relation to
Minister of Works of Malaysia. geology and hydrogeology. In Proc Fifth International
British Standards Institution, 1999. Code of Practice for Site Symposium on Landslides (C Bonnard ed.) Balkema
Investigations BS5930:1999. Rotterdam v1 3–35.
Hansen, A. 2007. Semi-automated geomorphological map-
ping applied to landslide hazard analysis. Ph.D. Thesis,

461
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Characteristics of rock failure in metamorphic rock areas, Korea

W. Park & Y. Han

Architectural Engineering Team, Civil Part, Samsung Corporation, Seoul, Korea

S. Jeon
School of Civil, Urban and Geosystem Engineering, Seoul National University, Seoul, Korea

B. Roh
Technical advisor Team, Samsung Corporation, Seoul, Korea

ABSTRACT: The metamorphic rock ranges in Korea very extensively, which has the unique characteristics
through crustal movement and complex metamorphic processes for a long time. So the metamorphic rocks have
discontinuities such as dislocation and fault as well as weak part such as fractured zone and fault gauge. For
these reasons, there exist great potential of collapses, so it is required great caution to construction and design
roads or tunnels in this area. This paper describes the past experiences of slope failure during the construction
of roads at the area of metamorphic complex, especially Gyong-gy gneiss, and analyzes the characteristics of
the failure patterns in order to prevent from loss of the life and property.

1 INTRODUCTION the type of gneiss and schist in the middle region and
sporadically puts in the whole country.
Length of the Korean Peninsula is about 1, 000 km and As it is well known, metamorphic rocks have many
the area is 223,000 km2 , especially 70% of the land complicated geological structures by the movement of
consists of the mountain. But despite of small area, crustal and metamorphism processes for a long time,
there exist various rock types and the complex geo- so they have irregular discontinuities, rock fractured,
logical structures from Pre-Cambrian to Cenozoic era. various scale fault and weak gouge.
As the characteristics of distribution of the rock, meta- In case of the fault in these rocks, the size is not
morphic rock of Cambrian period and plutonic rocks of large scale, but considering direction and location of
Mesozoic exist extensively in the middle of the Korean fratures, it is not easy to determine and predict.
Peninsula. Sedimentary rock and igneous rock which Discontinuities such as joints and foliations are very
include stratum of Paleozoic and after that period complex and these crevices are infilled by gouge and
scatter on these bedrock. According to roughly com- weak deposit material. So on excavating slope or pit
positions of the rock, the metamorphic rock is 40%,
igneous rock is 35% and sedimentary rock is 25%, and
in the report of slope failures, mostly rock collapses
were happened in metamorphic and sedimentary rock
area in the road constructions. Actually, many losses Gyonggy Province.
of life and property happened from repetitive slope Seoul
failure.
Therefore, it is important to understand geological
characteristics and collect data of the failure cases of
the each rock type in order to avoid natural disaster and
failures of artificial structures of rock. For this pur-
pose, this paper described characteristics of failures in
metamorphic rock area. Busan
Mokpo

1.1 Geological characteristics of Korea

Figure 1 shows the geological map of South Korea.
Most of metamorphic rock of Pre-Cambrian exists as Figure 1. The geological map of Korea.

463
and tunneling in rock, the failures were frequently Many failures in metamorphic rock area happened
happened by the disadvantage geological structures. immediately after cutting, and then as the time goes,
The sedimentary rock which is mostly mud stone, failures of sedimentary rock occurred much more. And
shale and sand stone from Cenozoic era ranges in the in case of the igneous rock, the slope became weak
south-east region, and most of the these rock, except by erosion of the rock and fast weathering, therefore
the Pohang’s mud stone, are very hard and mass state. swallow face-failure and the scour were easy to happen
This rock sometimes has some problems when the on the slope. On exposing the slope face, weathering
excavation directions are the same with the dip direc- velocity of some ingredients of the rock mass becomes
tion of sedimentary rock. Many engineers know how much faster and the rainfall and underground water
to design in this rock type, so comparatively the dam- infiltrate into discontinuities.
ages are low, but if the failures have happened this These unpredicted collapses brought many losses
area, the loss would be largely. of the construction period and the additional cost.
In case of the igneous rock, even if the rock is intru- The important points of these accidents were that
sive rock or eruptive rock, it can be seen easily in most of the failure could not predict in design. This
Korea. Representative igneous rock is granite which paper is the case study about investigation of the col-
ranges very largely in Seoul and Kyung sang Province lapse of soil and rock slope by cutting slope in the
and mostly eruptive rock such as the andesite and tuff metamorphic rock site which is located at Gyong-gy
exists in south region nearby Mokpo. Province and analyses the pattern of the failures.
The characteristics of this rock failure are that the
many failures were happened after cutting the slope by
erosion and weathering. 2 CHARACTERISTIC OF THE SITE
As the rock types, there are unique characteristics
of the movements, so for designing the slope or tunnel, 2.1 Geological characteristics
it is very important to apply the appropriate geological
structures. This investigation area was located at the Gyong-gy
Province near the Seoul and there was the Route 45
which was a new lying road in the South-North direc-
tion, especially most parts of the route passed the
1.2 Statistics of the failures mountain area, so major works were tunneling and
About the slope failures, the important things, except cutting slope.
geological problems, are related with the point time As characteristics of the geology, most of rock at
of excavation and elapsed time. Figure 2 shows the this area was metamorphic rock, called Gyong-gy
data of slope failures as the elapsed time of the slope gneiss which is the representative metamorphic rock in
excavation at highway construction site in Korea (You, South Korea. This rock has unique characteristics of
1997). the geological structures through crustal movements
As seen in the graph, though initial collapses were and metamorphism processes for a long time. Joints,
relatively rare, many slope failures occurred shortly dislocations, fault and weak part such as fractured
after cutting slope and opening to public, totally 85% zone and fault gauge were existed, and till some of
of the failures were these times. depth, discontinuity surfaces were covered by ferrite
oxide and very weak clay, also it is hard to predict the
direction of the discontinuity.
Figure 3 is the cut slope with containing the fault
which parallels the strike of slope. Discontinuities of
these structures were common in this site, because
the fault is the same direction with the road. In the
other side, upper and side of slope consisted of resid-
ual soil and weathered soil, and some of the complex
dyke rock. Figure 4 shows the plane failure and wedge
failure which was two persistent joint with the line
of intersection of the joints daylighting at the rock
face.

2.2 Design and construction the slope

Figure 2. Ratios of slope failure in term of elapsed time In order to design, designers need many data of the site
from cutting slope for there different rock types in Highway such as geotechnical investigation data and the detailed
construction sites in Korea (You, 1997). geological structures. But it is hard to get the all data

464
 Table 1. Standard Guide for slope’s dip.

Condition Height Dip Remarks

Soil 0∼5 m 1:1.2 Establish the 1 m ramp

Above 5 m 1:1.5 per 5 m height
Rock Weather rock 1:1.0 Establish the 3 m ramp
Hard rock 1:0.5 per 20 m height

Fault

a b

Figure 3. The exposed fault in the slope and this fault is the
same direction with the Yong-in fault.

Figure 5. Stereographic Analysis; a) Discontinuities at the

first design. b) after cutting.

Figure 4. Exposed discontinuity which can be collapsed in slope was finished the excavation, unpredicted fault
the slope. and joint were existed in the slope. Actually this slope
was happened the wedge failure and small scale plan
failure by the discontinuities. So in case of design the
slope with the standard guide, it is simple to use it, but
in case of the construction site of the road, owing to the it is always not safe.
limit of the investigation, so the standard of the slope
design, which need not the precious data, is used very
often.
3 CHARACTERISTIC OF FAILURE
The standard of slope design is practically the same
IN THIS SITE
as related organizations of the civil construction, and
most of the design criteria are based on the strength of
3.1 Characteristic of the slope failure
the rock without the rock type. But if the designer used
only these guides no concerning the geological infor- The rock failures are influenced on many factors such
mation, many problems can be occurred, and a few as the discontinuities, strength of rock and the time of
organizations recommend that the characteristics of construction. In case of the metamorphic rock, there
the geological structures are considered on the slope, are many collapses after finishing the excavation as
if need. result of the investigation. These results were the same
Table 1 shows the standard guide for design of slope of the investigation which was performed by Korea
at the Korea Expressway Corporation & Ministry of Expressway Corporation.
Construction and Transportation. Table 2 shows results of investigation of the relation
This guide can be used comfortably in case of lack rock type which was classified by the rock strength
of the geotechnical investigation data, but there are and elapsed time of the excavation. 66% of slope fail-
many potentialities of the dangerous failure, but in ure happened after finishing excavation, and 85% of
case of being applied to importance cutting slope. the failure occurred in the rock masses which were
As the briefly examples, the dips of the slope in above strength of weather rocks. These reasons that
this investigation site were determined by the standard the more slope was cutting, the much free surface in
guide such as Table 1 without considering direction of slope face was enlarged. As it is well-known, the free
discontinuity and geological structures. surface gives the many influences in the slope. By the
The Figure 5a shows stereographic analysis of the investigation, the causes of many failures were by the
slope at the initial design. As seen the result of the discontinuity of the slope, especially fault and joint.
stereographic analysis, the dangerous block could not On the other side, the failures of residual soil or
be found, but as the face mapping on slope after the completely weathered soil, which had no geological

465
Table 2. Number of slope failure.
a
On On After
Rock construction complete project Total

R.S 3 1 1 5 (15%)
W.R 2 5 − 7 (20%)
S.R/H.R 7 15 1 23 (65%)
Total 12 (34%) 21 (60%) 2 (6%) 35
∗ R.Sis residual soil; W.R is weathered rock; S.R is soft rock; b
H.R is Hard rock.

structure patterns, were caused by the soil weight, and

most of failure shapes were circular type in this part.
Figure 6a shows the circular failure, and it is typical
circular failure by the weight of soil mass. However Weathered Acidic D yke
geological structures such as dyke rock and foliation
are faintly reminded in slope on account of the different
weathered speed, so failures in soil occasionally hap- Figure 6. Soil failure; a. The circular failure through the
pened like the type of rock mass failures, especially in line of least resistance by soil weight. b. The circular failure
the metamorphic or sedimentary rock area. by dyke.
Figure 6b shows the circular failure by dyke rock.
Dyke rock which did a role as discontinuity which
divided into some soil layers, if this discontinuity was
the same direction with the cut slope, circular or other
shape failure would be occurred.
As seen the result of the Table 2, soil failures did
not often happen, but 2-case happened on the finished
excavation and opening the public, the three failures
happened the on the construction. In case of weak zone
in the slope, the characteristic of the movement was
same as soil rather than rock, so it is hard to predict
the slope failure in the area where the fractured zone
and mixed fault.
As seen the table 2, collapse of 85% is related
with rock, and large scaled failures were caused by
the fault and gouge which is called Yong-in fault and
this discontinuity was put to South-North direction.
Even if the fault were so small scale, when the fault
was created, nearby the crust was move at once, so
short and dangerous discontinuities were extensively
existed at this area. And as another reasons, there
Figure 7. This picture is large slope failure in the this site.
were fault gouge and infilled materials into the gap This slope’s height is 56 m and the width is 120 m. a. view
of joint. of the slope. b. Vertical discontinuity exposed by ground
Figure 7 shows the representative large scale failure settlement at the failure. c. Tension crack on the ridge.
before starting to excavate in this area. The height of
the mountain is about 70 m and width is 250 m. The
main cause of the collapse was fault which existed in
slope of lower part, and there was a thin thickness of the problem of compensation with inhabitants around the
fault clay between discontinuities. The strike of fault site, it was hard to do additional cutting and lots of
paralleled with direction of slope, so mostly mode of monetary problems were happened.
failures were plan failure. The sources of these problems were lack of properly
Due to this failure, 5 m of settlements and over geotechnical investigation, especially investigation of
the 50 cm of tension crack were occurred for several the fault and dangerous discontinuities is not easy and
weeks. Because this failure was large scale and the in usually case, small fault does not consider to design.

466
 Table 3 shows classification of the slope failure as of slope was 25 m and width that collapse occurs was
shape of the failures and number of the occurrences about 50 m.
are referred to table 2. The main cause was the unexpected fault which of
As above mentioned, because the discontinuity are dip of the fault was N40◦ E/63◦ SE.
the main cause, failures are happened the rock which Figure 10 is the stereographic analysis of the
related the discontinuity. Figure 9. As seen the results, the direction of the joint
According to the investigation of failures in the site, were changed as the construction, and could not reflect
there were several cases of wedge and plane failure but the fault at the first design because there were no data
fortunately the toppling failure did not happen. about the fault.
On the excavation, some of topping failure blocks As seen the Figure 10, although there were many
were in the slope, but reinforcement methods such as dangerous blocks in slope at the initial condition, but
the bolt and rock anchorage were used to prevent from designer used the standard guide, then the collapsed
failure through using. happened.
Figure 9 shows the plane failure and wedge failure
after cutting and passed after 1year in this site. Height

Figure 9. View of the failure in the slope: a: Slope shortly

after cutting; b: Wedge failure in the slope; c: Fault gouge
with 50 cm of width and slickenside in the slope; d: Tensile
Crack in the ramp.

a b
Figure 8. Rock face formed by persistent discontinuities:
a. plane failure formed by bedding planes parallel to the face
with continuous length over the slope (Gneiss on Route 45
near Yong-in); b. wedge failure formed by two intersecting
planes dipping out of the face.

Table 3. Types of the slope failure.

Type of Circle Planar Wedge Toppling

failure failure failure failure failure Figure 10. Stereographic analysis of the discontinuities
plane at the construction site: a. is state of the discontinu-
No. of ity before digging and b. is result of the face mapping after
occurrence 5 19 11 − 1st cutting. Dip of the fault was 63 degrees.

467
 If more accurately investigation data of this site Therefore, in order to prevent from the rock fail-
were existed in design, failure could be prevent by ure, it is important to collect the more data such as
using a few rock bolts. accurate geotechnical investigation which include the
characteristic of the geological structures and to eval-
uate slope stability by the face mapping after cutting,
4 CONCLUSION and through the international research on the slope
design, the more stable and rationally design method
In this paper, we investigated the characteristics of must be established.
slope failure in the metamorphic rock area which is
called Gyong-gy gneiss. This rock mass has many
unpredicted discontinuities from the metamorphism REFERENCES
process and crustal movement for along time.
Therefore most of the road and tunnel constructions Lee Dae-Sung (edited). 1987. Geology of Korea. South
of this region, many failures were happened by the Korea.
geological structures. 85% of the slope failures were Hoek, E. & Bray, J. 1977. Rock Slope Engineering, 3rd edn,
concerned discontinuities of the rock and most of the IMM, London.
Hoek, E. & Bray, J. 2004. Rock Slope Engineering, 4th edn,
failures happened after cutting the slope or opening IMM, London.
the public. Korea Expressway Corporation. 1996. Highway slope
As these reasons, the exposed discontinuities were manual.
revealed on the slope after excavation. Also in case of Park, W.S. 2002. The report of the slope stability in Yongin,
the soil parts, some of the failures were induced by Samsung Corp.
the reminded geological structures such as fold and You, B.O. 1997. A Study on Harzad Rating System and Pro-
foliation, especially dyke rock. tective Measures for Rock Slope, PH.D Dissertation, Dept.
As the problems of the design, the designer depends of Civil Eng., Hanyang University, Seoul, South Korea.
on the standard guides of the slope without no consi-
dering the characteristic of the geology, even if there
are many studies of slope stability.

468
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Shape and size effects of gravel grains on the shear behavior of sandy soils

S.N. Salimi, V. Yazdanjou & A. Hamidi

Tarbiat Moallem University, Tehran, Iran

ABSTRACT: The shear behavior of sandy soils containing gravel particles has been investigated by many
researchers. However, the effects of the shape and size of gravel particles have not been particularly evaluated.
In the present study the shear strength of sand-gravel mixtures with two different gravel grain sizes and shapes
is studied in loose, medium and dense states using large direct shear test. The results of this study indicate that
the gravel shape and size has a little effect on the shear strength of sand-gravel mixtures in low gravel contents.
Increasing the gravel content to higher values makes this effect more clear. This is more obvious when the gravel
particles are no more floating in the sandy soil matrix. The samples containing angular gravel particles generally
show higher shear strength and dilation compared to the mixtures containing rounded to sub-rounded gravels.
Also the samples containing larger gravel particles usually show more shear strength and dilation compared to
the samples with smaller gravel particles in the same gravel content.

1 INTRODUCTION Simoni and Houlsby (2004) performed 87 large

direct shear tests on sand-gravel mixtures with dif-
Fragaszy et al (1990, 1992) introduced a new method ferent gravel contents. They concluded that increase
to evaluate the shear strength of soils containing in gravel content enhances the dilatancy rate and the
oversized particles. This method was based on the critical state friction angle.
assumption that larger particles floating in a matrix In the present study a regular set of 84 large direct
of finer grained material do not significantly affect shear tests were performed using a 300×300×170 mm
the strength and deformation characteristics of mix- direct shear box apparatus to investigate the effects of
ture. In other words, while oversized particles are the shape and size of oversized particles on the shear
floating in the finer matrix without any contact, the behavior of sand-gravel mixtures.
strength and deformation characteristics are controlled
by the matrix part alone. However, for higher over-
sized contents, it is controlled by both the sand matrix 2 SAMPLE PREPARATION
and oversized particles. Therefore, in a floating state
the behavior of the soil containing oversized parti- A bad graded fine sand with sub-rounded to rounded
cles can be simulated by testing the matrix portion grains was used as the base soil. Also two gravel types
alone, provided that the model specimen is prepared were used as the oversized particles. The first type
in near field density. The near field density is the was river gravel with rounded to sub-rounded grains
density of matrix in vicinity of oversized particles and the second type was an angular to sub-angular
(Fragaszy et al. (1992)). Based on this concept the one. Two different gradations with maximum grain
shear strength of granular soils containing oversized size of 12.5 mm and 25 mm were considered for each
particles reduces by increase in oversized content pro- gravel type. Each one was mixed with the base sand
vided that the relative density of the mixture remains in different contents to prepare sand-gravel mixtures.
constant. The maximum and minimum void ratios of the mix-
Yagiz (2001) investigated the effects of the shape tures were measured according to the ASTM-D4253
and content of gravel particles on the shear strength and ASTM-D4254. Also the specific gravity of the
of fine sandy soils using direct shear tests. It was sand and gravel grains was measured as 2.74 and 2.64
concluded that the shape and content of gravel par- respectively according to ASTM-D854. The dry unit
ticles have important effects on the friction angle of weights of different mixtures in desired relative den-
the mixture. sities were computed. The weight of the soil required

469
Figure 1. Shows the shear stress-shear displacement curve Figure 2. Vertical displacement-shear displacement curves
for two mixtures containing 20% gravel of maximum size of of mixtures containing 20% of different gravel sizes.
12.5 and 25 mm.

2 surcharge: 1.5kg/cm2
to make samples in a specific relative density was Dr: 60%, gravel content: 40%

shear stress (kg/cm2)

selected. The gravel and sand portions were mixed 1.5
based on the desired weight percents. The whole mix-
ture was divided into three equal fractions and the soil
1
sample was made in three layers by purring the first
fraction and to compact it using a metal hammer in
order to fill one third of the shear box height. The two 0.5
max. size: 0.5 inch
other layers were placed using the same procedure. max. size: 1.0 inch
0
0 10 20 30 40
shear displacement (mm)
3 EFFECT OF THE GRAVEL SIZE
ON THE SHEAR BEHAVIOR Figure 3. Shear stress-shear displacement curves of mix-
OF SAND-GRAVEL MIXTURES tures containing 40% of different gravel sizes.

The tests performed under an overburden pressure

of 1.5 kg/cm2 on samples in a relative density of grains is the main reason for higher shear strength in
60%. As it can be seen the shear stress-shear displace- comparison to the mixtures with smaller gravel grains.
ment curves are almost coincided for both mixtures. Mixtures with 40% and 60% gravel contents were
Although the peak shear strength of mixture contain- prepared to investigate the effect of gravel size on
ing larger gravel grains is a little higher, the residual the shear behavior in higher gravel contents. Figure 3
strengths of both mixtures are the same. In fact in low shows the shear stress-shear displacement curve for
gravel contents, the gravel grains are floated in the sand samples with 40% gravel content. It can be seen that
matrix and there is little contact between them. There- the difference between the curves has been increased
fore the shear behavior of the mixture is controlled for different gravel grain sizes compared to the curves
mainly by the sand portion and the gravel grain size for mixtures with 20% gravel content. This is mainly
does not play an important role on the shear behavior due to the increase in the contacts between gravel
of the mixture. grains in the mixture in higher gravel content. The con-
The variation of vertical displacement with shear tact forces between particles are more for larger gravel
displacement for the above mentioned mixtures is grains compared to the smaller ones due to the more
shown in Figure 2. It can be observed that the vertical contact surface that leads to higher shear strength.
displacement of the mixture containing gravel grains The same comparison was made considering the
with maximum size of 25 mm is more than the one con- vertical displacement-shear displacement curves. It
taining gravel grains with maximum size of 12.5 mm. was concluded that larger grains cause higher verti-
It seems that the increase in gravel grain size increases cal displacement or dilation. The difference between
the vertical displacement or the dilation of mixture dilation of mixtures with different gravel grain sizes
due to the more vertical movement and slip or top- increases with increase in gravel content. Figure
ping of the larger gravel particles over each other. The 4 indicates this case for the mixtures with 40%
more dilation of the mixtures containing larger gravel gravel content. The differences of shear strengths and

470
dilation of the mixtures with different gravel grain displacement-shear displacement curves for samples
sizes are more obvious when the gravel content is 60% with 60% gravel content.
in mixture. This is shown in Figures 5 and 6 which indi-
cate the shear stress-shear displacement and vertical
4 EFFECT OF THE GRAVEL SIZE
ON THE FRICTION ANGLE
OF SAND-GRAVEL MIXTURES
2.5 surcharge: 1.5kg/cm2
Dr: 60%, gravel content: 40%
The variation of friction angle for mixtures containing
vertical displacement (mm)

2
gravel grains of different size prepared in different rel-
1.5
ative densities is shown in Figure 7. It can be concluded

0.5 max. size: 0.5 inch

max. size: 1.0 inch
0
0 10 20 30 40
shear displacement (mm)
-0.5

Figure 4. Vertical displacement-shear displacement curves

of mixtures containing 40% of different gravel sizes.

Figure 5. Shear stress-shear displacement curves of mix-

tures containing 60% of different gravel sizes.

Figure 6. Vertical displacement-shear displacement curves Figure 7. Friction angle of mixtures containing gravel
of mixtures containing 60% of different gravel sizes. grains with maximum size of 12.5 mm and 25 mm.

471
that the increase in gravel content leads to the increase rounded gravel grains as shown in Figure 9. In fact the
of friction angle for all the mixtures. Besides the mix- angularity of gravel grains results in more vertical dis-
tures containing larger gravel grain particles show placement or dilation. This is due to the increase in
higher friction angle. Also the difference between the overtopping of gravel grains during shearing.
friction angles of mixtures containing different gravel Figures 10 and 11 indicate the shear stress-shear dis-
size increases in higher gravel contents. placement and vertical displacement-shear displace-
ment curves for mixtures with 40% of different gravel
shapes. It can be observed that the effect of gravel
5 EFFECT OF THE GRAVEL SHAPE shape on the peak and residual shear strength and
ON THE SHEAR BEHAVIOR dilation of sand-gravel mixtures are more obvious in
OF SAND-GRAVEL MIXTURES higher gravel contents. The more shear strengths of
mixtures containing angular gravel grains are mainly
In order to study the influence of gravel shape on due to the higher dilation occurs in these mixtures
the shear behavior of sand-gravel mixtures, a set of compared to the mixtures containing rounded gravel
direct shear tests have been conducted on mixtures grains. The same trend can be seen when the gravel
containing angular and rounded gravel grains. content increases to 60% as shown in Figures 12
It was observed that in low gravel content of 20%, and 13.
the shear stress-shear displacement curves of mix-
tures containing angular and rounded gravel grains are
nearly coincided as shown in Figure 8. However, the
vertical displacement of the mixture containing angu-
lar gravel grains is a little more than the one containing

Figure 10. Shear stress-shear displacement curves of mix-

tures containing 40% of different gravel shapes.

Figure 8. Shear stress-shear displacement curves of mix-

tures containing 20% of different gravel shapes.

Figure 11. Vertical displacement-shear displacement

Figure 9. Vertical displacement-shear displacement curves curves of mixtures containing 40% of different gravel
of mixtures containing 20% of different gravel shapes. shapes.

472
Figure 12. Shear stress-shear displacement curves of mix-
tures containing 60% of different gravel shapes.
Figure 14. Friction angle of mixtures containing angular
and rounded gravel grains.

For mixtures containing angular gravel with max-

imum grain size of 12.5 mm in a relative density of
60%:

φ = 0.32Gc + 24.7 R2 = 0.99 (2)

In the above equations Gc is the gravel content

in percent and φ is friction angle of the mixture in
degrees.

Figure 13. Vertical displacement-shear displacement 7 SUMMARY AND CONCLUSION

curves of mixtures containing 60% of different gravel
shapes.
A regular set of direct shear tests performed on mix-
tures of sand and different types of gravel grains to
investigate the effects of gravel grain shape and size
6 EFFECT OF THE GRAVEL SHAPE on the shear behavior of sand-gravel mixtures. It was
ON THE FRICTION ANGLE concluded that the addition of each type of gravel
OF SAND-GRAVEL MIXTURES grains to the sandy soil increases the shear strength and
dilation of the mixture intensively. However, the addi-
The variation of friction angle for mixtures containing tion of larger gravel grains increases the shear strength
angular and rounded gravel grains in different gravel and dilation of mixture more compared to the smaller
contents is shown in Figure 14 in a relative density ones. Also the difference between shear strengths of
of 60%. The figure shows that increase of the angular mixtures containing gravel grains of different size is
gravel content in sandy soil increases the friction angle more obvious in higher gravel contents. The same
of the mixture more compared to the rounded gravel. results obtained for mixtures containing gravel grains
A relationship can be determined for mixtures con- with different angularity. Addition of angular gravel
taining rounded and angular gravel grains to estimate grains increases the shear strength and dilation of the
the friction angle of the mixtures containing certain mixture more than rounded gravel grains. The effect
amount of gravel using a linear regression as shown of gravel grain angularity is more obvious in higher
in equations (1) and (2). These equations may be used gravel contents. The friction angles of the mixtures
to determine the friction angle of sand-gravel mixture containing different amounts of gravel content were
with a relative accuracy. also studied and it was understood that the increase in
For mixtures containing rounded gravel with max- gravel content results in more friction angle. However,
imum grain size of 12.5 mm in a relative density of the friction angle of the mixtures containing angular
60%: gravel grains is higher compared to the one contain-
ing rounded gravel grains in equal gravel content. The
same result obtained for friction angle of mixtures con-
φ = 0.24Gc + 24.7 R2 = 0.99 (1) taining larger gravel grains compared to the mixtures

473
containing smaller gravel grains. For equal gravel con- Fragaszy, R.J., Su, W., Siddiqi, F.H. & Ho, C.L. 1992. Mod-
tent, the friction angle of the mixture containing larger eling strength of sandy gravel. Journal of Geotechnical
gravel grains is higher than the one containing smaller Engineering Division, ASCE 118(6): 920–935.
gravel grains. Yagiz, S. 2001. Brief note on the influence of shape and
percentage of gravel on the shear strength of sand and
gravel mixture. Bulletin of Engineering Geology and the
Environment 60(4): 321–323.
REFERENCES Simoni, A. & Houlsby, G.T. 2006. The direct shear strength
and dilatancy of sand-gravel mixtures. Geotechnical and
Fragaszy, R.J., Su, W. & Siddiqi, F.H. 1990. Effect of over- Geological Engineering 24(3): 523–549.
sized particles on the density of clean granular soils.
Geotechnical Testing Journal 13(2): 106–114.

474
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Nonlinear failure envelope of a nonplastic compacted silty-sand

D.D.B. Seely
IGES, Inc., Salt Lake City, Utah, USA

A.C. Trandafir
Department of Geology and Geophysics, University of Utah, Salt Lake City, Utah, USA

ABSTRACT: Accurate evaluation of shear strength of compacted fill materials represents a key issue in
performing reliable stability analyses of embankment slopes. In engineering practice, the shear strength of
compacted soils is expressed in terms of a linear failure envelope usually derived from consolidated-undrained
(CU) triaxial tests with pore water pressure measurements. Experimental evidence indicates, however, that the
failure envelope of many soils is not linear, particularly within the range of low effective normal stresses. The
present study addresses the nonlinear character of failure envelope of a nonplastic compacted silty-sand. Linear
and nonlinear strength functions were fitted to experimental results from CU triaxial tests to derive the strength
parameters of linear and nonlinear failure envelopes. For the investigated soil and typical range of mean effective
normal stresses of interest in embankment slope design, the nonlinear failure envelope appears to give a more
accurate representation of the experimental information compared to the linear strength model.

1 INRODUCTION D 422-63). A compaction test (ASTM D 698-00) was

also conducted in order to obtain the optimum water
The importance of incorporating the nonlinear content associated with the maximum dry unit weight
character of soil failure envelope in slope stability of the material. The results of the grain size analysis
computations has been demonstrated by two dimen- are shown in Figure 1, whereas the proportion of clay,
sional limit equilibrium (LEM) and finite element silt, and sand characterizing the soil are depicted in
(FEM) analyses (Trandafir et al. 2000) as well as three Figure 2. Based on the laboratory test results, the ana-
dimensional slope stability studies (Jiang et al. 2003). lyzed soil classifies as a nonplastic silty-sand (SM)
Previous research has shown that failure envelopes of according to the Unified Soil Classification System
various soils (e.g., silt, sand, compacted clay, and (USCS) (ASTM D 2487-06). The soil consists of
compacted rockfill) exhibit nonlinearity especially 66.4% sand, 30% silt, and 3.6% clay (Figure 2).
within the range of small normal stresses (Penman The compaction test yielded an optimum water
1953, Ponce & Bell 1971, Charles & Soares 1984, content of 16.14% and maximum dry unit weight of
Day & Axten 1989). In this context, the present
study is concerned with the nonlinear character of
the failure envelope of a compacted silty-sand repre- 100
Percent finer by weight

senting potential embankment material for earthworks

80
throughout Utah. Experimental data from a series of
consolidated-undrained (CU) triaxial tests with pore 60
pressure measurement are utilized to derive both the
linear and nonlinear failure envelopes of the analyzed 40
soil, and the results are discussed.
20

2 PRELIMINARY SOIL TESTING 0

0.001 0.01 0.1 1 10
Preliminary laboratory tests were performed on the Grain size (mm)
soil used in this study for the purpose of geotechni-
cal classification, including liquid and plastic limits Figure 1. Grain size distribution curve of the silty-sand used
(ASTM D 4318-05), and grain size analysis (ASTM in the experimental study.

475
 Sand of the maximum dry unit weight obtained from the
compaction test using 254 mm lifts, with the soil being
0 scarified between lifts. The initial dimensions of the
100 specimens were 61.37 mm in diameter and 152.4 mm
20 in length. The samples were then trimmed to a height to
80 diameter ratio between 2:1 and 2.5:1 prior to prepara-
40 tion in the triaxial cell. The moisture-dry unit weight
60 data points of the compacted specimens for triaxial
60
testing plot very close to the point of optimum on the
40 compaction curve, as seen in Figure 3.
The triaxial samples were percolated with CO2 to
80
20 aid in saturation prior to the initialization of the triax-
ial tests as recommended by Rad (1984). The samples
100
0 were backpressure saturated, ensuring a Skempton’s
Clay 0 20 40 60 80 100 Silt B pore pressure parameter of 0.95 prior to consolida-
tion. A strain rate of 0.06%/min was used for undrained
Figure 2. Ternary diagram showing the percentages of sand, shearing, as derived from the time rate of consolida-
silt and clay in the analyzed soil. tion data. The triaxial testing program involved CU
axial compression (AC) and axial extension (AE) tests
on the compacted soil specimens. The experiments
17.0 were conducted using a state-of-the-art micropro-
cessor controlled fully automated triaxial equipment
Dry unit weight (kN/m3)

manufactured by Geocomp Corporation (Dasenbrock

16.8 2006). The effective stress paths from triaxial tests are
shown in Figure 4.
16.6 Seven CU-AC samples were sheared up to 20–25%
axial strain with average data sampling every 0.1%
strain. The AC specimens demonstrated a dilative
16.4
response yielding an effective confining stress at fai-
lure larger than the initial (consolidation) effective
16.2 confining stress (Figure 4). Consequently, a CU-AE
8 10 12 14 16 18 20 22 test (characterized by a contractive tendency of the
Water content (%) soil before failure) was included in the experimental
program to capture the soil strength behavior within
Compaction curve
the range of low effective normal stresses (Figure 4).
ZAV Gs = 2.6
Triaxial test specimens
The AE specimen was sheared up to approximately
−10% axial strain with data sampling every 0.01%
strain.
Figure 3. Results of the compaction test performed in accor-
dance with ASTM D698-00. The water content-dry unit
weight characteristics of the soil specimens used in the triaxial
testing program are plotted on the same curve. 4 INTERPRETATION OF FAILURE IN
TRIAXIAL TESTING OF SILTY SOILS

16.82 kN/m3 (Figure 3). These were the target com- Failure in dense low-plasticity silty soils subjected to
paction parameters used to prepare the soil samples triaxial compression testing is difficult to define due
for triaxial testing. to a continuous dilative behavior of these materials
with increasing shear strain. Consequently, a variety
of criteria used to define the onset of failure for triax-
3 TRIAXIAL TESTING ial tests can be found in the literature (Brandon et al.
2006). Three of these criteria have also been evalu-
A total of eight samples were prepared for triaxial test- ated in the context of the triaxial test results shown
ing. The soil was blended with water at a target water in Figure 4, i.e., Skempton’s A pore pressure param-
content equal to the optimum water content derived eter = 0 (Brandon et al. 2006), peak deviator stress
from the compaction test, and allowed to attain the (σ1 −σ3 )max , and peak principal stress ratio (σ1 /σ3 )max .
steady moisture condition for a minimum of 16 hours Figure 5 graphically demonstrates all three of the pre-
prior to compaction. The soil specimens were hand viously mentioned criteria and their applicability to
compacted in the laboratory to approximately 100% triaxial test results in this study.

476
 300

40 kPa AE
20 kPa AC
40 kPa AC
76 kPa AC
200
132 kPa AC
188 kPa AC
q (kPa)

244 kPa AC
300 kPa AC

100

0
0 100 200 300 400 500 600
p' (kPa)

Figure 4. Results of the CU-AC and CU-AE triaxial tests. The values in the legend represent the initial effective consolidation
stresses for the triaxial tests.

300 effective consolidation stress of 300 kPa in Figure 5).

ESP For this reason, the A = 0 criterion was discarded.
200 The disadvantage of the peak deviator stress
q (kPa)

(σ1 −σ3 )max criterion applied to the tested soil is that it

100
ESP results in only one point on the failure envelope, i.e.,
the strength value at the maximum shear strain attained
TSP TSP at the end of the triaxial test (Figure 5). Compared to
0 the other analyzed criteria, the (σ1 −σ3 )max criterion
0 100 200 300 400 500 600 will provide the minimum number of experimental
p' (kPa) data points that can be used to derive the strength
Peak deviator stress, ( '1- '3) max parameters characterizing the failure envelope.
Peak principal stress ratio, ( '1 / '3)max The criterion selected to define the onset of failure
A=0 in this experimental study is the peak principal stress
ratio (σ1 /σ3 )max . As seen in Figure 5, the stress point
Figure 5. Graphical representation of various criteria to associated with (σ1 /σ3 )max is above the stress point
define the onset of failure in triaxial tests. Both samples in corresponding to A = 0 failure condition for the sam-
the figure were sheared up to 20% axial strain. The A = 0 ple with an initial effective consolidation of 132 kPa
failure condition occurs where the effective stress path, ESP, thus providing a smaller number of experimental data
and the total stress path, TSP, intersect. points characterizing the failure envelope in this case.
However, unlike the A = 0 criterion, the (σ1 /σ3 )max
failure condition is achieved in all of the performed
triaxial tests (thus including the sample with an initial
Brandon et al. (2006) recommends the condition effective confining stress of 300 kPa in Figure 5 that
A = 0 to define the onset of failure in low-plasticity did not achieve the A = 0 condition). Furthermore, the
silts. As illustrated in Figure 5, the A = 0 condition (σ1 /σ3 )max criterion appears to provide a significantly
seems an appropriate choice for the triaxial specimen larger number of experimental data points describing
with an initial effective consolidation stress of 132 kPa. the failure envelope compared to the (σ1 −σ3 )max crite-
It results in a large quantity of points on the failure rion. Figure 6 demonstrates the range of experimental
envelope described by the effective stress path above data points located on the failure envelope based on
the A = 0 stress point. However, the A = 0 condition is (σ1 /σ3 )max failure condition, in relation to axial strain,
not achieved for some of the triaxial test results in this εa , for a triaxial test with an initial effective confining
study (e.g., the triaxial specimen with an initial mean stress of 244 kPa.

477
 4.0 300
3.5
3.0 200

q (kPa)
'1/ '3

2.5
Range of experimental data used 100
2.0 in deriving failure envelope
1.5
0
1.0 0 100 200 300 400 500 600
0.0 2.5 5.0 7.5 10.0 12.5 15.0 17.5 20.0 p' (kPa)
a (%) Power, q = 0.761p' 0.946 ; R2 = 0.999
Linear, q = 0.547p' + 2.70 ; R2 = 0.997
Figure 6. Principal stress ratio plot for sample with an effec-
tive confining stress of 244 kPa showing the range of data
Figure 8. Linear and nonlinear failure envelopes obtained
used in failure envelope regression analysis. The diamond
from fitting linear and nonlinear strength functions to the
point represents the peak principal stress ratio.
experimental data.

300
are shown in Figure 8. A high coefficient of deter-
200 mination R2 of 0.997 and 0.999 obtained for the linear
q (kPa)

and nonlinear regression analyses indicates that both

strength functions provide a very good description of
100 the experimental information which in fact covers a
range of mean effective normal stresses, p , within 25
0 to 500 kPa (Figure 7). For the interval of very small
0 100 200 300 400 500 600 mean effective normal stresses (i.e., p < 25 kPa), the
p' (kPa) nonlinear strength function predicts smaller strength
values compared to the linear model (Figure 9).
Figure 7. Experimental stress points on the failure envelope Pariseau (2007) has shown that both forward and
according to the principal stress ratio failure definition. backward extrapolation of a particular linear fit
applied to the experimental data can overestimate the
available shear strength of the material. This aspect is
By selecting the (σ1 /σ3 )max criterion to define the also demonstrated in this study by the q-axis intercept
onset of failure in triaxial testing, a reasonably large of the linear failure envelope, unlikely for a cohesion-
number of experimental stress points describing the less nonplastic soil such as the analyzed silty-sand
failure envelope was obtained. This will allow for a (Figure 9), and the deviation of the linear and non-
more accurate evaluation of the shear strength param- linear failure envelopes in the range of mean effective
eters characterizing the linear and nonlinear strength normal stresses above 350 kPa (Figure 8).
functions derived from regression analysis. Figure 7 Reliance on the R2 value alone to assess the
displays the experimental failure envelope data from accuracy of a strength function in describing the
triaxial testing according to the (σ1 /σ3 )max failure experimental information should be approached with
condition. caution. Pariseau (2007) pointed out that a common
but erroneous procedure is to interpret the failure enve-
lope with the greatest R2 value as the ‘‘best fit’’ failure
criterion. Care must be taken in evaluating the ‘‘best
5 LINEAR AND NONLINEAR FAILURE fit’’ for the range of normal stresses available from
ENVELOPES FROM REGRESSION laboratory testing as long as this range does not coin-
ANALYSIS cide with the interval of normal stresses for the design
problem of interest.
A regression analysis was performed using the exper- In case of compacted embankment slope applica-
imental data set shown in Figure 7. Conventional tions, the mean effective normal stress typically does
least squares minimization methods were employed not exceed 200 kPa. However, the experimental results
to derive the strength parameters for a typical lin- in Figure 7 are within a larger p interval (i.e., 25 to
ear kf function and a nonlinear empirical power law 500 kPa), thus the R2 parameter characterizing the
relationship. The resulting linear and nonlinear fail- linear and nonlinear strength functions in Figure 8 cor-
ure envelopes derived from the regression analyses responds to this large range of mean effective normal

478
 '
'

Figure 9. Expanded view of the linear and nonlinear failure envelopes together with the experimental data in the range of
small mean effective normal stresses.

stresses. Therefore, in order to investigate the accu-

racy of strength predictions within the interval of mean
effective normal stresses of interest, an approach based
on cumulative squares of residuals has been used. The
residual for a given p value was calculated as the dif-
ference between the experimentally measured q value
and q predicted by the linear or power strength func-
tions in Figure 8. The cumulative squares of residuals
was plotted against the mean effective normal stress
(Figure 10) to demonstrate the ‘‘fit’’ of the two strength
models within the range of stresses applicable to com-
pacted embankment slopes (i.e., 0 to 200 kPa mean
effective normal stress). As seen in Figure 10, for p
within 0 to 200 kPa, the nonlinear strength model pro-
vides either a comparable or a much lower value of the Figure 10. Demonstration of the ‘‘best fit’’ failure criterion
cumulative squares of residuals compared to the linear for the given range of stresses using the cumulative residual2 .
strength function. Thus, it may be concluded that for The nonlinear failure criterion provides a more accurate rep-
the analyzed compacted silty-sand and for the range resentation of the experimental data, having an overall lower
of mean effective normal stresses of interest in this cumulative residual2 for the given range of stresses.
study, the failure envelope described by the nonlinear
strength function is a better representation of the actual
strength of the soil.
valid descriptions of the experimental information.
However, a closer examination of the accuracy of
predictions within the interval of mean effective nor-
6 CONCLUSIONS mal stresses of interest for compacted embankment
slopes revealed a better ‘‘fit’’ of the nonlinear strength
Linear and nonlinear strength functions were fitted function to the triaxial data. The derived nonlinear
to laboratory triaxial test data to derive the failure failure envelope also predicted lower strength values
envelope of a compacted silty-sand. For the range of than the linear strength model within the range of very
experimental mean effective normal stresses, both lin- small normal stresses uncovered by the experimental
ear and nonlinear strength models provided equally information. This departure from linearity is usually

479
associated with smaller computed safety factors in a Classification System), Annual Book of Standards, Vol. 4,
conventional slope stability analysis based on the non- ASTM International, West Conshohocken, 2007.
linear strength model, implying therefore a safer slope Brandon, T.L., Rose, A.T. & Duncan, J.M. 2006. Drained and
design. undrained strength interpretation for low-plasticity silts.
J. Geotech. Geoenviron. Eng., 132(2): 250–257.
Charles, J.A. & Soares, M.M. 1984. Stability of compacted
rockfill slopes. Geotechnique 34(1): 61–70.
ACKNOWLEDGMENTS Dasenbrock, D.D. & Hankour, R. 2006. Improved soil
property classification through automated triaxial stress
Funding for this research was provided by the Under- path testing. Proc. GeoCongress, Atlanta, Feb. 26–Mar. 1
graduate Research Opportunity Program (UROP) at 2006.
the University of Utah. The authors would like to thank Day, R.W. & Axten, G.W. 1989. Surficial stability of
IGES, Inc. for open access to their extensive laboratory compacted clay slopes. J. Geotech. Eng. 115(4): 577–580.
facilities that were used in the present experimental Jiang, J.C., Baker, R. & Yamagami, T. 2003. The effect of
investigation. strength envelope nonlinearity on slope stability compu-
tations. Can. Geotech. J. 40(2): 308–325.
Pariseau, W.G. 2007. Fitting failure criteria to laboratory
strength tests. J. Rock Mech. & Mining Sciences, 44:
REFERENCES 637–646.
Penman, A. 1953. Shear characteristics of saturated silt
ASTM 2007. Standard D 4318-05: Standard Test Methods for measured in triaxial compression. Geotechnique 15(1):
Liquid Limit, Plastic Limit, and Plasticity Index of Soils, 79–93.
Annual Book of Standards, Vol. 4, ASTM International, Ponce, V.M. & Bell, J.M. 1971. Sear strength of sand at
West Conshohocken, 2007. extremely low pressures. J. Geotech. Eng. 9(4): 625–638.
ASTM 2007. Standard D 422-63: Standard Test Method for Rad, N.S. & Clough, G.W. 1984. New procedure for saturat-
Particle-Size Analysis of Soils, Annual Book of Standards, ing sand specimens. J. Geotech. Eng. 110(9): 1205–1218.
Vol. 4, ASTM International, West Conshohocken, 2007. Trandafir, A., Popescu, M. & Ugai, K. 2001. Two dimen-
ASTM 2007. Standard D 698-00: Standard Test Method sional slope stability analysis by LEM and FEM consid-
for Laboratory Compaction Characteristics of Soil Using ering a non-linear failure envelope. Proc. 40th Annual
Standard Effort (12,400 ft-lbf/ft3 (600 kN-m/m3 )), Conf. of Japan Landslide Society, Maebashi, August
Annual Book of Standards, Vol. 4, ASTM International, 2001: 219–222.
West Conshohocken, 2007.
ASTM 2007. Standard D 2487-06: Standard Classifica-
tion of Soils for Engineering Purposes (Unified Soil

480
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

An investigation of a structurally-controlled rock cut instability at a metro

station shaft in Esfahan, Iran

Ali Taheri
Zaminfanavaran Consulting Eng. (ZAFA), Esfahan, Iran

ABSTRACT: On July 29th 2006 at 1.30 A.M, an overall slope failure took place at eastern wall of a 28 m-depth
metro station shaft in Esfahan, Iran. The failure caused some serious damage to the city utilities and loss of one
man life. The instability occurrence was about 10 month after the shaft completion, where, the shaft walls had
been supported by a regular grid of grouted dowels, steel mash and shotcrete.
According to the geotechnical investigation results carried out after the failure, the main cause of the rock mass
instability at the shaft wall was a major shear joint existence in the rock mass, trending almost parallel to the slope
face and dipping towards the shaft bottom with about 60 degrees angle. This structural feature was unrecognized
before the failure incidence.
In this paper, by describing the induced failure mechanism, the dominant role of the existing natural plan of
weakness on the rock instability is analyzed and the unsuccessful effect of the used rock supporting system is
discussed. It has been concluded in this paper that the possibility of the structurally-controlled rock failures
should be taken into account seriously in design and execution of any earth structures, especially in urban areas.

1 INTRODUCTION the project area. These deposits which are denoted

as ‘‘Shemshak Series’’ are overlain by a slightly
The Esfahan Metro Project is a large scale infrastruc- cemented, coarse grained alluvium, which its thick-
ture project which has attracted international interest ness is around 9 m at the proposed site.
form inception as it passed through the city of Esfa-
han with 2500 years history and architecture and home 2.2 Structural features
to a number of highly treasured world heritage sites. The dominant structural features of the bedrock mass
On the other hand, the geological and hydrogeologi- are bedding plane and three systematic joint sets, with
cal conditions of the project area is various and rather the orientations given in Table 1.
complicated. There is no visible geological discontinuity in the
The line 1 of the Esfahan subway passes through covering alluvial deposits.
bedrock at southern part of the city and through soft
deposits at the middle and northern parts. The kargar 2.3 Groundwater condition
Station shaft instability occurred in the discontinuous There is a shallow, low transmissible groundwater
bedrock mass due to sliding on a major shear joint, aquifer at the southern part of the Esfahan city, which
10 months after completion of the ground excavation has been formed in the alluvial deposits and shallow
and support installation. layers of the bedrock mass. The main recharge source
In this paper, by describing the geological con-
ditions of the site and explanation of the installed
support system of the station cut, the induced failure Table 1. Orientations of discontinuity system in the rock
mechanism has been discussed and conclusions and mass.
recommendations are given in order to prevent such
instabilities in similar projects. Orientation (deg.)

Discontinuity type Dip direction Dip amount

2 GEOLOGY
Bedding plane 185–200 35
2.1 Lithology Joint set 1, J1 310 85
Joint set 2, J2 110 75
The lower Jurassic (Lias) deposits, comprising shale Joint set 3, J3 010 60
and sandstone alternation, form the bedrock of

481
of this aquifer is water infiltration form green areas, 4 THE INSTALLED SUPPORT SYSTEM
as well as, leakage from water and sewage lines. The
groundwater flow direction is generally from south to In order to support the proposed station shaft walls, a
north, with hydraulic gradient of about 3%. The ori- passive supporting system has been designed and exe-
ginal groundwater table at the proposed site lies in the cuted during the excavation process. In designing this
range of 6.2 to 8.8 m. bellows the ground surface. system, the rock mass had been considered as a equiv-
alent continuous media, with potential failure plane as
non-structurally controlled (similar to soil). The wall
3 GEOTECHNICAL ASPECTS supporting system was comprised of φ32 mm grouted
dowels with 12 , 8 and 6 m. lengths, at 2 × 2 m. regu-
In order to assess the effect of the geological dis- lar grid, and 30 cm. thickness shotcrete, reinforced by
continuities on stability of the proposed rock mass, two 10 × 10 × 0.8 cm. weldmeshes (Figure 2).
the orientation of the four distinguished discontinuity
sets have been plotted on a streonet, together with the
orientation of the shaft cut face (Figure 1). 5 THE FAILURE MECHANISM
It is evident from this streoplot that the existing dis-
continuities were unlikely to be involved in the slope When the overall failure of the shaft wall started to
failures. So, the potential slope failure had been con- take place at 1:30 A.M., there were five persons work-
sidered as non-structurally controlled and the slope ing inside the shaft, which one of them had not chance
stability analyses were performed on the basis of this to survive. Another worker who was witness of the
assumption. This led to design and execution of the incident reported that the failure started from bottom
following slope supporting system. of the shaft by mass sliding towards the opposite wall
In designing the executed supporting system, by with a loud voice, and followed by toppling of soil,
assuming the rock mass as an equivalent continuum rock and shotcrete blocks from high levels of the wall
media, the shear strength of the rock mass had been (Figure 3). It should be added that 10 days before fail-
determined on the basis of the Hoek-Brown failure ure incidence, some tension cracks had been formed at
criterion, as follows (Hoek et al. 2002). upper surface of the shaft, which nobody paid attention
to it.
σ1 = σ3 + 49(0.0176σ3 + 0.0003)0.508 (1) According to the geotechnical investigation results,
carried out after the failure incidence, a major shear
joint (local fault plane), trending almost parallel to
where σ1 and σ3 are the major and minor effective
the slope face and dipping towards the shaft bottom
principle stresses at failure, in Mpa.
with about 60 degrees angle, was the main cause for
the failure occurrence. This fault plane which had not
been recognized before the failure incidence was filled
by a soft and saturated gouge of about 30 cm thickness.

Figure 1. Streoplot of the discontinuity system in the rock

mass. Figure 2. Wall supporting system of the station shaft.

482
Figure 3. General view of the failed shaft wall.

Figure 4. Plan and cross section of the failed shaft wall.

As shown in Figure 4, the induced failure had a
dual mechanism; sliding of a discontinuous rock on
the existing fault plane and then rotation and toppling • Besides the insufficient supporting system, the poor
of the soil blocks and shotcrete slabs from higher levels installation of the grouted dowels intensified the
of the wall (slide head toppling). instability problem of the shaft wall. For instance,
by inspecting the failure plane after removal of the
debris, it was found that just a limited number of
the dowels were failed in tension, while, most of
6 THE MAIN REASONS OF THE INSTABILITY them were pulled out from failed or remained rock
masses (Figure 5).
In spite of the supporting system installation for stabi- • As the failure took place 10 months after completion
lization of the shaft walls, about half of the eastern of the shaft excavation, the sliding of the poten-
wall was failed in a few seconds, 10 months after tially unstable rock block can be related to increase
completion of the shaft excavation. of groundwater pressure on the fault plane, due to
By consideration of the induced failure mechanism the high water infiltration rate from adjacent green
and taking into account all the parameters affecting areas in hot summer.
the stability of the slope, the main reasons of the shaft
wall instability were reported as follow.

• Design of the wall supporting system by assuming 7 FAILURE MECHANISM ANALYSIS

the rock mass as a continuum media with equiva-
lent geomechanical parameters. While, the existing In order to analyze the induced failure mechanism,
fault plane caused the rock mass to behave as a windows grogram called ‘‘ROCPLANE’’ has been
discontinuous, with structurally controlled failure used. This program is an interactive software tool
incidence. for assessing the stability of planar sliding blocks in

483
 rock slopes. It also allows users to estimate the sup-
port capacity required to achieve a specified factor of
safety.
To do this analysis, the Mohr-Coulomb shear
strength parameters of the fault plane have been deter-
mined by performing the Consolidated-Undrained tri-
axial tests on a number of representative undisturbed
samples of the fault gouge material in the laboratory.
The mean shear strength parameters determined in this
manner are as follows:
• Cohesion, c = 0.5 kg/cm2 ,
• Angle of internal friction, φ = 30 degrees.
By assuming the groundwater pressure distribution
as shown in Figure 6, the factor of safety of the so
supported slope was found as 1.1, without taking into
account the retaining effect of the reinforced shotcrete.
In this case, the total retaining force of the installed
grouted dowels was found 157.5 ton/m, assuming all
the dowels were mobilized simultaneously. While, by
factor of safety of 1 the retaining force of the dowels
drops to 118.5 ton/m, i.e. 75% of the expected total
force, and by taking into account the retaining effect
of the reinforced shotcrete, it was found that around
50% of the installed grouted dowels were not mobi-
lized simultaneously with the others, at the verge of
the failure.

Figure 5. Pull out of the dowels from the failed rock mass. 8 CONCLUSIONS AND RECOMMENDATIONS

According to the induced failure mechanism and rea-

sons, the following conclusions and recommendations
can be given.
• Regarding the dominant role of the major geological
discontinuities on stability of rock slopes, in pre-
liminary designing of the wall supporting systems,
the potentiality of structurally controlled failures
should be taken into account, especially in urban
areas. To do this, a good practice is to carry out
some sensitivity analysis by assuming a range of
possible orientations and shear strength parameters
for the major structures in rock masses.
• To confirm the geological structure of the slope,
further geological mapping together with evidence
on groundwater and tension cracks, would pro-
vide information for review of the situation to
decide upon the best means of slope stabilization,
in addition to the drainage measures.
• Due to the considerable effect of the supporting sys-
tems installation quality on stability of rock masses,
the proper execution of the rock supporting designs
is highly recommended. In this sense, it is desirable
that unlike the studied case, the passive supporting
systems be mobilized simultaneously. Otherwise,
Figure 6. Geometry of the failed slope. the resisting capacity of the supporting systems will

484
 be decreased and may not be able to sustain the Haraz Rah Consulting Eng. Group. 2006. A study of the
destructive forces of the unstable rock blocks. Kargar Station eastern wall failure. Technical report, in
• In order to control the stability of the slopes and Persian, Tehran-Iran.
efficiency of the ground stabilization systems, it is Hoek, E. & Bray, J.W. 1981. Rock Slope Engineering, 3rd.
recommended that to perform some instrumentation ed. London: IMM.
Hoek, E., Carranza-Torres, C.T. & Corkum, B. 2002. Hoek-
and monitors surface movements and subsurface Brown failure criterion-2002 ed., proc. North American
deformations of the ground, during execution and Rock Mechanics Society meeting in Toronto in July 2002.
operation of the shafts in the urban areas. Ortigao, J.A.R. & Sayao, A.S. 2004. Handbook of Slope
Stabilization. Berlin: Springer.
Simons, N., Menzies, B. & Matthews, M. 2001. Soil and
ACKNOWLEDGMENTS Rock Slope Engineering. London: Thomas Telford.
US Army Corps of Engineers. 2003. Engineering and Design
The author would like to record his appreciation of Slope Stability, Manual No. 1110–2-1902.
help from Mr. H. Mansoori Broojeni in collection and Wyllie, D.C. & Mah, C.W. 2004. Rock Slope Engineering,
Civil and Mining, 4th ed., Based on the 3rd. ed, by Hoek,
analysis of the failure data. E. & Bray, J. , London: Spon Press.
Zaminfanavaran Consulting Eng. (ZAFA). 2006. Concluding
report of the instability at the Kargar Station eastern wall.
REFERENCES in Persian, Esfahan-Iran.

Alamoot Bridge & Building Eng. Co. 2006. Geological

report of the failure event at the Kargar Station eastern
wall, Report No. EURO-ALMT-RO3-RPT-GEO-2003-
(Rev. 0), in Persian, Tehran-Iran.

485
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Yield acceleration of soil slopes with nonlinear strength envelope

A.C. Trandafir
Department of Geology and Geophysics, University of Utah, Salt Lake City, Utah, USA

M.E. Popescu
Department of Civil and Architectural Engineering, Illinois Institute of Technology, Chicago, Illinois, USA

ABSTRACT: This paper presents and discusses the results of a pseudostatic slope stability analysis aiming
to address the influence of strength envelope nonlinearity on the computed yield acceleration of a uniform
slope in a homogeneous soil. Published linear (Mohr-Coulomb) and nonlinear strength envelopes derived
from the same experimental database were employed in the analysis. For the variety of slope geometries ana-
lyzed in this study, the nonlinear strength envelope always resulted in a smaller yield acceleration compared
to the traditional Mohr-Coulomb strength envelope. The difference between yield accelerations associated with
the two failure criteria appears to increase with decreasing slope height and increasing slope inclination. For
the same slope geometry, the critical sliding surface corresponding to the yield acceleration for the nonlin-
ear strength envelope is shallower than the critical sliding surface associated with the Mohr-Coulomb strength
envelope.

1 INTRODUCTION static safety factors compared to those provided

by the linear (Mohr-Coulomb) strength envelope
Assessment of seismic displacements of slopes based (Maksimovic 1979, Charles and Soares 1984, Popescu
on Newmark sliding block method (Newmark 1965) et al. 2000, Trandafir et al. 2001, Jiang et al. 2003).
represents a routine engineering practice. Newmark Given the interdependence between static safety fac-
model consists of one-block translational or rota- tor and yield acceleration of a potential sliding mass,
tional mechanism along a rigid-plastic sliding surface the importance of considering the strength envelope
which is activated when the ground shaking accel- departure from linearity in seismic slope stability eval-
eration exceeds the yield acceleration of the sliding uations may be easily inferred; the result may be a
mass. The yield acceleration is defined as the earth- smaller value of the computed yield acceleration and
quake acceleration required to bring the sliding mass therefore a higher vulnerability of the analyzed slope
to the limit equilibrium condition corresponding to a to earthquake induced instability. This paper exam-
safety factor of 1.0. The yield acceleration is the yield ines through a parametric study the influence of soil
coefficient (ky ) multiplied by the gravitational accel- strength envelope nonlinearity on the computed yield
eration (g). The yield coefficient (ky ) is obtained from acceleration of a uniform slope in a homogeneous
conventional pseudostatic slope stability analyses and soil. Published linear Mohr-Coulomb and nonlinear
depends, among other factors, on the shear strength strength envelopes derived from the same experimen-
properties of the material along the sliding surface. tal database are employed with pseudostatic slope
Soil shear strength is generally expressed in terms stability computations to obtain the yield acceleration
of Mohr-Coulomb linear failure criterion τ = c + σ of the sliding mass.
tan φ, with c (cohesion) and φ (angle of internal
friction) representing the conventional shear strength
parameters. There is however considerable experi-
mental evidence showing that the strength envelope 2 SLOPE GEOMETRY AND SHEAR
of many soils is not linear particularly within the STRENGTH ENVELOPES
range of small effective normal stresses (e.g., Bishop
et al. 1965, Charles and Soares 1984, Atkinson and Figure 1 shows the geometric variables (i.e., slope
Farrar 1985, Maksimovic 1989). Furthermore, slope height H and slope inclination parameter b) for the
stability analyses conducted for materials exhibit- parametric study, and the forces acting on the slid-
ing a nonlinear failure envelope resulted in lower ing mass in a pseudostatic slope stability analysis to

487
 Ground surface the shear strength estimate in the range of small nor-
mal stresses obtained from the projection of the linear
strength envelope in the range of larger normal stresses
b is unsafe.
kW
H 1
W =17.7 kN/m3

Sliding b=2 (NL) b=2 (MC)

surface b=2.5 (NL) b=2.5 (MC)
b=3 (NL) b=3 (MC)
Figure 1. Slope geometry and forces acting on the slide
mass.
0.55

150

Yield acceleration (g)

Mohr-Coulomb strength envelope 0.45
Nonlinear strength envelope
100
(kPa)

0.35

50
0.25

0
0 50 100 150 200 250 0.15
4 6 8 10
(kPa)
Slope height, H (m)
Figure 2. Linear and nonlinear strength envelopes consid-
ered in the analysis. Figure 3. Yield acceleration versus slope height for var-
ious slope inclinations; NL—nonlinear strength envelope;
MC—Mohr Coulomb strength envelope.

determine the yield acceleration of a specific slope.

H=4m (NL) H=4m (MC)
The linear (Mohr-Coulomb) and nonlinear strength
H=7m (NL) H=7m (MC)
envelopes used in the analysis are depicted in Figure 2.
These envelopes were derived by Jiang et al. (2003) H=10m (NL) H=10m (MC)
using least-square estimates on experimental results
from numerous (i.e., 103) laboratory triaxial tests 0.55
on heavily compacted Israeli clay. The conventional
shear strength parameters characterizing the linear
Yield acceleration (g)

0.45
(Mohr-Coulomb) failure envelope (Figure 2) are c =
11.7 kPa and φ = 24.7◦ . The nonlinear strength
envelope is described by a power-type relationship 0.35
τ = Pa A(σ/Pa )δ (Baker 2004), with Pa represent-
ing the atmospheric pressure, whereas A and δ are
0.25
the dimensionless strength parameters of the nonlinear
strength function. For the nonlinear failure envelope
in Figure 2, A = 0.582 and δ = 0.857. 0.15
By comparing the sums of squares of residuals,
Jiang et al. (2003) noticed equally valid descriptions
provided by both linear and nonlinear strength mod- 0.05
els of the available experimental data associated in fact 1.5 2 2.5 3
with normal stress levels greater than 35 kPa. It is to be Slope parameter, b
noted that triaxial tests reported by Atkinson and Farrar
(1985) indicate that the nonlinear shear strength enve- Figure 4. Yield acceleration versus slope inclination for
lope apparent in the range of small stresses becomes various slope heights; NL—nonlinear strength envelope;
linear in the range of larger normal stresses. Therefore MC—Mohr Coulomb strength envelope.

488
 1.08

1.06

1.06
1.04

1.08
1.10
2
1.0

1.12
1.14

1.02
1.16

1.04
1.06
1.18

1.08
F s =1

F s =1
a) b)

15.0 15.0

12.5 12.5
Elevation (m)

Elevation (m)

10.0 10.0

7.5 7.5

kyNL = 0.251
5.0 kyMC = 0.479 5.0

2.5 2.5

0.0 0.0
05 10 15 20 25 05 10 15 20 25

Horizontal distance (m) Horizontal distance (m)

Figure 5. Critical sliding surface corresponding to the yield coefficient of a 5 m high slope with an inclination b = 2:
(a) linear strength envelope model (MC); (b) nonlinear strength envelope model (NL).

For the particular soil considered in our analy- software that can handle both linear and nonlinear
ses (Figure 2), the nonlinear strength model predicts strength envelopes in a limit-equilibrium slope sta-
smaller strength values than linear (Mohr-Coulomb) bility analysis. In Slope/W procedure, the nonlinear
strength envelope for normal stresses below 35 kPa. strength envelope is introduced as a general data
The following section of this paper presents and dis- point function consisting of a series of (σ, τ) values
cusses the results of a parametric study performed to characterizing the failure envelope. For each analyzed
illustrate the effect of the nonlinear failure envelope in slice, Slope/W computes the local slope angle (φ) and
the range of very low normal stresses (i.e., <35 kPa) cohesion intercept (c) of the tangent to the nonlinear
on the yield acceleration. failure envelope, as a function of the normal stress at
the base of the slice (GEO-SLOPE International, Ltd.,
2002). Consequently, in slope stability analysis with
3 PARAMETRIC STUDY nonlinear strength envelope using Slope/W module,
the linear strength parameters (c, φ) are different for
The effect of strength envelope nonlinearity on the each slice.
yield acceleration of the sample slope depicted in In limit-equilibrium slope stability analysis, the
Figure 1 was studied for various slope heights and seismic coefficient (k) associated with a safety factor
inclinations. The yield coefficient (ky ) giving the yield of 1.0 represents the yield coefficient (ky ) of the sliding
acceleration (ky g) of the slide mass was determined mass. Therefore, k was gradually increased in the anal-
from a pseudostatic slope stability analysis along cir- ysis until the calculated safety factor reached 1.0. For
cular sliding surfaces based on Bishop’s simplified each input k value, the minimum factor of safety was
method of slices. The analysis was conducted using determined by enabling the automatic search option in
the Slope/W module of the GEO-SLOPE OFFICE Slope/W to locate the pole of the critical failure surface
package (GEO-SLOPE International, Ltd., 2002), a across a grid of potential slip circle centers (Figure 5).

489
 Figures 3 and 4 present the yield acceleration ver- design, if the relevant range of normal stresses along
sus slope height and, respectively, slope inclination, the sliding surface of the investigated slope is over-
as computed for the linear (MC) and nonlinear (NL) looked in laboratory evaluations of Mohr-Coulomb
strength envelope models shown in Figure 2. Overall, strength parameters. The discrepancy between yield
the nonlinear strength model resulted in smaller yield accelerations computed using linear and nonlinear
acceleration values compared to the linear strength strength models becomes larger for smaller height
model. The difference between the computed yield slopes and steeper slopes. This is related to the fact
accelerations based on the two strength envelope mod- that in smaller height slopes and steeper slopes, the
els becomes more pronounced with decreasing slope zones of smaller normal stresses which are within the
height (Figure 3). For a slope inclination correspond- range of pronounced strength envelope nonlinearity
ing to b = 2, this difference increased from 33% have more extent.
(H = 10 m) to 51% (H = 4 m) expressed as per-
centage of the yield acceleration associated with the
linear (MC) strength envelope. Additionally, the yield
acceleration given by the NL model is less sensitive to REFERENCES
variations in slope height compared to the yield accel-
Atkinson J.H. & Farrar D.M. 1985. Stress path tests to
eration determined for the MC model. On the other measure soil strength parameters for shallow landslips.
hand, the yield acceleration corresponding to the NL Proc. 11th Int. Conf. on Soil Mechanics and Foundation
model shows a more abrupt increase with decreasing Engineering, San Francisco: 983–986.
slope angle compared to the MC based yield accelera- Baker R. 2004. Nonlinear Mohr envelopes based on triaxial
tion (Figure 4). The difference between the NL and data. J. Geotech. and Geoenv. Eng., 130(5): 498–506.
MC yield accelerations increases significantly with Bishop A.W., Webb D.L. & Lewin P.I. 1965. Undisturbed
increasing slope inclination. For example, this differ- samples of London Clay from the Ashford Common shaft:
ence increased from 31% (b = 3) to 69% (b = 1.5) Strength effective normal stress relationship. Geotech-
expressed as percentage of the yield acceleration asso- nique 15(1): 1–31.
Charles J.A., & Soares M.M. 1984. The stability of slopes
ciated with the linear (MC) strength envelope for a in soils with nonlinear failure envelopes. Canadian
slope height H = 4 m (Figure 4). The differences in Geotech. J., 21: 397–406.
terms of depth and geometry between critical sliding GEO-SLOPE International Ltd. 2002. Computer program
surfaces corresponding to the yield coefficient asso- SLOPE/W for slope stability analysis. User’s guide,
ciated with the NL and MC strength models for a Version 5, Calgary, Alberta, Canada.
slope with H = 5 m and b = 2 are illustrated in Jiang J.C., Baker R., & Yamagami T. 2003. The effect of
Figure 5. Apparently, the type of failure envelope strength envelope nonlinearity on slope stability compu-
has a significant influence on the critical sliding sur- tations. Can. Geotech. J. 40(2): 308–325.
face. The nonlinear strength model always resulted in Maksimovic M. 1979. Limit equilibrium for nonlinear failure
envelope and arbitrary slip surface. Proc. 3rd Int. Conf.
a much shallower critical sliding surface compared to on Numerical Methods in Geomechanics, Aachen, A.A.
the linear (Mohr-Coulomb) strength model. Balkema, Rotterdam, Vol. 2: 769–777.
Maksimovic M. 1989. Nonlinear failure envelope for soils.
J. Geotech. Eng., 115(4): 581–586.
4 CONCLUSIONS Newmark N.M. 1965. Effects of earthquakes on dams and
embankments, Géotechnique 15(2): 139–159.
The parametric study presented in this paper illustrates Popescu M., Ugai K. & Trandafir A. 2000. Linear versus non-
the significant effect of strength envelope nonlin- linear failure envelopes in LEM and FEM slope stability
earity in assessment of seismic slope stability. For analysis. Proc. 8th Int. Symp. Landslides, Cardiff, A.A.
Balkema, Rotterdam, Vol. 3: 1227–1234.
slope materials exhibiting nonlinear strength envelope Trandafir A., Popescu M., & Ugai K. 2001. Two dimensional
within the range of low normal stresses, an analysis slope stability analysis by LEM and FEM considering
based on the traditional Mohr-Coulomb failure crite- a non-linear failure envelope. Proc. 40th Annual Conf.
rion may lead to considerable overestimate of the yield of Japan Landslide Society, Maebashi, August 2001:
acceleration and therefore to unsafe seismic slope 219–222.

490
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Evaluation of rockfall hazards along part of Karaj-Chaloos road, Iran

A. Uromeihy
Dept. of Engineering Geology, Tarbiat Modares University, Tehran, Iran

N. Ghazipoor & I. Entezam

Engineering Geology Sect. Geological Survey of Iran, Iran

ABSTRACT: Karaj-Chaloos Road is a vital route that connects Tehran and the southern part of Alborz to the
northern part and the resort areas along the Caspian Sea. The road is located in the Central Alborz Mountains
where the occurrence of rockfall hazards is always considered to be very high. The event of May 28th 2004
earthquake triggered a large number of rockfalls along the road which caused human casualties and huge damages
to the infrastructures along the road. The aim of this paper is to evaluate the potential of rockfalls along part
of the route between Pol-e-zangoleh and Marzan-abad. The study area has a very complicated geological a
geomorphological conditions. Since most of the slope instabilities along the road are of rockfalls type, rockfalls
theory proposed by Evans and Hungr (1993) was used together with CONEFALL program (introduced by
Jaboyedoff, 2003) to analyze and to predict the potential of rockfalls along the road. Many factors such rock
type, slope morphology, drainage pattern and the seismic activity were considered in this research as main
affecting factors. Based on GIS program a rockfall hazard zonation map was prepared. It was found that about
45% of the rout has a high potential of rockfall activity and slope morphology is the most affecting factor for the
development of rockfalls. The influences of other factors such as rock type, fault alignment were found to be less.

1 INTRODUCTION paper. Figure 1, illustrates the location of the study

area along the main road. Due to the rough topography
Rockfalls are the most dominant type of slope insta- features and deep cuts along the road, most of rockfall
bilities in mountainous areas. They can cause a large hazards occurred along this part of the road.
number of casualties and huge economic losses along Initiation of rockfall events along the study area
road and other infrastructures. Rockfall can be defined can be sourced from different geological, geomor-
as the detachment and falling of rock blocks down the phologic, climatologic and human-related processes.
slope of a rock body. Earthquake also has a clear influence on trigger-
Evaluation of rockfall potential along the road in ing rockfall occurrence. Recently, the magnitude 6.3
mountainous areas was the interest of many resear- Baladeh Earthquake of May 28th 2004 is reported to
chers in recent years. For examples; Aksoy & cause 28 deaths due to the rockfall along the Chaloos
Ercanoglu, (2006), evaluate the rockfall initiation road (Bolourchi et al. 2007). Details of the earthquake
mechanism along highway in forest preservation and mechanism are investigated by Tatar et al. (2007).
urban settlement areas in Turkey. Batterson et al. Since most of the slope instabilities along the road
(2006), investigate the risk of geological hazards are of rockfalls type, rockfalls theory proposed by
including the rockfalls in Newfoundland, Canada. Evans & Hungr (1993) was used together with CONE-
Wasowski & DelGaudio, (2003), evaluate seismically FALL program (introduced by Jaboyedoff, 2003) to
induced mass movement hazard in Caramanico, Italy. analyze and predict the potential of rockfalls along the
Okura et al. (2000), studied the effect of rockfall road. Based on GIS data, a rockfall hazard zonation
volume on run-out distance in Japan. map was also prepared.
Karaj-Chaloos Road is a main route that connects
the Capital Tehran and the southern part of Alborz
Mountain to the northern part along the Caspian Sea. 2 GEOLOGICAL SETTING
Evaluation of rockfall potential along part of this road
between Pol-e-zangoleh and Marzan-abad, with a total Karaj-Chaloos road with North-South alignment cut
length of about 40 kilometers, was considered in this through the Alborz Mountain and passes various types

491
 Figure 2. Geological setting of the area between Pol-
e-zangoleh and Marzan-abad.

Figure 1. Geographical location of the studied area along

the Karaj-Chaloos road.
north-west south-east alignment such as Siahbisheh,
Table 1. Summary of geological units in the area. Vali-abad, and Dona faults. The second sets have
north-east south-west alignment named as Mekarood,
Geology time Formation’s names Rock type Dezbon, Mojlar and Marzan-abad faults and they dips
generally towards the south. The first and second sets
Pre-Cambrian Kahar, Soltanieh Sandstone, siltstone, of the faults are generally concentrated to the south-
shale, dolomite
ern and northern part of the area respectively, (Stoklin,
Paleozoic Zaigon, Lalon, Mila, Shale, siltstone,
Mobarak, Dorood, sandstone, dolo- 1974). Examples of rockfall potential along the road
Roteh mite, limestone, are shown on Figures 3a and 3b.
basalt
Mesozoic Elika, Shemshak, Limestone, shale,
Chaloos sandstone, coal,
tuff, basalt 3 CONEFALL THEORY
Cenozoic Karaj, Hezar-darreh Tuff, alluvial
Conefall theory is a simple way to demonstrate the
free movement of a block along a slope. A block can
propagate from its source point making an angle φp
of geological features and structures. The geological with horizontal. The space where a block can prop-
setting of the site consists of a sequence of sedimentary agate from a grid point is located within a cone of
rocks from Pre-Cambrian to Quaternary. The rocks slope φp with a summit placed at the source point.
include laminated layers of sandstone, shale, lime- It can be shown that the mean velocity of block is
stone and dolomite of different ages and some patches given as a function of the difference in altitude h
of volcanic rocks such as basalt and tuff. A summary between the cone and the topography. The details of
of the most dominant geological units are presented the theory can be found in Evan and Hungr (1993). A
in Table 1. The main geological features and rock computer program named as CONEFALL is designed
distribution of the area are also shown on Figure 2. by Jaboyedoff (2003), to estimate roughly the potential
There are two sets of thrust and reverse faults in of rockfall prone area. The principle of the program
the area, (Vahdati, 2001). They include faults with is to define a maximum run-out distance for block

492
propagation regarding the source point. The angle with 4 ZONATION OF ROCKFALL HAZARD
horizontal of the line joining the stop point and the
source φp may vary according to the slope of topog- In order to perform a precise evaluation of rockfall
raphy. Figure 4 shows the relationships between angle potential, the area was divided into three quadrants
φp and the maximum run-out distance. named as Mekarood, Dezbon and Pol-e-zangoleh from
north to south respectively as shown on Figure 5. The
required data for the analysis were collected from 47
stations along the road. The locations of each station
were selected regarding their past history of rockfall
occurrence and specially their response to the more
recent earthquake event of May 28th 2004.
The CONEFALL program was run twice consid-
ering two cone angle of 40 and 45 degrees. The
cone angles were selected regarding the common slope
angle along the road in the area. The results showed
that as the cone angle is increased the run-out distri-
bution of block at the base of the slopes decreases.
Figure 5 shows a rockfall hazard zonation map of the
area. The zonation map was correlated to other sheet
data by the aid of GIS program and following results
were obtained:
• Potential of rockfall is higher along the alignment
of main faults.
• The drainage pattern has no great effect on the
development of rockfalls.

Figure 3. Potential of rockfall along the Chaloos road.

Figure 4. Relationships between angle φp and the maximum Figure 5. Rockfall hazard zonation map along Chaloos
run-out distance (after Jaboyedoff, 2003). road (between Pol-e-zangoleh and Marzan-abad).

493
• Rockfall potential is higher where the Pre-Cambrian Batterson, M.J., McCuaig S. & Taylor, D.M. 2006. Mapping
rocks are outcropped. and assessing risk of geological hazards on the Northeast
• Slope angle of natural slopes and cuts along the road Avalon Peninsula and Humber Valley, Newfoundland. In
greatly influences the generation of rockfalls. Current Research, Newfoundland Department of Natu-
ral Resources, Geological Survey, Canada. Report 06–1,
147–160.
Bolourchi MJ. Entezam I. Mahmoudpour M. & Ansari F.
5 CONCLUSIONS 2007. Investigation of rockfall hazard in Chaloos road.
Report Number 85–5-1 Geological Survey of Iran. 209
The CONEFALL theory and program were used to pages, (in Persian).
evaluate the potential of rock fall hazards along part Evan, S. & Hungr, O. 1993. The assessment of rockfall haz-
of Chaloos road. The rockfall zonation map showed ard at the base of talus slopes. Canadian Geotechnical
that the rockfall potential is higher in places where Journal, vol. 30, pages: 620–636.
the elevation is higher, the cuts are deeper. Also it Jaboyedoff, M. 2003, CONEFALL 1.0 user guide, Open
was found that the development of rockfalls is directly report-soft 01, Quanterra, www.quanterra.org, 15 pages.
Jaboyedoff, M. & Labiouse, V. 2003. Preliminary assessment
related to the alignment of the main faults and other of rockfall hazard based on GIS data. Technology roadmap
fractures. The type of rock also showed great influence for rock mechanism, South African Institute of Mining and
on the generation of rockfalls. The number of rock- Metallurgy. 575–578.
falls in sandstone were distributed among the whole Okura Y. Kitahara H. Sammori T. & Kawanami A. 2000. The
areas while in limestone were restricted to the deep effect of rockfall volume on rumout distance. Engineering
cuts along the road path. The potential of rockfall Geology, Elsevier, vol. 58, Issue 2, 109–210.
is highest in Dezbon Quadrant where Pre-Cambrian Stoklin J. 1974. Northern Iran Alborz Mountain, in
rocks are dominant. While the distribution of rockfalls Mesozoic-Cenozoic orogenic belts, Geological Society
in Pol-e-zangoleh Quadrant comes in second place London, spec. pub. No. 4.
Tatar M, Jackson J. Hatzfeld D. & Bergman E. 2007. The
where Paleozoic rocks have greater outcrops. Finally 2004 May 28 Baladeh earthquake (Mw 6.2) in the Alborz,
the occurrence of rockfalls in Mekarood Quadrant is Iran: overthrusting the South Caspian Basin margin, par-
lowest in which the topographic reliefs are less and titioning of oblique convergence and the seismic hazard
Mesozoic rocks are more frequent. of Tehran. Geophysical Journal International, vol.170,
issue 1, 249–261.
Vahdati F. 2001. Geological description of Marzan-abad
REFERENCES map, scale 1:100,000. Geological Survey of Iran, Sheet
No. 6262.
Aksoy, H. & Ercanoglu, M. 2006. Determination of rock- Wasowski J. & DelGaudio V., 2003. Evaluating Seismically
fall sources in an urban settlement area by using a rule induced mass movement hazard in Caramanico Terme
base fuzzy evaluation. Nat. Hazards Earth Syst. Sci, (Italy). Engineering Geology, Elsevier, vol. 67, Issues
6, 941–954. 3–4, 281–296.

494
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Coupled effect of pluviometric regime and soil properties on hydraulic

boundary conditions and on slope stability

R. Vassallo & C. Di Maio

Dipartimento di Strutture, Geotecnica, Geologia Applicata, Università degli Studi della Basilicata, Italy

M. Calvello
Dipartimento di Ingegneria Civile, Università degli Studi di Salerno, Italy

ABSTRACT: The case of a landslide in Southern Italy was studied by modeling pore pressure distribution under
steady and transient conditions. The goal was to interpret the in situ piezometric and inclinometric measurements,
and to investigate the effects of the pluviometric regime on the slope stability. Steady state analyses show that
the hypothesis of zero pore pressure at the ground surface gives the best fit between theoretical and experimental
results, despite the nearly arid climate. The results of the transient state analyses show that the condition of
zero pore pressure is appropriate for both rainy and non-rainy periods because of the particular soil properties,
i.e. heterogeneity, compressibility and hydraulic conductivity. This condition generates a distribution of pore
pressures which can justify the observed horizontal displacements.

1 INTRODUCTION was carried out, including boreholes, undisturbed

sampling, laboratory testing for the determination of
This paper focuses on the study of an unstable slope index and mechanical properties of soils. In situ mon-
located in Tricarico (Fig. 1). The town of Tricarico itoring of pore pressures and horizontal displacements
is situated in Southern Italy, Basilicata Region, in the was carried out over several years.
hill zone of the Lucanian Apennine chain, north of The analysis presented herein is centred on the inter-
the Basento river. The historical centre rises on a cal- pretation of in situ measurements carried out on the
carenite plate which overlies a formation of Pliocenic Carmine slope, along section AA’ reported in Figure 1.
Clay. The newest quarters are mostly located in the Particular attention was devoted to the study of the
areas where a more recent clay formation outcrops and pore pressure distribution and of its influence on slope
overlies the calcarenites. stability.
This town has always suffered from slope instability
problems. In the 1950s a landslide in the Saracena area,
north of the town, caused severe damages to buildings 2 IN SITU AND LABORATORY
and infrastructures. As a consequence, the most part INVESTIGATIONS
of the community living there had to move to other
quarters in the 1960s and 1970s. Nowadays this zone As mentioned above, the subsoil of Carmine slope
is nearly uninhabited. consists of a clay formation overlying a calcarenite
In the zone of expansion, west of the town, on the formation. In this zone, five continuous coring bore-
Carmine slope, many buildings constructed in the last holes were driven, each equipped with two Casagrande
ten years suffered significant deformations. The most piezometers. These boreholes are indicated with the
serious damages occurred in some buildings situated letter ‘‘S’’ in Figure 1. Two further continuous cor-
north-west, near the line where calcarenites outcrop. ing boreholes (i.e. S9 and I9 in the same figure) were
The eastern sector of the town was only affected equipped with full length inclinometer casings. By
by local instability problems and fissuring of road rapidly extracting small soil specimens from the inner
pavements. part of remoulded borehole samples and sealing them
In October 2002 the University of Basilicata started hermetically (i.e. undrained conditions) it was possible
a study of the above mentioned three zones of to measure water contents. The obtained in situ water
Tricarico, funded by the Basilicata Administrative content profiles are reported in Figure 2. Five core-
Region. A very extensive geotechnical investigation destruction boreholes were also driven, each equipped

495
Figure 1. Plan view of the investigated zone.
Figure 3. Grain size distribution of soil samples from the
Carmine slope.

S11 S14 S10

w (%) w (%) w (%)
0 15 30 45 0 15 30 45 0 15 30 45
0 0 0
respectively) was found and the location of their inter-
5 5 5
face was in good agreement with borehole results. As
Organic soil and/or
10 10 10
fill material it will be shown, the integration of geophysical surveys
15 15 15 Soft or medium cla with the geotechnical investigation was quite useful to
20 20 20
Stiff grey clay model the pore pressure distribution of the slope with
Very stiff grey clay a good reliability.
25 25 25
Clayey sand Many undisturbed samples were used in the labo-
30 30 30
Marly clay
ratory to measure index properties and to carry out
35 35 35
Calcarenite
mechanical testing, in particular oedometer, triaxial
40 40 40
and direct shear tests. The fine-grained soil ranges
from clay with silt to clayey-sandy silt (Fig. 3). In the
45 45
Casagrande chart this soil is classified as high plastic-
50 50 ity inorganic clay. In the activity chart most data points
are located between the Na-montmorillonite and the
illite lines. Peak friction angle from triaxial tests ranges
Figure 2. Stratigraphy and water content profiles for bore- between 17◦ and 24◦ and residual friction angle from
holes S11, S14 and S10.
direct shear tests ranges between 7◦ and 14◦ .
The results of the consolidation stages of oedomet-
ric tests were used, under the hypothesis of the Terzaghi
with a full length inclinometer, indicated with the letter one-dimensional model, to estimate values for the
‘‘I’’ in Figure 1. coefficient of consolidation cv and for the hydraulic
The clayey material is about 42 m thick in bore- conductivity k in the axial direction and to investigate
holes S11 and S14 and 27 m thick in borehole S10, the influence of the void ratio, e, on those parame-
as reported in Figure 2. Verticals S11, S14 and ters. The variation of k with e observed experimentally
S10 belong to section AA’, used to analyze the pore (Fig. 4), along with the significant reduction of poros-
pressure distribution and the stability conditions. ity with depth indicated by the water content profiles
Electrical Resistivity Tomographies (ERT) were reported in Figure 2, suggested that the clay forma-
also used to recognize the position of the contact tion could be divided into two sub-layers. The upper
between clay and calcarenite where no direct observa- layer, about ten meters thick, can be characterized
tion was available. A longitudinal tomography cross- by an hydraulic conductivity one order of magnitude
ing verticals S14 and S10 and several transversal greater than the lower layer. It will be illustrated in
tomographies were carried out (Fig. 1) as described the following how this hypothesis was corroborated
in detail by Perrone et al. (2007). A very clear differ- by an extensive parametric study of the pore pressure
ence of electrical resistivity between an upper and a distribution within the slope and by the comparison of
lower material (associated with clay and calcarenite, calculation results with in situ measurements.

496
 1·10-9 3 PORE PRESSURE ANALYSIS
I9C1
I9C2 3.1 Steady state analyses
1·10-10 I9C3
Figure 7 shows the 2-D model used for pore pressure
S10C1
calculations and the position of the verticals S11, S14
S10C2 and S10 with piezometers. The interface between clay
1·10-11 S10C3 and calcarenites was initially obtained schematically
S10C4 by drawing a straight polyline through the limit found
S12C1 in boreholes S11, S14 and S10 and the points were cal-
1·10-12
carenites outcrop. However, this choice was validated,
S12C2
at a later stage, by tomography results.
S12C3 Several analyses were performed using the finite
1·10-13
element code SEEP/W (by Geoslope International
0.2 0.4 0.6 0.8 1.0
Ltd), considering different boundary conditions and
assuming different values of the anisotropy ratio. For
Figure 4. Hydraulic conductivities obtained by interpreting each analysis, the calculated values of water pres-
oedometer test data. sures were compared with the values measured at the
internal points of the domain.
At first, the numerical simulations were carried out
by considering only two homogeneous and isotropic
materials: the upper soil (clay) was characterized by
0 a conductivity three orders of magnitude lower than
S11upp
5 [@9.5m] calcarenite. Zero water pressure was assumed at the
10 S11low ground surface. Lateral boundary conditions able to
15 [@42.5m] optimize the results for the lower piezometers were
S14upp found and left unchanged in subsequent analyses.
20 [@9m]
S14low
These boundary conditions are also shown in Figure 7.
25
[@43m] Then, an heterogeneity within the clay was intro-
30
S10upp duced, based on the observed significant reduction
35 [@12m] of porosity with depth and on the relationships found
40 S10low
experimentally between conductivity and porosity.
[@28.5m]
45 An upper layer of altered soil, about 10 m thick, was
distinguished from a lower, less permeable, one. The
latter was given a conductivity three orders of magni-
tude lower than calcarenites. A parametric study of the
Figure 5. Piezometric measurements. hydraulic properties of the clay formation was carried
out, looking for those which could minimize the dif-
ferences between measured and calculated total heads.
As it was mentioned above, two Casagrande Figure 8a plots the difference  between mea-
piezometers were placed at different depths in each sured and calculated values as a function of the ratio
of five verticals of the Carmine slope. Piezometric kupp /klow of the conductivities of the two clay lay-
measurements started in May 2003 and were carried ers, considered isotropic. Figure 8b plots the same
out frequently for about two years. Subsequent less difference as a function of the anisotropy ratio kx /ky
frequent measurements were also carried out in 2006 of the whole clay layer considered as homogeneous
and 2007. Figure 5 shows the depth of the piezometric (the direction x of maximum principal conductivity
level from the ground surface as a function of time for was assumed parallel to the slope). The conditions
the six piezometers considered in this study. One can kupp /klow = 1 and kx /ky = 1 are relative to homo-
observe that total heads decrease significantly with geneous and isotropic clay having a conductivity 1000
depth along the verticals of boreholes S11 and S14. In times lower than the conductivity kc of the calcarenite.
particular, piezometer S11low , after a short initial tran- As the ratio kupp /klow increases, || at first signif-
sient, always resulted empty. Inclinometric readings icantly decreases and then stays practically constant
started in November 2003 and the last measurements for a ratio greater than 20. On the other hand, as the
were taken in April 2007. The observed displacements ratio kx /ky increases, assuming kx = 10−3 • kc , ||
of the ground surface since the installation of the incli- increases. Therefore, as it is also highlighted by the
nometric casings are about 3 cm for borehole I11, 4 cm square root of the sum of squares plotted in the same
for borehole I14 and 1 cm for borehole I10. The pro- figures, the hypothesis which gives the best agreement
files reported in Figure 6 show that most displacements with the in situ measurements is that of heterogeneous
take place in the upper 10 meters. and isotropic clays, with kupp /klow = 20. It is worth

497
 660m 0 1 2 3 4 cm

640
0 1 2 3 4 cm
620

600 0 1 2 3 4 cm
I11
580

560 I14
50m
540 I10
Calcarenite upper limit

Figure 6. Profiles of horizontal displacements in April 2007 (zero: November 2003).

S11upper Conductivity values

measured hmax= 651.42m
calculated h = 650.94m klow = 1 · 10 -3 kc
S11lower S14upper kupp = 2 · 10 -2 kc
measured hmax < 609.5m measured hmax = 622m
calculated h = 599.98m calculated h = 621.2m
S14lower S10upper
measured hmax = 594m measured hmax = 592m
Upper clay (k=kupp) calculated h = 594.0m calculated h = 592.16m
S10lower
Lower clay (k=klow) u=0 measured hmax = 587m
calculated h = 586.71m

Calcarenite (k=kc)
h=const.
600

h=const.
598

596

594

592

590

586
588

584

582
∂h/∂n=0
/ n=0

Figure 7. Results of a steady state analysis: optimized solution.

S11upp S14upp S14low S11upp S14upp S14low

S10upp S10low SRSS S10upp S10low SRSS
Square Root of the Sum of Squares (m)

10 60 10 60
Square Root of the Sum of Squares (m)

(a) (b)
50 50
0 0

40 40
-10 -10
30 30

-20 -20
20 20

-30 10
-30 10

kx (clay) = 10-3 kc (calcarenite)

-40 0 -40 0
1 10 100 1000 1 10 100 1000
k upp / k low
kx / ky

Figure 8. Effect of heterogeneity (a) and anisotropy (b).

498
noting that this value is in agreement with that coming 654
from oedometer tests and water content profiles. The 653
total head contours and the calculated total head values 652

at the piezometers, versus the measured ones, are also 651

650
shown in Figure 7 for the final and optimized solu-
649
tion. The very good agreement between calculations
648
and measurements proves the accuracy of the proposed 0 50 100 150 200 250 300 350
time (days)
interpretation.
The validity of the chosen lateral boundary condi-
tions was further proved by some analyses. Different
boundary conditions were used above the phreatic Figure 9. Effect of a 6 months dry period followed by a 6
months wet period. The starting condition is a steady state-
surface in the calcarenites to account for possible fis-
distribution of pressures obtained using u = 0 at ground
suring and for unsaturated flow. The domain was also surface.
extended on both sides up to where the calcarenite
outcrops and a significant reduction of the hydraulic
conductivity above the phreatic surface was taken into
two horizontal lines plotted in this graph represent
account for this soil. These changes did not signifi-
the minimum and the maximum total head recorded
cantly influence the results at the piezometers and a
during the period of measurement. Total head at the
very good agreement with the measurements was still
ground surface starts with a u = 0 condition, then
obtained.
decreases as a consequence of the dry period and then
immediately returns to a u = 0 condition as a conse-
3.2 Transient analyses quence of rain. Because of this boundary condition, in
As mentioned above, the best agreement between cal- the 9.5 m deep-piezometer the total head decreases
culations and measurements is obtained with u = 0 at slowly during the dry period and then returns to it
the ground surface. Given that the clay is water satu- initial value more rapidly during the wet period. In
rated up to the surface, this condition would certainly other words, the condition u = 0 prevails. Further-
derive from a steady rainfall with an intensity greater more, total head always stays in the range between the
than the soil hydraulic conductivity. Yet, since the cli- maximum and the minimum measured values. In the
mate is nearly arid in Tricarico, further analyses were other piezometers, the effect of dry periods is even less
carried out to evaluate which other process can make significant.
u = 0 at the ground surface as the most significant Further analyses were carried out using different
condition. To this aim, the transient flow caused by ‘‘rain functions’’, i.e. different cycles of wet and dry
a cycle of alternating dry periods (normal unit flux periods, starting both from a wet period- and a dry
qn = 0 at ground surface) and wet periods (u = 0 at period- steady state distribution of pore pressures.
ground surface) was studied. These analyses included the case of 4 dry weeks fol-
Two different steady state cases, i.e. u = 0 and lowed by 4 wet weeks and that of 3 dry weeks followed
qn = 0 at the ground surface, were used as initial con- by 1 wet week. The results are reported in Figure 10 for
dition of the transient analyses. The conductivity of the S11upp , which was still the piezometer most influenced
lower clay was chosen two orders of magnitude higher by the imposed rain functions.
than the value measured in the laboratory because of For the analyses which started from a wet period-
in situ large-scale heterogeneity, fissuring, etc. The steady state (Fig. 10a), the total head always fluctu-
conductivities of the other two materials were calcu- ates around values close to that relative to u = 0 at
lated from the same ratios used in steady state analyses. the ground surface. On the other hand, the analyses
Therefore, the selected values were kupp = 2·10−8 m/s which started from a dry period- steady state (Fig.
10b) indicated that after a few years of alternating wet
and klow = 1 · 10−9 m/s for intact and altered clay,
and dry periods the total head again fluctuates around
respectively, and kc = 1 · 10−6 m/s for calcarenite.
values close to that relative to u = 0 at the ground
Upper and lower clays were assumed to have the same
surface. Therefore, the condition u = 0 at the ground
compressibility, deduced from oedometer test results.
surface confirms to be appropriate even during dry
The calcarenite was assumed to have a compressibility
periods.
significantly lower than the clay one.
A first analysis was carried out considering a
6-month dry period followed by a 6-month wet period,
starting from the steady state condition which resulted 4 STABILITY ANALYSIS
from u = 0 at the ground surface. The evolution of
the total head at the ground surface and in the upper The trend of total head contours shown in Figure 8 is
piezometer at vertical S11 is reported in Figure 9. The clearly a consequence of the subsoil stratigraphy and

499
 654 S11 (a)
653 (a) S14
652
S10
651
650
649
648
0 50 100 150 200 250 300 350 400 450 50
time (days)
ground surface f.1 ground surface f.2 2.0
S11upp f.1 S11upp f.2 (b)
h max h min

safety factor
660 1.5
650
640
630 1.0
620
610
600 0.5
(b) 10 15 20
590
0 50 100 150 200 250 300 350 400 450 500
time (days)
surface 1 (5m deep) surface 2 (7.5m deep)
ground surface f.1 ground surface f.2 surface 3 (10m deep) surface 4 (12.5m deep)
S11upp f.1 S11upp f.2 surface 5 (15m deep)
h max h min

Figure 11. Results of stability analyses: (a) tentative slip

Figure 10. Effect of cyclic successions of wet and dry
surfaces; (b) safety factor as a function of friction angle.
periods. The starting condition is a steady state- distribution
of pressures obtained using: (a) u = 0 at ground surface and
(b) qn = 0 at ground surface. Rain functions are indicated peak friction angles measured on undisturbed samples
as f.1 (4 dry weeks followed by 4 wet weeks) and f.2 (3 dry extracted from boreholes S11, S14, S10.
weeks followed by 1 wet week).

5 CONCLUSIONS
of the hydraulic properties which were inferred by a
combined use of boreholes, tomography and labora- The results of pore pressure analyses showed that the
tory measurements. A confirmation of the validity of interpretation of in situ measurements is satisfying if
the chosen model came from the comparison between a 10-meters thick altered-clay layer characterized by
computed and measured pore pressures. greater void ratio and conductivity than the lower clay
A detailed study of pore pressure distribution was is considered. The condition u = 0 was imposed at
essential to carry out an accurate slope stability the ground surface, despite the nearly arid climate
analysis. Figure 11 shows for several hypothetical of Tricarico. The transient analyses proved that the
slip surfaces the safety factor (lower value between soil properties of the studied slope, i.e. heterogeneity,
those obtained from Janbu and Morgenstern & Price along with compressibility and hydraulic conductiv-
methods) as a function of the friction angle. ity values, make the condition of zero pore pressure
The safety factor stays almost constant with the appropriate for both rainy and non-rainy periods. The
depth of the slip surface in the upper clay layer. On obtained trend of total head contours significantly
the other hand, it considerably increases with depth in influences the safety factor, which is lower in the
the lower clay layer. This is a clear consequence of the altered clay layer, in agreement with the inclinometric
particular pore pressure distribution. In fact, the upper measurements performed over the last 4 years.
layer has a pore pressure regime quite similar to that
of an infinite slope, with lines parallel to the ground
surface. Differently, total head contours in the lower REFERENCES
clay layer are nearly horizontal, and this results in pres-
Perrone, A., Piscitelli, S. Lapenna, V., Loperte, A., Di Maio,
sures lower than those which would be predicted for an
C. & Vassallo, R. 2007. Electrical resistivity tomography
infinite slope at the same depth. This seems in agree- and geotechnical techniques for the stability analysis of
ment with the displacements measured during almost the Tricarico landslide. Thirteenth European Meeting of
four years, which are more relevant in the uppermost Environmental and Engineering Geophysics, Instanbul,
10 meters. Turkey, 3–5 September 2007.
The mobilized friction angle on the slip surface at
the interface between upper and lower clay is about
17◦ . This value is within the range of residual and

500
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Mechanical characters of relaxing zone of slopes due to excavation

H. Wang
College of Environment and Resources; Fuzhou University, Fuzhou , China

X.P. Liao
Northwest Research Institute of China Railway Engineering Corporation, Lanzhou, China

ABSTRACT: The analysis for relaxing zone is very important for stability analysis of slopes, and it is difficult
to confirm the boundary of relaxing zone because of excavation. Based on a typical case study by finite element
analysis for excavation, the mechanical character and the distribution regulation for relaxing zone of slopes,
especially the variety of differential principal stress and plastic zone, are researched in this paper. The results
shows that: the change of differential principal stress tensor which is obtained by subtracting the stress tensors
before and after excavated can be used to confirm the range of relaxing zone; and the distribution and development
of the negative differential principal stress zone is almost the same as the plastic zone; in the end, the stress
attenuation plastic relaxing zone near excavation surface, the original stress zone far from excavation surface
and the stress concentration elastic compressing zone between above two are partitioned.

1 GENERAL had been used to confirm the width of unloading

zones. The excavation disturbed zone in the permanent
The analysis for relaxing zone is very important shiplock slopes of the Three Gorges Project had been
for stability analysis and design of slopes, and it studied using investigate, engineering physics explo-
is attached importance to by researcher for rock ration, test and monitor by J.H. Deng et al. (2001).
mechanics at home and abroad. Recently, with the The relation between the longitudinal wave speed of
large-scale development of infrastructure in China, rockmass, rockmass deformation modulus and the
especially hydropower high slopes and the deep stress of rockmass had been studied after rock slope
cut slopes along expressway, some research results excavated using the dynamics theory of wave, and
about relaxing zone of slopes can be summarized as then the change of rockmass deformation had been
follows: researched, the thickness of rockmass relaxing band
Unloading rockmass mechanics had been founded had been predicted.
by Q.L. Ha et al. (2001) by systemically research- In general, the research on the relaxation of rock
ing anisotropy, strength theories, scale effect, and slope till now is based on qualitative analysis and
rheology. Estimate method for excavation disturbed semi-quantitative analysis, and is now in the stage
zone of slopes had formed using numerical simulation of research independently and express respectively.
technique and engineering test method by Q. Sheng Some important problems about the relaxation should
(2002), and the relative research result had been used to be deeply studied, such as the production pro-
in the Three Gorges Project. The formation principle cess of relaxing zone, the evaluation method for the
of unloading zones of high slopes had been analyzed reducing effect for strength parameter of rock or soil,
by R.Q. Huang (2001), and the basic regularities of and the method to confirm the boundary of relaxing
secondary stress distribution due to excavation had zone because of excavation. To deep into the rela-
been studied. The stress field and the relative dis- tive research, the basic concept and physical meaning
placement field of the deep cut slope had been studied of relaxation due to excavation are discussed in this
by S.G.Xiao (2003), and the saltation position of dis- paper firstly, and then the mechanics principle of relax-
placement or safety factor within influence area due ation have been emphatically studied using numerical
to excavation. The distributing range and the variety simulation technique; in the end, the partition stan-
character of the relaxing zone in centrifugal model dard, confirm method on the relaxing zone have been
test had been researched, and the displacement value expounded.

501
2 THE BASIC CONCEPT OF THE RELAXING
ZONE FOR SLOPES DUE TO EXCAVATION

The relative research on the relaxation of rockmass at

home and abroad began from the excavation of the tun-
nel, and the relaxing zone, the compaction zone and the
original rock zone usually had been partitioned mainly
based on the change of stress for rockmass. Similarly,
the relaxation of slope is refer to the behavior or the
phenomena about the relaxation of rockmass because
of the change of rock stress during excavating. The def-
inition of the relaxing zone of slopes due to excavation Figure 1. Computing mesh of slope excavation by FEM.
can be comprehended as the zone near the excavation
surface within which the rockmass stress level reduces,
frees, releases or adjusts to a new balance level. And In this paper, the variety of differential principal
the phenomena of the relaxation are refer to the swell stress tensor and the change regulation or distribu-
deformation near the excavation surface, the stretch tion of plastic zone are meanly discussed to clarify
and broaden of rock joints or the produce of new joints, the mechanical character of the relaxing zone due to
the loosening of rock or soil, the producing of plastic excavation. It is necessary to define the differential
zone, and the failure of slopes locally or holistically. sigma 1 which is obtained by subtracting the stress
According to the comparison of theory analysis, tensors and calculating the differential principal stress
numerical simulation and case test, the basal viewpoint tensor at each node. And differential sigma 1 is not
on the mechanical character and its engineering effect obtained by simply subtracting the stress magnitudes,
about the relaxing zone due to excavation have formed, since this would not account for changes in principal
and can be summarized as follows: stress balance with stress orientation at different stages. The variety of
the rockmass has been breached during excavating, the differential sigma 1 and the plastic zone of slopes
then the stress state has been adjusted, so the relaxing during all the process of excavation are showed from
zone come to being. During this process, the stress Figure 2 to Figure 11.
level within the local area beyond the load limitation, Firstly, we study the evolvement regulation of the
the deformation and the failure forms relevantly, and differential sigma 1 and the plastic zone of slopes dur-
then the physical and the mechanical character have ing excavating. When the 1st and the 2nd excavation
been weakened, and the failure of rock mass has been step, stress concentration appear near the foot of the
accelerated. slope, but the excavated slope surface is still in the
elasticity state, also we can say the excavated slope
is in the elastic state. When the 3rd excavation step,
stress concentration increase near the foot of the slope,
and some area of rockmass is in the plastic state. In
3 THE MECHANICAL CHARTER OF the other hand, we can find that the plastic zone is
RELAXING ZONE DUE TO EXCAVATION in the upper and the elastic stress concentration zone
is in the lower of the foot of the slope, and these two
To study the mechanical charter of relaxing zone of zones are coterminous. We can say the excavated slope
slopes, a finite element method calculation model for is in the plastic state within local area. When from
slope excavation has been discussed in this paper. This the 4th to the 7th excavation step, the plastic zone
model has been excavated in 8 steps to form a slope expands obviously, and the differential sigma 1 within
of 64 m highness, the slope angle is 63 degree. The the plastic zone is negative, near the plastic zone is
elastic-plastic model which use Mohr-Coulomb yield the elastic stress concentration zone within which the
criterion as the yield function has been used to simu- differential sigma 1 is positive, and the area of plas-
late rockmass, and the calculation has been performed tic zone or stress concentration zone all enlarge. We
with Phase 2 software. This analysis model have some can say the excavated slope is in the extending plas-
assumed condition, which can been expounded as tic state. When the 8th excavation step, the plastic
follows: (1) only considering the gravity field; (2) con- zone is continuous, the max shear stain and volume
sidering the model as problem of plane strain, (3) joints stain develop directionally, and the slip surface come
and ground water are not considered in this model, to being (as shown in the figure 12). In this time, the
(4) using strength reduce method to calculate the safety area within which the differential sigma1 is negative
factor of slopes. The computing mesh can be drawed or positive are all enlarge, and the change value all
as figure.1, and the relative parameters are listed in increase exquisitely. The safety factor is equal to 1.0.
table 1. We can say the excavated slope is in the failure state.

502
 Table 1. Rockmass parameters of slope excavation model.

Unit Young’s Poisson’s Tensile Cohesion Friction Friction angle Dilation

weight modulus ratio strength Cohesion (residual) angle (residual) angle
γ /kN · m−3 E/kPa υ T /kPa c/kPa c/kPa φ/◦ φ  /◦ ψ/◦

25 500000 0.3 5 100 85 40 35 20

Sigma 1
kPa
–600.00
–536.00
–472.00
–408.00
–344.00
–280.00
–216.00
–152.00
–88.00
–24.00
40.00
104.00
168.00
232.00
296.00
360.00

Figure 5. Yield zone of the 3rd excavation step.

Figure 2. Differential Sigma 1 of the 2nd excavation step.

Figure 6. Differential Sigma 1 of the 5th excavation step.

Figure 3. Yield zone of the 2nd excavation step.

Figure 7. Yield zone of the 5th excavation step.

Figure 4. Differential Sigma 1 of the 3rd excavation step.

the value decrease with the distance to the excavated

Secondly, we study the distributing area of the dif- surface increasing. This zone can be defined as the
ferential sigma 1 and the plastic zone. There are local stress attenuation plastic relaxing zone. Near the plas-
area near the excavated surface within which the differ- tic relaxing zone, there are local area within which
ential sigma 1 is negative, and the final failure surface the differential sigma 1 is positive, the maximal dif-
appears within which. As shown in the figure 12, the ferential sigma 1 is near 360 kPa, and the grads of the
minimum differential sigma 1 is near −600 kPa, the change value is very great. This zone can be defined
minimum value lay near the upper foot of slopes, and as the stress concentration elastic compressing zone,

503
 Sigma 1
kPa
-600.00
-536.00
-472.00
-408.00
-344.00
-280.00
-216.00
-152.00
-88.00
-24.00
40.00
104.00
168.00
232.00
296.00
360.00

Figure 8. Differential Sigma 1 of the 7th excavation step.

Figure 11. Yield zone of the 8th excavation step.

Figure 9. Yield zone of the 7th excavation step.

Figure 12. Max shear strain and stress trajectories.

Figure 10. Differential Sigma 1 of the 8th excavation step. Figure 13. Variety of Differential Sigma 1 in the level
referenced line.

and these two zones above transform quickly. With the far from excavated surface in which there are a lit-
distance to excavated surface increasing, there are a tle change of the differential sigma 1, we call it the
wide area within which the stress change little, and the original stress zone.
deformation are also little. This zone can be defined According to the analysis expounded above, and
as the original stress zone. To deeply comprehend the combining the case test about tunnel and slope at
meaning of these three zones, we study the variety home and abroad, the stress attenuation plastic relax-
of differential sigma 1 in the level referenced line as ing zone near excavation surface, the original stress
shown in the figure 13. Within the area that the distance zone far from excavation surface and the stress concen-
to excavated surface is less than 8 m, the minimum tration elastic compressing zone between above two
differential sigma 1 is about −450 kPa, showing con- can be partitioned. The rockmass in the plastic relax-
suming relaxing effect of rockmass. With the distance ing zone step into plastic state, joints splay and slip
increasing to 12 m, the differential sigma 1 transform within which, and the volume of rock mass increases,
to zero, this zone is the transition of plastic zone and the elastic wave speed value decreases, the carry-
elastic zone. With the distance increasing to 20 m, the ing capacity of relaxing zone weakens, and then the
differential sigma 1 increase rapidly to 320 kPa, and reside strength can be used to simulate their mechani-
then with the distance increasing to 45 m, the differen- cal character. The rockmass in the stress concentration
tial sigma 1 decrease slowly to zone. This zone shows zone is in the elastic always, and the joints in which
the stress concentration effect of rockmass. In the area close yet, the volume of rockmass decrease, the elastic

504
wave speed value increase, and the carrying capac-
ity increase too. For safety considering, the initial
carrying capacity evaluation can be used in numer-
ical simulation. The rockmass in the original stress
zone is ultimately changeless because it is disturbed a
little.

4 THE PARTITION STANDARD

AND CONFIRM METHOD OF THE
RELAXING ZONE DUE TO EXCAVATION

As summarized above, the differential sigma 1, plastic

zone and the weaken of mechanical character are the
basic characteristic for the relaxing zone due to exca-
vation, and can be used as the partition standard. It is
need to point out that the designer ordinarily lack case
test data because of little money or difficulty to test, Figure 14. Flow chart to confirm the relaxing zone due to
so we can use the basic regulation of the relaxing zone excavation.
which are researched from theory analysis, numerical
simulation, and case test to establish its partition stan-
dard and confirm method. The partition standard of the
stress attenuation plastic relaxing zone, the stress con-
centration elastic compressing zone and the original
stress zone is shown in the table 2.
To enhance the veracity and the applicability of this
partition principle, it is necessary to build a applied
confirm method and operation flow on the relaxing
zone. The basic flow during design process to con-
firm the boundary of relaxing zone of slopes due to
excavation is shown in the figure 14.
In the process of confirming the relaxing zone,
data processing is the most important flow. In gen-
eral, the differential sigma 1 is the most representative Figure 15. Distribution chart of relaxing zone of slopes due
to excavation.
parameter. Strictly speaking, the field within which
the differential sigma 1 is negative can be confirmed as
the stress attenuation zone, and the field within which differential sigma 1 is more than 50 kPa can be con-
the differential sigma 1 is positive can be confirmed firmed as the stress concentration elastic compressing
as the stress concentration zone. But considering the zone; the field between above two is the transitional
computing error of FEM, we recommend 50 kPa as field, and the field far away these two zone is the origi-
the partition standard. Also we can confirm the field nal stress zone. The distribution chart of relaxing zone
within which the differential sigma 1 is less than of slopes due to excavation is as shown in figure 15.
−50 kPa can be confirmed as the stress attenuation From the figure 15, we can see this analysis method
plastic relaxing zone, and the field within which the can simply confirm the boundary of relaxing zone,

Table 2. Partition principle for relaxing zone of slopes due to excavation.

The stress attenuation The stress concentration The original

Partition principle plastic relaxing zone elastic compressing zone stress zone

Differential sigma 1 negative positive Approximate

equal to zero
Yield state and strain Plastic, large strain Elastic , small strain Elastic, little
character strain

505
and its physical meaning are clear. It is need to point tion elastic compressing zone between above two
out the plastic zone is almost the same as the distri- are partitioned.
bution and development of the negative differential
principal stress zone, and the maximal shear strain
and the deformation field in the state of limit equilib-
rium are very typical, these character can be used as REFERENCES
the accessorial criterion.
D.X. Nie. 2004. The study on rock mass deforming para-
menters and relaxing thickness of rock high slope.
5 CONCLUSION Advance in Earth Sciences. 19 (3); P472–47.
J.H. Deng, Z.F. Li & X.R. Ge. 2001. Disturbed zones and dis-
Based on the relative research on the relaxing zone for placement back analysis for rock slopes. Chinese Journal
of Rock Mechanics and Engineering. 20 (02); P171–174
tunnel and slope at home and abroad, the mechanical [in Chinese].
character of relaxing zone of slopes due to excavation Q.L. Ha. 2001. Study on the anisotropic unloading rock
has been studied using numerical simulation, and the mass mechanics for the steep-high rock slope of the three
partition standard and confirm method of the relax- Gorges Project permanent shiplock. Chinese Journal of
ing zone has been built, the main conclusion can be Rock Mechanics and Engineering. 20 (05); P605–610 [in
summed up as follows: Chinese].
Q. Sheng. 2002. Excavation disturbed zone of deep cut-
1. The definition, phenomena, mechanical character ting rock slopes and mechanics behaviour of engineering
and its engineering meaning on the relaxing zone rock mass. The dissertation for doctor degree of Institute
for slopes due to excavation has been enucleated. of Rock and Soil Mechanics; The Chinese Academy of
2. The change of differential principal stress tensor Sciences. P48–68 [in Chinese].
which is obtained by subtracting the stress tensors R.Q. Huang, F. Lin & D.J. Chen. 2001. Formation mech-
before and after excavated indicates the essence of anism of unloading fracture zone of high slopes and its
relaxing zone by discussing the analysis result of engineering behaviors. Journal of Engineering Geology.
9 (03); P228–229 [in Chinese].
numerical simulation. S.G. Xiao & D.P. Zhou. 2003. Determination and numer-
3. The distributing and regulation of differential ical analysis method of relaxation region for cutting
sigma 1, plastic zone and the shear stain can be used slope. Journal of Southwest Jiaotong University. 38 (03);
to as the partition standard of the relaxing zone. P318–321 [in Chinese].
4. Based on the mechanical character of relaxing X.Y. Zhao, H.T. Hu & L.X. Pang. 2005. Study on unloading
zone, the stress attenuation plastic relaxing zone effect and width of unloading zones in excavating of soil-
near excavation surface, the original stress zone far like material slopes. Chinese Journal of Rock Mechanics
from excavation surface and the stress concentra- and Engineering. 24 (02); P710–711 [in Chinese].

506
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Deformition characteristics and stability evaluation of Ganhaizi landslide

in the Dadu River

Yunsheng Wang, Yaoming Sun, Ou Su, Yonghong Luo & Jiuling Zhang
National Key Laboratory, Chengdu University of Technology, Chengdu, China

Chunhong Zhou & Shengfeng Zhang

East China Investigation and Design Institute Under China Hydropower Engineering Consulting Group Corporation,
Hangzhou, China

ABSTRACT: Ganhaizi landslide, about 19790 × 104 m3 , is located on the right bank of the Dadu River. There
are two cracks and one creep body in the accumulation of Ganhaizi landslide. The potential slippage volume of
the accumulation involved by each deformation body is more than 10 × 104 m3 . Cracks 1 and 2 are located in
the middle part of Ganhaizi landslide, while the creep body is situated at the south of the landslide body. Several
springs are found to the southwest of the staggered platform. The spring flows perennially and keeps limpid in
flood period. The accumulation of Ganhaizi landslide is so huge that it may frighten the upstream hydropower
project (3 km away). Therefore, its stability evaluation is extremely important. The analysis shows that rainstorm
or earthquake will induce Ganhaizi landslide only partly reactive, but the whole landslide is stable. The area
of potential instability is small, and it cannot surge or dam Dadu River again. In a word, the accumulation of
Ganhaizi landslide has little influence on the hydropower project.

1 INSTRUCTION on top of the back edge is Eluo village platform. The

average grade of this platform is about 10◦ , the most
The Ganhaizi landslide, situated on the right of the width of the platform is 700 m, and this platform is
Dadu River, is a giant cut-bedding one, with an accu- primarily covered by Gazha colluvial accumulation.
mulation volume about 15950 × 104 m3 . It once The boundary of landslide at upstream side is new set-
dammed Dadu River. The road across the front of the tlement; Its downstream boundary is located in Guan
landslide sank apparently. Additionally, there are two stacked village of Bawang town downstream is a steep
deformed cracks and one creep body in Ganhaizi land- to gentle common boundary in land form.
slide. The characteristics above shows the landslide is On the basis of the interior terrain of landslide, its
partly under the deformation. Therefore, the landslide relief is descending from north to south. The north
stability is vital to the upstream hydropower station (upstream), located in north to Xiaobawang village,
construction which is just 3 km away. is a protuberant platform. This platform extends from
the north boundary of landslide to the back edge of
landslide, its altitude is between 2130 m to 2200 m, the
largest width of this platform is 520 m, and the average
2 THE CHARACTERISTICS OF THE
gradient is 10◦ ; The south relief is low, but the area is
LANDSLIDE
big, and its area occupies 2/3 to the landslide body
surface area. Civilian houses of Xiaobawang village
2.1 The basic characteristics of the landslide
repose on a depression to the south of the platform,
Ganhaizi landslide, on the right bank of Dadu River, and its gradient is comparative gentle.
lies at downstream of Bawang ditch. The front edge There is a gentle platform (Xiaobawang platform) in
altitude is 1950 m, its back edge altitude is 2540 m, the back edge of Ganhaizi landslide downstream from
its length is 1580 m from east to west, and its average west to east (Figure 2), whose altitude is about 2200 m
width is 2070 m from south to north. The area of the and the average gradient is 12◦ . Under the altitude
landslide is 3190 × 103 m2 , the average thickness is of 2200 m, there gradient of the terrain is steep and is
about 50 m, and its volume is about 15950 × 104 m3 . about 20◦ . From the altitude of 2200 m at Xiaobawang
According to the field investigation (Figure 1), the platform to the one of 2050 m at the front edge of
back edge altitude of Ganhaizi landslide is 2540 m, landslide is an accumulating platform, its middle part

507
 Smx4-1 Cutline
Smx5
Lower member of the fourth
Qcol+dl+el Qal
Smx4-3
Smx
4-1
formation of Maoxian Group of
Silurian

A Smx4-2 4-2
Middle member of the fourth
A' Smx formation of Mao-clan Group of
Silurian
Genba

Qsef+col valley
4-3
Upper member of the fourth
formation of Maoxian Group of
Smx
Silurian
Ganhaizi landslide
Qcol
Qdel
The fifth formation of Maoxi
Smx5
B Eluo village Smx4-2 an Group of Silurian

Smx
4-1 Qcol+dl B'
C C' The section line

Dadu
The Guan stockaded village River
of Bawang town 4-2 Qel+dl The boundary of landslide
Smx Qcol+dl

Figure 1. The plan of Ganhaizi landslide.

Cutline
Extrapolated Initial
form line
Extrapolated form
sef
Q line after sliding
Present form line
Smx3
Smx
del Extrapolated present
Q Slide surface
sel col
Q Muscovite-biotite-
quartz schist
4-1
Smx Smx
4-2
Quartzite

Boulder gravel

Figure 2. The A-A section of Ganhaizi landslide.

gradient is about 8◦ , the front gradient of the platform We find the checked-up accumulation nipped in the
facing river changes from 30◦ to 35◦ . middle layer of the sand and pebble of the terrace II
There is few gullies at the superficial of Ganhaizi downriver of Yan’eryan valley in the left bank of Dadu
landslide, no wide incised gullies. River, which shows that the landslide took place when
Associating with clay layer of checked-up accumu- the terrace formed. According to regional time-dating,
lation in Bawang (Figure 4), we conclude that Ganhaizi the second-level terrace formed between late Pleis-
landslide once damed Dadu River, as Figure 2, bro- tocene to initial Holocene, according to time dating of
ken line indicates the conjecture terrain before the the clay, Ganhaizi landslide formed 14∼17 thousands
landslide slide, dot line indicates the shape of accu- years age.
mulation slide later. The Dadu River undercut and
increased the free face in the front edge of slope,
2.2 The characteristics of the composition and the
owing to channel was narrow, and the right bank was
slide bed of the landslide
high and steep (the altitude of shoulder of slope is
2400 m), it is easy to check-up Dadu River (Chen The accumulation of landslide takes on echelon lon-
De-chuan, 2004). By the time dating, the checked-up gitudinally, in top of which is a wide-gentle platform,
accident happened in Pleistocene, and dammed the its average gradient is 12◦ (Figure 3). This platform
river for long time. Finally, the river unceasingly cor- has thin deluvium in superficial, whose gravels mainly
rades the accumulation to form the present terrain (real vary from 20 cm to 30 cm. Below the platform, its aver-
line). age gradient is steep, approximately 29◦ , and the slope

508
 Altitude
(m)
The platform of Ganhaizi
small Bawang landslide Fiercely
weathering Dadu River
pseudo-bed
rock

Q col+dl
S mx4-1 Q del
4-2 4-2
S mx S mx

Distance (m)
Middle member of the Lower member of the fourth
cutline Extrapolated boundary
of quaternary overlay
4-2
S mx fourth formation of
S mx4-1 formation of Mao-clan muscovite-biotite-quartz
schist
quartzite boulder gravel
Mao-clan Group of Group of Silurian
Silurian

Figure 3. The B-B section of Ganhaizi landslide.

one reaches 80 cm, its lithology is muscovite-biotite-

quartz schist. The surface of the platform is the slope
wash, in which the average broken particle diam-
eter is about 10 cm and the slope wash has some
stratification which thickness is about 0.5 m. The
vegetation on it is poor. Most part of the Ganhaizi
landslide is located on the stratum of Smx4-1 , Smx4-2 ,
the lithology of which is mainly garnet muscovite-
biotite schist, quartzite and has few laminated marble
additionally.
The attitude of the bedrock in the district of land-
slide is N40∼60◦ W, NE∠50◦ ∼65◦ basically, and the
whole dip of stratum is steep to upstream, and the
intersectant angle to river is 45◦ . The back cliff of land-
slide is overlaid with Gazha colluvial and deluvium
Figure 4. Silt-clay layer of checked-up accumulation in the accumulation in which there appears a few bedrock,
north of Bawang village. the bedrock in the back cliff lithological character-
istic is quartzite and garnet muscovite-biotite-quartz
schist of Smx4-1 and Smx3 . The left (upstream) is cov-
ered with Gazha colluvial and deluvium accumulation,
is composed of lots of blocky stones, about 40% to the the bottom rock is garnet muscovite-biotite-quartz
total, especially the gentle part in the middle of accu- schist and quartzite of Smx4-1 . The right (downstream)
mulation downstream, its average particle diameter widely spread bedrock, and its lithology is grey and
between 50 cm to 70 cm, and the biggest can be reach grayish brown color garnet muscovite-biotite schist,
3 m. The component of block and debris in the slope is muscovite-biotite-quartz schist and biotite schist of the
mainly composed by quartzite and muscovite-biotite- stratum of Smx4-2 .
quartz schist. The low district downstream side is According to the adit, the belt of slide is composed
covered by thicker soil layer, the vegetation and arbor is of clay with gravels, about 30 cm thick. The site test
rich, the soil layer contains few block and broken stone, shows the φ of the belt is 23◦ , and C is 0.001 MPa.
its particle diameter is between 5 cm to 10 cm, and its
lithology is muscovite-biotite-quartz schist and garnet
muscovite-biotite schist. The Xiaobawang village is 3 THE DEFORMED CHARACTERISTICS
located in this region. OF THE LANDSLIDE
The middle of the Ganhaizi landslide downstream is
a protuberant platform, which makes up of broken and The chief deformed characteristic of Ganhaizi land-
blocky stone with silty soil, and partly includes huge slide is that there are two deformed cracks and one
block with original rock sequence that weathering creep body. The crack lies in the middle of Gan-
and broken fiercely, its lithology is muscovite-biotite- haizi landslide in the right bank of river upstream,
quartz schis. The average particle diameter of block the avalanching locates in front of the remained land-
stone in the slope is about 20 to 30 cm, and the largest slide in the left bank of river, and the creep body is

509
on the right bank of Ganhaizi landslide downstream scarp below the platform upstream, extending 50 m
(Figure 5). ahead northwest, the slide direction is towards north-
No.1 fracturing crack: It is located in the middle part east, and after sliding it formed a sunk ditch, the width
of Ganhaizi landslide towards its upstream, presented of which is 50 m, and the depth is 40 m, the surface
echelon. It presents the fracturing crack directed to grade is about 30 to 35◦ . The back cliff of this creep
N30◦ E near the Daodaoyi, the width is about 20 to body had been fractured, and formed two staggered
30 cm, the eyeable depth is over 1 m, the length is over level scarp, each scarp highness is about 2 m, and the
50 m, and this crack vanishes to the Dujize house. potential slippage volume is 10 × 104 m3 . We found
The landslide superficial layer sliding partly brings on several springs to the southwest of the staggered scarp,
the crack, and the crack grows in the interior of the the distance is about 30 m. The volume of each spring
landslide, where the average grade is about 20◦ . There outlet is approximately 0.01 l/s. After convergence,
is a scarp in the front of the crack with the distance the volume is about 0.05 l/s, the water is colorless and
about 100 to 150 m between them. The grade of the tasteless, its temperature is about 15◦ C. The spring
scarp is between 30◦ and 35◦ , the height is about 40 flows perennially, and keeps limpid in flood period.
to 50 m. This crack is caused by the part slide of the Accordingly the downstream concave trench
superficiality of landslide. Its extending is long, and formed as early as the trench upstream, thus, the reac-
its controlled area is big, besides the stability is weak tivation of this creep body is that Dadu River corroded
in present, therefore, if this part fails, the estimated laterally and earthquake or other geological action
volume is 15 × 104 m3 . induced (Li Xue-ping & Tang Hui-ming 2005).
No.2 fracturing crack: It is located in the middle
part of colluvial and deluvium accumulation of Gan-
haizi landslide towards its upstream, extending south
4 THE STABILITY EVALUATION
to north, and vanishing after prolonging 10 m. This
OF LANDSLIDE
crack mainly developed at the back of the isolated
stone, its particle diameter is about 2 to 3 m, and
According to the investigation of Ganhaizi landslide
the depth is about 2 m, the width is 5 m. Generally,
and limit equilibrium method, the whole Ganhaizi
there is a scarp in front of the crack approximately
landslide is stable (K is 1.56 in nature state), but the
5 to 10 m. According to the characteristics of this
superficiality may be kept on sliding locally (Chen
crack, we know that this crack is caused by the falling
Dong-liang, et al. 2002), for example, the No.1 frac-
isolated rock impinging against the superficial uncon-
turing crack upstream, but the potential unstable area
solidated colluvial accumulation or by the landslide
is small, and its volume is not big, consequently, the
superficial deformed, and this fracturing crack con-
influence to the hydropower station is smaller. The
trols the deformed area finitely, thus, it has little effect
stability of the remained accumulation of Ganhaizi
to the stability to the whole landslide (Xu Hua, et al.
landslide in the left bank of Dadu River is poor. The
2005).
colluvial accumulation in front edge of Ganhaizi land-
Creep body No.3 (Figure 5) is situated at the south
slide is unconsolidated, and the gradient of free face
of the landslide body. The back edge situated in the
of main scarp is quite steep, those factors may led to
secondary fall or slide below the platform of Genba
valley.
Dadu
Smx 4-1 River

Smx 4-3 5 CONCLUSION

1
Eluo village Genba valley
The remained Ganhaizi landslide in the two bank of
Q sef+col Dadu River is likely to reactive partly, and the chief
Q del 2 deformed characteristic of Ganhaizi landslide is that
Ganhaizi landslide there are two deformed cracks, one avalanching and
one creep body. No.1 fracturing crack extending is
long, and the area it controlled is big, thereby, the
Smx 4-2 stability is lower in present. No.2 fracturing crack con-
cutline
trols limited area, it has little effect to the stability to
1 fall and its number 3
crack and its number al
the whole landslide. The creep body will react on the
2
Q function of rainstorm or earthquake, its stability is low.
Analysis above shows that rainstorm and earth-
Figure 5. The distribution of deformed phenomena in quake will induce Ganhaizi landslide part slide or fall
Ganhaizi landslide. secondarily, but the whole stability is well. The area

510
of potential instability is small, and the volume is lit- the Three Gorges Reservoir area [J], Hydrogeology and
tle, it cannot surge or dam Dadu River. In a word, the Engineering Geology, 4(4), 28–31.
remains accumulation of Ganhaizi landslide has little Li Xue-ping & Tang Hui-ming, 2005. Application on Likeli-
influence on the hydropower project. hood Ratio Test Used for Sensitivity Factors of Area Slope
Stability [J], Journal of Yangtze River Scientific Research
Institute, 10(5), 37–48.
Chen Dong-liang, Ying Jing-hao & Yin Guo-sheng, 2002.
REFERENCES Analysis on the Forming Cause Mechanism of Loess
Landslide in the Area of Crossing Yellow River Project
Chen De-chuan, 2004. Stability study on Bawangshan land- along the Middle Route of South to North Water Trans-
slide near ErtanReservior [J]. Sichuan Water Power, 9(3), fer Project [J], Journal of North China Institute of Water
21–23. Conservancy and Hydroelectric Power, 4(4), 49–51.
Xu Hua, Li Tian-bin & Xiao Xue-pei, 2005. Formation
mechanism and stability prediction of Andu landslide in

511
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Landslide-prone towns in Daunia (Italy): PS interferometry-based

investigation

J. Wasowski & D. Casarano

CNR-IRPI, Italy, Bari, Italy

F. Bovenga & A. Refice

CNR-ISSIA, Italy, Bari, Italy

R. Nutricato & D.O. Nitti

Dipartimento Interateneo di Fisica, Politecnico di Bari, Italy

ABSTRACT: Persistent Scatterers Interferometry (PSI) and satellite radar imagery can be used to detect
very slow displacements (mm-cm per year) of targets (PS) exhibiting coherent radar backscattering properties
(mainly man-made structures). Here we present results of the PSI application to the Daunia Apennines, which
include many hilltop towns affected by landslides. Examples from the towns Casalnuovo Monterotaro and
Pietramontecorvino are used to illustrate that the interpretation of PS data on urbanised slopes can be difficult,
because their movements may arise from a variety of processes: i) volumetric strains within soils, ii) natural
or anthropogenic subsidence or uplift, iii) settlement of engineering structures, iv) deterioration of man-made
structures, v) extremely slow slope deformations that may or may not lead to failure. Where true landslide
movements are detected, they likely regard long-term post-failure displacements involving clay-rich materials.

1 INTRODUCTION failure hazard, PSI results can be used to provide a

preliminary distinction between conditions of stability
Since slope failures are a world-problem, an effec- and instability.
tive approach to landslide hazard reduction seems to However, with the exception of urban landslides,
be through exploitation of Earth Observation systems, the density of persistent radar targets suitable for
with focus on long term monitoring, early detec- interferometric measurements on unstable rural and
tion and warning (e.g. IGOS GEOHAZARDS 2004; peri-urban slopes is typically low and this makes
Wasowski et al., 2007a). In particular, space-borne difficult PSI analysis, as well as introduces uncertain-
synthetic aperture radar differential interferometry ties in the assessments of true ground motions (e.g.
(DInSAR) techniques are useful in landslide investiga- Bovenga et al., 2006; Colesanti & Wasowski 2006).
tions, because of their capability to provide wide-area Furthermore, the interpretation of the exact geotech-
coverage (thousands km2 ) and, under suitable con- nical significance of millimetric to centimetric yearly
ditions, spatially dense information on small ground displacements currently detectable by PSI can be diffi-
surface deformations (e.g. Colesanti et al., 2003). Fur- cult, because i) such slow ground surface deformations
thermore, advanced multi-pass DInSAR methods such may arise from a wide variety of causes and, there-
as Persistent Scatterers Interferometry (PSI; Ferretti fore, their presence on slopes may not always reflect
et al., 2001) and Small Baseline Subset (SBAS; Lanari shear movements or occurrence of landslides, and
et al., 2004) overcome the limitations of conventional ii) most radar targets correspond to man-made objects
DInSAR and extend the applicability of radar inter- (e.g. houses) and thus their structural behaviour (and
ferometry from regional to local-scale engineering ground-foundation interactions) should be taken into
geology investigations of ground instability (Colesanti account.
& Wasowski, 2006; Ferretti et al., 2006). PSI tech- In this study we present the results of the appli-
niques can be used to detect and monitor very slow cation of the SPINUA PSI technique (Bovenga
displacements of targets (PS) exhibiting coherent radar et al., 2004) to landslide-prone towns in the Daunia
backscattering properties. Since the presence of slow Apennines (Southern Italy). We expand upon the
deformations represents evidence of potential slope initial works of Bovenga et al. (2005, 2006) by

513
 By using C-band radar data, the detectable
velocities range theoretically from about 1 mm/y up
to 29 cm/y; however, recent assessments (Crosetto
et al., 2007) indicate that displacement rates lower than
1.5–2 mm/y and higher than 15–20 cm/y are hardly
measurable in real experimental conditions.
In spite of the above-mentioned limitations, the pos-
sibility to measure, relatively quickly, deformations
on slope surfaces over wide areas (103 ÷ 106 km2 ),
coupled with the regular update capabilities of satel-
lite sensors, opens new perspectives in regional
scale detection of slope hazards and monitoring of
slow landslides (Colesanti et al. 2003; Colesanti &
Wasowski 2004, 2006; Farina et al., 2006; Hilley
et al. 2004; Wasowski et al., 2007a).
Figure 1. Filtered SAR amplitude image of the Daunia area
(28 km × 27 km) showing the predominantly moderate relief
hillslope topography. In addition to the towns of Casalnuovo 3 THE STUDY AREA
Monterotaro (marked by C) and Pietramontecorvino (P), the
white rectangles (few km in width) enclose several other The area studied is located in the Daunia region
towns selected for the Permanent Scatterers (PS) analysis. (Southern Italy), characterized by gentle hills and low
Inset shows location of Daunia in Southern Italy. mountains, only locally exceeding 1000 m above sea
level. Daunia belongs to the highly deformed area
between the frontal thrusts of the Apennine chain
and the western-most part of the foredeep (Dazzaro
providing additional analysis of the urban areas of
et al., 1988). The chain units are characterised by a
Pietramontecorvino and Casalnuovo Monterotaro
series of tectonically deformed flysch formations of
(Fig. 1). We also focus on some difficulties in inter-
pre-Pliocene age.
preting the exact origin of the PS movements detected
The clay-rich flysch units are more prone to land-
by exploiting the European Space Agency (ESA)
sliding, compared to the formations containing higher
ERS-1/2 satellites data.
proportion of lithoid intercalations (sandstones, lime-
stones). The widespread presence of clayey materials
with poor geotechnical properties is the underlying
2 BACKGROUND ON PSI cause of landsliding. Furthermore, as a result of
the tectonic history of the Apennines, the geologi-
Detailed information on PSI techniques is available cal materials are intensely deformed and hence also
in specialised remote sensing publications (e.g. Fer- rock units are susceptible to slope movements. In gen-
retti et al., 2001), and its practical applicability has eral, the activity of landslides in the Daunia Apennines
already been addressed in engineering geology lit- is characterized by seasonal remobilisations of slope
erature (e.g. Colesanti et al., 2003; Bovenga et al., movements, typically related to rainfall events. Indi-
2006). Here we mention only some basic aspects of vidual meteoric events have been the most frequent
DInSAR and PSI. For a recent comprehensive review triggers of landslides, even though the mean annual
of radar-based remote sensing and PSI applications rainfall is modest (in the order of 600–700 mm per
for landslide assessment, the reader is referred to year).
Colesanti & Wasowski (2006). Although mass movements appear widespread
Space-borne synthetic aperture radars are active throughout the entire region (Zezza et al., 1994), there
microwave systems capable of recording coherently are relatively few studies published on landslides in
the electromagnetic signal backscattered from the Daunia. The better documented events are concen-
Earth surface. With two or more SAR images acquired trated within or in the immediate proximity of the
over the same area during successive satellite passes urban areas. In the 1990’s there has been an appar-
it is possible to detect ground surface movements ent increase in landslide activity in several urban and
occurred between the SAR acquisitions along the Line peri-urban areas. It is probable that the stability of
Of Sight (LOS) direction. For the ERS satellites (inci- slopes bordering the hilltop towns has gradually wors-
dence angle 23◦ ), the LOS unit vector has director ened because of residential development over recent
cosines of about 0.9 (up–down), 0.4 (east–west), and decades. This has led in some cases to reactiva-
0.05 (north–south), respectively. Thus, the maximum tions of pre-existing old landslides. Furthermore, the
sensitivity is for vertical displacements. urban expansion onto marginally stable hillslopes and

514
improper land use has led to increases in damaging
first-time failures (Wasowski et al., 2007b).

4 PSI ANALYSIS OF TWO TOWNS

To illustrate the potential of the PSI technique as well

as current difficulties in PS data interpretation we
present the results concerning two towns: Casalnuovo
Monterotaro and Pietramontecorvino (Fig. 1). These
two cases are considered representative of the geologic
and geomorphologic conditions encountered in other
towns of Daunia.

4.1 PSI processing

The initial analysis (Bovenga et al., 2005) involved an
area of 28 × 27 km2 , enclosing 10 towns affected by
slope instability problems (Fig. 1). We used a SAR
dataset of both descending and ascending ERS-1/2
acquisitions (40 and 35 scenes, respectively) covering
the period 1995–1999. The dataset was pre-processed
adopting optimized solutions for co-registration, rela-
tive calibration, and re-sampling (see Bovenga et al.,
2005, 2006 for more details). The analysis was lim-
ited to small image windows of size ranging from 5 to
15 km2 (Fig. 1). The windows enclose towns, exclud-
ing vegetated rural areas that typically contain very few
PS. This strategy was adopted to ensure an adequate Figure 2. Distribution and average line of sight (LOS)
distribution of PS candidates needed for a reliable cor- velocity of radar targets (PS marked by black symbols) in
rection of the atmospheric signal. Such approach was Casalnuovo Monterotaro. Negative and positive velocities
also considered as most practical given that landslides indicate, respectively, movements away from and towards the
with high socio-economic impact in the study area are sensor; PS within the velocity range −2–2 mm/y are assumed
motionless. Background image is a 1997 orthophoto. Loca-
those affecting urban centres. tions of relict and recent landslide scarps (dashed and dotted
white lines, respectively), are after Zezza et al. (1994).
A small watercourse draining the valley at the eastern periph-
4.2 Interpretation of the PS results: The case ery of the town is shown in light gray. Note also ascending
of Casalnuovo Monterotaro radar satellite acquisition geometry (white arrows).
Figure 2 shows that the great majority of PS, which
falls within the built up area, results to be stable. This
is not surprising as the central part of the town devel- indicative of either downward (subsidence) or down-
ops along a flat topped N-S trending ridge made of slope (eastward) movements, or a combination of both.
the Flysch di Faeto Formation. This Miocene age fly- The presence of moving PS in the east-facing head
sch, consisting of an alternation of marly limestones, portion of the valley, in the vicinity of old landslide
calcarenites and marly clays, is relatively less prone scarps, could suggest landslide origin for the detected
to landsliding than the so-called Argille Bentonitiche movements. However, in cases like this, in situ mon-
(clay-rich unit of Miocene age including bentonite itoring data are needed to demonstrate whether the
clays). The latter crop out at the eastern and western predominant displacements are indeed in downslope
peripheries of the town (Zezza et al., 1994). or vertical directions.
There are, however, two small groups of moving PS Nevertheless, inspections of the slope stability con-
situated in the eastern periphery of the town, on gently ditions in the eastern part of the town (conducted in the
inclined head portion of a local valley (Fig. 2). These recent years for the Department of Civil Protection)
PS show low velocity (from −3 to −4 mm/y) move- revealed that several buildings and retaining walls,
ments. The negative sign stands for the displacements including those with moving PS, show signs of distress
away from the radar sensor and, given the ascending and have suffered recurrent damage (cracks) since the
satellite acquisition geometry, can be interpreted as 1990s. This confirms the reliability of the PS results.

515
Because the clusters of moving PS are small, they unit, also referred to as Varicoloured Clays Zezza
probably point to local site instabilities rather than to et al. (1994), is known for its high suceptibility to
true landslide movements. The cut and fill re-shaping landsliding.
of the valley head during the post-second world war Pietramontecorvino landslide legacy has been
development of the town has obliterated the evidence examined on a detailed scale by Zezza et al. (1994),
of landslide legacy, but the presence of artificial fill who identified and mapped very old and more recent
mantling the clay-rich slope substratum could be a landslide scarps, quiescent, mappable movements, as
cause of the local ground settlements and structure well as active and quiescent landslide zones (Fig. 4).
instability. Variations in local drainage conditions and However, it is apparent that the pattern of moving
in water input to the slope could also play an important and motionless PS bears no specific relation with the
role by inducing volumetric changes in the soil. distribution of pre-existing landslides. One possible
exception can be identified at the southern periphery of
the town, where a significant number of slowly moving
4.3 Interpretation of the PS results: The case PS falls within and near the limits of a N-S elongated
of Pietramontecorvino landslide zone (Fig. 4), interpreted as active by Zezza
The PS pattern in the Pietramontecorvino area is much et al. (1994). Our field inspections showed that the
different from that of Casalnuovo Monterotaro. Sev- recent landslide activity in this area is conditioned by
eral zones in the town centre and at its northern and the erosion activity of a local torrent at the slope base.
southern outskirts include clusters of moving radar tar- However, again, the PS one-dimensional LOS
gets (Fig. 3). Indeed, Pietra Montecorvino is not a motion data alone are insufficient to resolve the nature
hilltop town, because it develops mostly on a SSE fac- of the observed displacements, i.e. whether they repre-
ing slope mantled by large landslides. However, Fig. 3 sent predominantly vertical or downslope movements
does not reveal any obvious link between the distribu- or a combination of both. The fissures observed on
tion of moving PS and the landslides. As in the case of
the landslides, both moving and stable PS are present
also in the slope areas occupied by the Flysch Rosso
Formation of pre-Pliocene age (Fig. 3). This clay-rich

Figure 3. Distribution and average line of sight (LOS)

velocity of radar targets (PS marked by black symbols)
in the Pietramontecorvino area. Negative and positive Figure 4. LOS velocity of radar targets (black symbols) in
velocities indicate, respectively, movements away from the Pietramontecorvino area. Background image is a 1976
and towards the sensor; background geological map from airphoto showing photo-interpreted (after Zezza et al., 1994)
http://www.apat.gov.it: clay-rich Flysch Rosso Formation landslide features including: old scarps (semicircular barbed
(in grey colour) of pre-Pliocene age and three old landslides lines, dashed if uncertain), quiescent, mappable movements
(in white). General downslope direction is to SSE. Note (with Vs and Us standing, respectively, for slides and flows)
also descending radar satellite acquisition geometry (black and active and quiescent landslide zones (respectively cross-
arrows). hatched and hatched areas).

516
 of the PS movements (e.g. local ground settlements
versus true slope movements or structure instability).
In some instances, however, even a simple site
inspection can allow for a straightforward interpreta-
tion of PS data. This is the case of a single moving PS
situated near the SW corner of the town’s football field
(Fig. 5). Its position coincides with a small rotational
slide: we inspected the site only after obtaining the
PS results and found a semi-circular scarp with minor
settlement of the field ground, as well as a locally
rotated gabion (Fig. 6). The field is also guarded
by metallic wire-frame sustained by iron poles: these
objects could be associated with the PS behaviour and
they were locally deformed by the slide. Although
the time of the initial failure is not known, the PS
displacement time series (Fig. 5) indicate that the
movements were taking place in the period covered
by radar imagery (1995–1999). Interestingly, this is
the fastest (−8 mm/y) moving radar target in Pietra-
montecorvino. It seems that in this case PSI allowed
us to capture the slow post-failure rotation of the slide
head, and that the PS motion reflects predominantly
sub-vertical displacements.
Figure 5. Top: close-up on the southern portion of the Nevertheless, interpretations based on a single PS
town centre with local distribution and average LOS velocity should generally be viewed as a limit case. Clearly,
of radar targets (black symbols), superimposed on a recent
clusters of moving radar targets are needed for a
ortophoto of Pietramontecorvino. White arrow points to a
small landslide. Bottom: Displacement time series of the PS reliable detection of ground instability.
located near the slide.

5 DISCUSSION AND CONCLUSIONS

The PSI results indicate that the majority of PS in

Casalnuovo Monterotaro are stable; these PS are con-
centrated within the historical centre of the town. Very
slowly moving radar targets are present in two clus-
ters situated at the eastern border of the town. This is
in agreement with the observation that displacements
related to landslide events or simply to ground instabil-
ity occur mainly in the peripheries of the Daunia hilltop
towns, where the urban expansion is more recent and
the man-made structures acting as PS are located close
to the potentially unstable slopes (Bovenga et al.,
2006).
In the case of Pietramontecorvino the deformations
detected by PSI are present not only at the town’s
Figure 6. Small rotational slide at the SW corner of Pietra- peripheries, but also in the centre. Indeed, this town
montecorvino football field, whose movements were detected is known for the presence of instability problems
by PS interferometry (cf. Figure. 5). Note semi-circular scarp affecting many buildings. Although a major part of
with minor ground settlement. Pietramontecorvino appears to be sited over a large,
very old landslide, we consider unlikely that the slowly
moving PS (3–5 mm/y) reflect its activity. Instead,
some buildings confirm the persistence of the con- it is suggested that processes such as settlements of
ditions of ground and/or structure instability, but the engineering structures, volumetric changes and post-
surface expression of the past landslide events has failure creep of the clay-rich Pietramontecorvino soils
been substantially altered by recent town development and colluvia, could be responsible for the localised
and man activity. Clearly, in situ investigations would deformations detected by PSI. Nevertheless, a link
be needed to distinguish between the possible causes seems to exist between the moving radar targets and

517
recent landslide activity in the southern periphery of Colesanti, C. & Wasowski, J. 2006. Investigating land-
the town (Fig. 5). slides with satellite Synthetic Aperture Radar (SAR)
Even though in several cases the PS displacement interferometry. Engineering Geology, 88 (3–4), 173–199.
fields show clear evidence of moving objects in urban Colesanti, C., Ferretti, A., Prati, C. & Rocca, F. 2003. Moni-
and peri-urban areas, local knowledge of the investi- toring landslides and tectonic motions with the Permanent
Scatterers Technique. Engineering Geology, 68 (1), 3–14.
gated area and site inspections are required to interpret Crosetto, M., Agudo, M., Raucoules, D., Bourgine, B.,
the significance of PS motion data and to identify the de Michele, M., Le Cozannet, G., Bremmer, C.,
mechanism of the detected deformations. In general, Veldkamp, J.G., Tragheim, D., Bateson, L. & Engdahl,
on slopes, surface displacements over time might be M. 2007. Validation of Persistent Scatterers Interferome-
found to be in a downslope direction but such defor- try over a mining test site: results of the PSIC4 project,
mations might not necessarily always reflect shear Proc. ENVISAT Symposium, Montreux, Switzerland (CD-
movements or movements leading to shear failure, i.e. ROM).
to landsliding. With the exception of ‘‘natural’’ PS Dazzaro, L., Di Nocera, S., Pescatore, T., Rapisardi, L.,
(e.g. corresponding to rock outcrop targets), without Romeo, M., Russo, B., Senatore, M.R. & Torre, M. 1988.
Geologia del margine della catena appenninica trail F. For-
an appropriate in situ investigation, several different tore ed il T. Calaggio (Monti della Daunia—Appennino
interpretations of the very slow PS displacements are Meridionale). Mem. Soc. Geol. It. 41: 411–422.
possible. Our PS results showing ground surface defor- Farina, P., Colombo, D., Fumagalli, A., Marks, F. &
mation changes over time on landslide susceptible Moretti, S. 2006. Permanent Scatterers for landslide
slopes are very promising. However, the geotechni- investigations: outcomes from the ESA-SLAM Project.
cal parameters and geological boundary uncertainties Engineering Geology, 88 (3–4), 200–217.
which control PS displacements need to be inves- Ferretti, A., Prati, C. & Rocca, F. 2001. Permanent Scatter-
tigated and better understood before they can be ers in SAR Interferometry. IEEE Trans. Geoscience and
used directly for landslide hazard/risk zonation or for Remote Sensing, 39 (1), 8–20.
Ferretti, A., Prati, C, Rocca F. & Wasowski, J. 2006. Satel-
predicting (warning) of potential instabilities. lite interferometry for monitoring ground deformations
in the urban environment. Proc. 10th IAEG Congress,
Nottingham, UK (CD-ROM).
ACKNOWLEDGEMENTS Hilley, G., Burgmann, R., Ferretti, A., Novali, F. & Rocca F.
2004. Dynamics of Slow-Moving Landslides from Per-
This work was supported in part by the European Com- manent Scatterer Analysis. Science, 304, 1952–1955.
munity (Contract No. EVGI 2001-00055—Project IGOS GEOHAZARDS 2004. GEOHAZARDS theme report:
LEWIS). Images were provided by ESA under the For the monitoring of our Environment from Space and
CAT-1 project 2653. from Earth. European Space Agency publicatiom, 55 p.
Lanari, R., Mora, O., Manunta, M., Mallorqui, J.J.,
Berardino, P. & Sansosti, E. 2004. A Small Baseline
Approach for Investigating Deformations on Full Res-
REFERENCES olution Differential SAR Interferograms. IEEE Trans.
Geoscience And Remote Sensing, 42, 1377–1386.
Bovenga, F., Refice, A., Nutricato, R., Guerriero, L. & Wasowski, J., Ferretti, A. & Colesanti, C. 2007a. Space-
Chiaradia, M.T. 2004. SPINUA: a flexible processing Borne SAR Interferometry for Long Term Monitoring of
chain for ERS / ENVISAT long term interferometry, Pro- Slope Instability Hazards. Proceeding of the First North
ceedings of ESA-ENVISAT Symposium 2004, Saltzburg, American landslide Conference, Vail, USA. (CD-ROM).
Austria (CD-ROM). Wasowski, J., Casarano, D. & Lamanna, C. 2007b. Is
Bovenga, F., Chiaradia, M.T., Nutricato, R., Refice, A. & the current landslide activity in the Daunia region
Wasowski, J. 2005. On the application of PSI technique to (Italy) controlled by climate or land use change? Proc.
landslide monitoring in the Daunia mountains, Italy, Pro- International Conference on ‘‘Landslides and Climate
ceedings of FRINGE 2005, ESA-ESRIN, Frascati, Italy Change—Challenges and Solutions’’, Ventnor, UK,
(CD-ROM). 41–49.
Bovenga, F., Nutricato, R., Refice, A. & Wasowski, J. 2006. Zezza, F., Merenda, L., Bruno, G., Crescenzi, E. &
Application of multi-temporal differential interferometry Iovine, G. 1994. Condizioni di instabilità e rischio da frana
to slope instability detection in urban/peri-urban areas. nei comuni dell’Appennino Dauno Pugliese. Geologia
Engineering Geology, 88 (3–4), 218–239. Applicata e Idrogeologia, 29, 77–141.
Colesanti, C. & Wasowski, J. 2004. Satellite SAR interfer-
ometry for wide-area slope hazard detection and site-
specific monitoring of slow landslides. Proc. Interna-
tional Landslide Symposium—ISL2004 Rio de Janeiro,
Brasil, 795–802.

518
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Basic types and active characteristics of loess landslide in China

Weijiang Wu, Dekai Wang & Xing Su

Geological Hazards Prevention Institute, Gansu Academy of Sciences, Lanzhou, Gansu, China

Nianqin Wang
Department of Geology and Environment Engineering, Xi’an University of Science and Technology, Xi’an,
Shanxi, China

ABSTRACT: Loess landslide disasters are widely distributed in loess area in china. Four kinds of basic types
can be divided which are homogeneous loess landslide, loess interface landslide, loess-mudstone layer plane
landslide and loess-mudstone cutting layer landslide. The characteristics of loess interface landslide and loess-
mudstone layer plane landslide are low sliding velocity, short sliding distance and the sliding body has low
stability after sliding. The characteristics of homogeneous loess landslide and loess-mudstone cutting layer
landslide are high sliding velocity, long sliding distance and the sliding body has high stability after sliding,
except its scarp. Basic laws mentioned above have some guiding significance and practical value in economic
construction.

1 INTRODUCTION 2 TYPE CHARACTERISTIC

One of the most basic classification of landslide is The rock-soil is not only the material base of a land-
based on the main material the sliding body composed slide developing, but also the problem should be
of [1] . Loess landslide which is widely distributed in the identify firstly in research and prevention of the land-
Loess Plateau in northwest of China is one of the major slide. Therefore, the classification method based on
special soil landslide types. But some large landslides the material composition is the best method to reflect
are not composed entirely of loess, and contained a the basic characteristics of the landslide. Controlling
lot of soft rock formed in Neogene, Cretaceous and so by the regional composition of lithology conditions, in
on[2][3][4] except for the loess in the different periods. addition to the loess in different periods which devel-
Because they are in loess region some scholars also ops pure loess landslides, argillaceous rock formation
call them loess landslide. Thus, loess landslides types which is widespread distributed in the Neogene, the
in the traditional concept are different with the actual Cretaceous and Jurassic of the Mesozoic and Ceno-
situation. zoic Erathem, is also an easy-sliding stratum because
The most of difficult questions are to forecast the of the poor cementation and low mechanical strength
sliding distance and hazard areas when deal with in the Loess Plateau of the northwest in China. So
the dangerous situation of the landslide. If the types many landslides in the region, especially some large
and activity characteristic can not be realized fully, landslides whose sliding body are often composed of
to reduce the disaster lose furthest would be diffi- loess and the underlying soft rock joint, are mixed type
cult. Generally, loess landslide may collapse, mostly landslide, has gone beyond the traditional concept of
belongs to high-speed or high-speed long-distance the loess landslide which is composed of the pure loess.
landslide. In fact, to different types of loess landslide The concept in broad sense and classification of loess
the difference of sliding and resurrection are quite dif- landslide are needed. Its basic characteristics of the
ferent. Therefore, it is significance for bring forward various types of loess landslides can be summarized
the new concept of loess landslide in broad sense and in Table 1.
studying the characteristics of different types of loess Comparing to the traditional concept of loess land-
landslides. This can help to have the correct under- slide, The new classification is more in line with
standing and forecasting of deformation and damage, the actual development of landslide in the loess area
activities and hazard areas of landslide. and make the concept be more clearly. At the same

519
Table 1. The basic types and characteristics of loess landslide.

Types of loess Position of the Constructed

landslide sliding belt profiles

Pure loess landslide Homogeneous The main sliding surface

loess landslide develops in loess
formation.

Loess inter-face The main sliding surface

land-slide is in the interface of
high water content or
saturated loess and
mudstone.

Mixed loess land-slide Loessmudstone The main sliding surface

layer plane develops in the weak
landslide formation and
inter-layer.

Loessmudstone The main sliding surface

cutting layer cuts the underlying
landslide stratum, controlled by
joints, fracture and
other structural plane.

time it consider the relationship between development Heifangtai in Gansu Province is above 300 × 104 m3 ,
location of the sliding surface and stratum and con- but their sliding distance is only about 50 m. The vol-
ducive to understand the law of loess landslide and its ume of Jiaojia landslide, the Tanhuaguichang factory
prevention. landslide is generally only 10 × 104 –40 × 104 m3 , but
the sliding distance is more than 300–500 m. Such
phenomena cannot be explained and predicted with
Scheidegger’s theory or Moriwaki • width formula.
3 ACTIVITY CHARACTERISTICS
There are nearly relationship among sliding veloc-
ity, sliding distance and type of loess landslide by the
The sliding velocity, sliding distance and other char-
massive loess landslide Statistics survey.
acteristics of loess landslide are very important signif-
icance in the landslide disaster prediction, prevention 1. The homogeneous loess Landslide generally devel-
and control. Nowadays the main common forecast ops in the steep parts of loess slope, the main sliding
methods of landslide are: the slope theory was offered surface develops in the homogeneous loess, slid-
by Austria scholar A.E. Scheidegger[5] , the formula ing plane is approximate arc, controlled by vertical
was offered by Japanese scholar Moriwaki • width[6] , joints with the steep scarp (generally around 60◦ ),
the corresponding formula and the method of the slid- and the sliding plane is smooth and straight, be
ing velocity and sliding distance are also given by conducive to release the sliding energy rapidly, and
Jia-zheng Pan Academician, Si-Jing Wang Academi- become a high-speed slide. The sliding distance
cian, Yu-shu Fang and other scholars in China[7][8][9] . will be farther while there are open terrain con-
The statistical models are restricted by many influence ditions (Figure 1 and 2). Others such as Tianshui
factors. So it is difficult to predict the sliding veloc- Forging Machine landslide, has 80 m high, with
ity, sliding distance accurately, sometimes results are sliding distance is nearly 200 m and average sliding
opposite to the actual situation. For example, in recent velocity is 10 m/s around.
years, the volume of Huang-Ci landslide, Jiayouzhan 2. Loess interface landslide always develops in loess
landslide and Bengzhan station landslide occurred at slopes of Loess Hills Area with the dip angle from

520
 Figure 3. The section of Luoyugou valley loess interface
landslide in Tianshui city, Gansu.
Figure 1. The section of Tanhuaguichang homogeneous
loess landslide at Heifangtai, Gansu.

Figure 4. The section of Renjiawan loess interface landslide

in Lantian county, Shananxi.

Figure 2. The section of Qiewa homogeneous loess land- 3. Loess-mudstone layer plane landslide occurs in the
slide at Chongxin County, Gansu. slope site where the underlying stratum dip angle
always from 10◦ to 30◦ according with the direction
10◦ to 20◦ , many examples found in Tianshui, of the slope. The main sliding surface of this kind of
Tongchuan, Lantian cites. Malan loess deposits on landslide develops along the relatively soft forma-
the ancient topography which is up and down with tions or interlayer in underlying stratum. Because
draping form, and the underlying stratum mainly of the sliding of the bottom sliding mass, the upper
are formed in Cenozoic. The upper loess has good loess body forms tension cracks. Although this type
permeability, the lower mudstone and sandstone landslide have large sliding potential energy, the
have poor permeability. So they formed a double slope structure, the shape and nature of the main
heterojunction structure of the slope. The ground- sliding plane determine that its sliding velocity is
water got together at interface, so the interface area low and sliding distance is short. Firstly, the dip
was wet soft plastic state in a long-term and become angle of the main sliding plane is gentle, and its
the weak-soft structure surface of landslide. There- length is large relatively, it is not conducive to slid-
fore, the loess upper in the slope can slide easily ing. Secondly, the strength of the rock-soil is low,
along the weak-soft structure plane in the above the cohesion has basically released during the long-
area where there are abundant precipitations, form- term shear deformation, sliding body maintains a
ing loess interface landslide (Figure 3 and 4), often balance while main stabilizing force is friction of
occurs during the rainy season. the main sliding plane. Therefore, the cohesion
Because of high water content, low mechanical reduced gradually and caused the damage of the
strength near interface, the sliding velocity of loess rock-soil. On the other hand it can not create a
interface landslide is low, sliding distance is short. tremendous push force to make the sliding body
Jiaoping highway toll stations landslide caused by slide in high-speed and long distance.
slope cut, the main sliding surface is less than 5◦ . Huangci landslide, gas station landslide and
Since 1993 sliding always accompany with rainy pumping stations landslide at Heifangtai region
season, massive sliding soil buried and blocked the in Gansu, Wenchangge landslide, Qingbaishi land-
highways, impact the transportation seriously. It is slide , Wushan Dingjiamen landslide at Dajiatai in
low-speed sliding landslide[10] . Lanzhou, all these are this kind of loess landslide

521
 with the features of low sliding velocity and short sliding body, the slope will be stability by the
sliding distance. For example, the main sliding pro- effect of the part of a lock-solid role temporarily.
cess of Huangci landslide spent about 90 minutes, With the time passes, the scale and failure rate
the sliding distance is only 30–60 m although the of the landslide are increasing and accelerating
terrain is open in the front of the landslide (Figure 5 continuously, the stress of mudstone on the lock-
and 6). solid phase is concentrated consistently, length is
4. Loess-mudstone cutting layer landslide develops in shorter, the entire slope loses its stability and slide
the slope areas where the underlying stratum incli- completely when reach critical length and failure
nation is opposite to the direction of the slope, with instantaneously.
steeper slope and greater sliding potential energy.
This type landslide Developed through progres- According to experimental data, mudstone in lock-
sive deformation and damage by the effects of solid phase has high cohesion, its peak intensity
gravity and groundwater. Firstly, in the process, effect to present brittle failure. Therefore, the slid-
because of unloading and rebound, groundwater ing body will lose stabilizing force when the failure
acting, stress relative concentration at the toe of of the lock-solid phase happen, meanwhile get great
the landslide, the fracture develops shearing and corresponding pushing force. Add up the effects of
creeping is relatively strong, shear-creeping phase vaporization and a speed cushion when slide in high-
formed. It results the changing of stress and defor- speed, form high-speed and long-distance landslide
mation of upper loess slope, formation of the tensile when the terrain is open.
stress near the top of the landslide, emergence of Chana landslide in Longyangxia Reservoir and
tension cracks at the back of landslide, forma- Shaleshan landslide in Gansu are such kind of land-
tion of tensile paragraph at top. Shear-creeping slides. Part of the sliding body of Chana landslide
phase and rupturing phase develop, the stress con- surpasses the Yellow River and arrives at the other side
centrates in the mudstone of the central part of of the bank, sliding distance is more than 2000 m, slid-
ing velocity is 41 m/s based on experience counting[11] .
Shaleshan landslide happened on 7 March of 1983.
The height between anterior border and posterior bor-
der is 321 m (Figure 7 and 8), the furthest sliding

Figure 5. The loess-mudstone layer plane landslide at Taohe Figure 7. Saleshan loess-mudstone cutting layer landslide,
River, Gansu. Gansu Province.

Figure 6. The section of loess-mudstone layer landslide at Figure 8. The section of Saleshan loess-mudstone cutting
Taohe River, Gansu. layer landslide, Gansu Province.

522
distance is 1050 m in less than one minute, the aver- large-scale revival of landslide occurred in August
age speed is 20 m/s and the quickest can be 32 m/s. of 1994 after five years later, damaged the highway.
It killed 220 people and overwhelmed four villages. Similarly, Huangci landslide occurred in January
Such landslide often cause serious economic losses of 1995, the revival of large-scale landslide had
and people death. happened after 11 years later in May of 2006,
buried10 yards of residenter.
4. To loess-mudstone cutting layer landslide, Slid-
4 THE REVIVAL CHARACTER ing body was accumulated in the flat valley area
after the sliding in high-speed and long-distance,
According to the investigation data, the stability of sliding potential energy was released completely
sliding body and the scarp of different kinds of the basically, so the sliding body has high stability, slid-
landslide are different and the characters of sliding for ing body will be stable even anterior border of the
the second time or revival are also different. landslide is eroded by water, artificial excavation.
That was called ultra-stable by Professor Guang-
1. To homogeneous loess landslide, the potential tao Hu[10] . Sliding body of Shaleshan landslide,
energy has released after sliding in high-speed and Chana landslide and Wolong Temple landslide all
long-distance. Its stability is high. And it is always were in ultra-stable. The stable coefficient of Shale-
not overall revival or occur local collapse in the shan landslide which happened in 1983 was 5.5
scarp sometimes. For example, the stable coeffi- using Bishop Law to count. The scarp is high and
cient of Tianshui Forging Machine landslide is 2.3. steep when the landslide slides, develops nearly
Even when the factory was rebuilt the sliding body parallel tensile-tension crack by the effect of slid-
was cleared 60 m in the toe, the sliding body is still ing strongly at the scarp. So the stability of the
in stable state. But because the scarp of the land- scarp is poor and may slide once again. There was
slide is 50 m high and 45◦ in dip angle, its stability is big scale sliding at the scarp of Wolong Temple
worse and collapse often. Since 1985, the scarp of landslide in 1970[13] . There was small scale sliding
Panjizhai landslide has slided 8 times and the slid- of the scarp of Shaleshan landslide, and the vol-
ing scale become bigger and bigger. The landslide ume is 300 × 104 m3 when the large scale sliding
harms to State Road 310. happened on 26 March of 1986.
2. To loess interface landslide, the sliding potential
energy can not release completely because of low
sliding velocity and short sliding distance of, slid- 5 CONCLUSION
ing body will slide once more when faces with rain,
heavy rain, water erosion, or excavation. Some 1. The loess landslide can be divided into four basic
landslide is in perennial creep slowly. And its slid- types according to rock-soil formation and location
ing velocity will be quicker in rainy season or after of the sliding surface in loess areas: homogeneous
heavy rain. loess landslide, loess interface landslide, loess-
There are 3 times of large scale sliding at mudstone layer plane landslide and loess-mudstone
Zhangjiabanpo ancient loess interface landslide cutting layer landslide.
in Tianshui over the past 50 years. It destroyed 2. Sliding velocity is low and sliding distance is short
5 houses from 1973 to 1974. And a number of to the homogeneous loess landslide, sliding body
cracks occurred after heavy rain in August of 1990, is in a more stable state, may be revival because of
one of the fissures was 7 cm width and 60 m long, the poor stability of the scarp.
stagger down 40–50 cm. It was sliding again in 3. Sliding velocity is low and sliding distance is short
2001. The Jiaoshuwan ancient landslide in Tianshui to the loess interface landslide, the sliding body is
relived during the Autumn in 1990, was seriously in unstable state, may revive to slide by the effect
harmfal to the city and had to invest heavy fund to of motivating factors.
control[12] . The revival of landslides at Lantian in 4. Generally, Sliding velocity is low and sliding dis-
Shaanxi belong to such kind. tance is short to loess-mudstone layer plane land-
3. To loess-mudstone layer plane landslide, its slid- slide, sliding potential energy has not been released
ing distance is short, most of the sliding body completely, may revive to slide by the effect of
remains at main sliding surface, sliding potential motivating factors.
energy was not released completely, the sliding 5. The loess-mudstone cutting layer landslide has the
body may be in whole or partial revival because of characters of high-speed and long-distance, the
the rainfall, irrigation water or artificial excavation. stability of the sliding body is usually high after
Sliding distance of gas station landslide at Heifang- the sliding potential energy was released basically,
tai in Gansu in March of 1989 was only 35–45 m, but high-steep scarp is instability, may slide once
sliding velocity was also low, stability was poor, again.

523
REFERENCES disaster prediction. Landslide Paper anthology
in 1987 the National Colloquium Landslide.
[1] Xu, Bang-dong. 2001. Analysis and Prevention Chengdu: Sichuan Science and Technology
of landslide. Beijing: China Railway Press. Press, 117–124.
[2] Wu Wei-Jiang & Wang Nian-Qin. 2006. Land- [9] Fang Yu-shu. 1993. Research of the large high-
slide disasters in Gansu. Lanzhou: Lanzhou speed landslide forecasting. The natural slope
University Press. stability analysis and Huayingshan slope defor-
[3] Wu Wei-Jiang & Wang Shou-ying. 1989. Land- mation seminar collection. Beijing: Geological
slide mechanisms of Shaleshan. Landslide Paper Press. 92–102.
anthology in 1987 the National Landslide Collo- [10] Hu Guang—tao, et al. Landslide dynam-
quium. Chengdu: Sichuan Science and Technol- ics. 1995. Beijing: Geological Press. 36–56,
ogy Press. 125–139.
[4] Wu Wei-Jiang, et al. Characteristics and causes [11] Wang Cheng-hua. 1989. Longyangxia Reservoir
of the Huangci landslide[J]. Journal of Gansu Dam hydropower project near large landslide
Science, 1996 Supplement 73–78. prediction. Landslide Paper anthology In 1987
[5] Guo Chong-yuan. 1982. Super landslide and the the National Colloquium Landslide. Chengdu:
counting of the velocity. Landslide Collection, 190–197 Sichuan Science and Technology Press.
Part3. Beijing: China Railway Press 87–193. [12] Zhang Shi-wu(c) Wang Nian-Qin, et al.
[6] Moriwaki • Width. 1989. Landscape Forecast- 1996. The basic characteristics and control of
ing of sliding distance (Nian-Qin Wang transla- Jiaoshuwan landslide in Tianshui. Journal of
tion). Geological and railway roadbed. The third Gansu Science. Supplement 55–60.
period. 42–47. [13] Baoji in Shaanxi Province Irrigation Author-
[7] Pan Jia-zheng. 1980. Resistance sliding stabil- ity. 1976. The stability and control measures of
ity of buildings and landslides analysis. Beijing: Wolong Temple landslide. Landslide Collection.
Water conservancy Press, 120–132. Beijing: The Railwa Press, 147–152.
[8] Wang Si - Jing & Wang Xiao-ning. 1989. Anal-
ysis energy of Large-High-speed landslide and

524
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Investigation of a landslide using borehole shear test and ring shear test

Hong Yang
MWH Americas, Inc., Walnut Creek, California, USA

Vernon R. Schaefer & David J. White

Iowa State University, Ames, Iowa, USA

ABSTRACT: This paper presents a case history using the Borehole Shear Test (BST) as the in-situ test method
to investigate a landslide involving glacial till in Iowa, USA. Shear strength values of the soil were measured on
the walls of boreholes using an in-situ BST. The investigation was supplemented by measuring the residual shear
strength values of the soil on remolded soil samples in the laboratory using the Ring Shear Test (RST). Back
analyses were also performed on the slope to evaluate the possible shear strength values and slope conditions
at failure. The results show that the average shear strength value of the soil obtained by the BST is comparable
with the residual strength value obtained by the RST and the shear strength value obtained by back-calculation.
The slope failure could have occurred at a water table higher than that measured during the slope investigation,
as the failed slope is stable under the currently slope geometry and water table conditions.

1 INTRODUCTION the shear strength values and the slope stability analy-
sis results, the possible failure conditions and current
An important aspect of landslide investigation is to stability of the slope are evaluated.
determine and evaluate shear strength of the soils
in the field or in the laboratory. One of the many 2 SITE CONDITIONS
in-situ techniques is the BST which can directly mea-
sure soil shear strength parameter values (i.e., the The landslide is located next to highway E57 near
internal friction angle, φ  , and the cohesion, c ). The Luther, Iowa, USA. The regional geology indicates
BST is essentially a direct shear test that is performed that the soils and landscapes near the project site
on the wall inside a borehole. Normal stress is applied formed in glacial till deposited by the most recent, the
to the wall of the borehole through a pair of shear Wisconsin glaciations; and the soil deposits are gener-
plates, and the peak shear stress is measured in-situ ally up to one hundred meters in thickness. No bedrock
separately and concurrently. Thus, the φ  and c values outcrops or other soil types were observed during the
of the soils are determined from the Mohr-Coulomb field investigation. Examination of the aerial photos
failure envelope (e.g. Handy & Fox 1967). The BST indicates that the slide occurred sometime between
is normally considered a consolidated-drained test 1994 and 2002. The head scarp and the humps gen-
(Demartinecourt and Bauer 1983). erally appeared old in 2003 when the slide was first
In the laboratory, the RST has been used to deter- investigated; and bushes and vegetations were well
mine the residual shear strength of the soils. The RST grown on the slope surface, as indicated in Figure 1.
involves continuously shearing a remolded soil sam- The surface soil near the scarp appeared relatively
ple to a large displacement using a torsional ring shear loose.
apparatus such as the Bromhead ring shear apparatus The slope has an overall sloping angle of about
(e.g. Bromhead 1979; ASTM 2002). Thus, the shear- 16 degrees (H : V = 3.5 : 1), a maximum length of
ing resistance of the soils at very large displacement 85 m and a maximum height of 23 m. The width of
represents the soil condition in a landslide with a large the slope is about 80 m along the highway. The scarp
movement. of the slope near the top has a maximum height of 5 m.
In this study, the BST and RST were used to eval- The slope also has a few cracks near the middle and
uate the shear strength of the soils in a landslide in the toe of the slope (Figure 1). The maximum widths
Iowa. The shear strength values were also compared of the cracks are about 0.3 m. There is also a hump
with those obtained from back-calculations. Based on near the toe of the slope. A 2-m wide shallow ditch

525
Figure 1. Slope photograph showing the head scarp and the cracks on the slope.

was located at the toe of slope (beside the highway).

The surface features suggested that a relatively large
movement had occurred resulting the landslide.
Six boreholes were drilled manually on the slope
during a field investigation in 2004. The boreholes
were located either near the head scarp, at middle of the
slope or near the toe of the slope. The maximum depth
of the boreholes was 4.1 m. The boreholes showed
that the slope was made of yellowish brown glacial
till which was generally soft to medium stiff. Ground
water table was observed to be located near the bottom
of the boreholes after boring. The ditch at the toe of
the slope was wet indicating the water table was shal-
low near the toe. The water table near the crest of the Figure 2. Shear stress versus normal stress for the borehole
shear tests.
slope appeared relatively high, which could have been
contributed by the agricultural activities in the large
area of agricultural land behind the crest of the slope.

3 RESULTS OF FIELD AND LABORATORY

INVESTIGATIONS

The BSTs were conducted in the boreholes to obtain

the shear strength of the soils. A total of nine BST
were performed near or below the ground water table.
Shear stresses versus normal stresses for the tests are
presented in Figure 2, and the test results are summa-
rized in Table 1. The results show that the φ  value

526
 200
Normal stress = 395 kPa
180

160
297
140
Shear Stress (kPa)

120
199 B13-1 199
100
199
80

60 101 B11-1 101

101
40 52.0
52.0
27.5 27.5 52.0
B12-3

0
150 200 250 300 350 400 450
Displacement (Distance Traveled) (mm)

Figure 3. Shear stress versus displacement for the ring shear tests.

Table 2. Summary of the index test results of the soil 200

B11-1 (3.7m)
samples. B12-3 (3.4m)
150 B13-1 (3.1m)
Shear stress (kPa)

Atterberg
Grain Size Limit 100

Sample Depth Sand Slit/Clay LL PI 50

No. (m) (%) (%) (%) (%) USCS

0
0 50 100 150 200 250 300 350 400
BH11-1 3.7 48 52 28 14 CL Normal stress (kPa)
BH12-3 3.4 32 68 31 17 CL
BH13-1 3.1 47 53 27 14 CL
Figure 4. Shear stress versus normal stress for the ring shear
tests.

ranged from 12◦ to 36◦ , and the c value varied from 3.8 kPa. The average residual shear strength value
1 to 12 kPa. The average shear strength value was could be represented by φr value of 25.3◦ and cr value
represented by a φ  value of 21.3◦ and a c value of of 1.3 kPa.
7.8 kPa. Basic properties for representative soil sam-
ples were also investigated and the results are summa-
rized in Table 2. The results indicate that the glacial 4 SLOPE STABILITY ANALYSES
till could be classified as sandy, silty clay and had low AND RESULTS
plasticity.
The RSTs were conducted on three remolded soil Slope stability analyses were performed assuming that
samples and the results are presented in Figures 3 the soil condition was uniform for the slope. This
and 4. The shear stresses versus displacements at was because the ground conditions appeared relatively
various test stages (various normal stresses) for the simple based on the site geology and field investiga-
samples were plotted to determine the corresponding tion, and the slope was of relatively small size. The
ultimate shear stresses (Figure 3). Then, the ultimate slip surface was assumed circular passing the observed
shear stresses were plotted against the corresponding scarp at the top of the slope and a point near the toe
normal stresses, and the residual shear strength val- of the slope based on the estimated pre-failure slope
ues of the soil samples were obtained (Figure 4). The surface (the area of ‘‘loss’’ of the pre-failure slope near
RST results are also summarized in Table 3, which the top equals the area of ‘‘gain’’ of current slope near
indicates that the glacial till has residual shear strength the toe, as shown in Figure 5). Thus, the centers of the
values of residual friction angle, φr ranging from 24.3◦ rotations are all located on the line that is perpendicu-
to 26.0◦ and residual cohesion, cr ranging from 0 to lar to and bisects the slope surface. The most probable

527
 values from the BST and the lowest shear strength
values from the RST (to be conservative) were used as
inputs of soil properties for the analyses.
The results of the slope stability analyses and the
corresponding shear strength values are summarized
in Table 4. A few observations can be made from the
results: 1) FS values obtained from the Morgenstern-
Price method and the Bishop simplified method are
essentially the same for each of the analyses. 2) FS
based on the pre-failure slope surface are generally
smaller than the FS based on the current slope sur-
face (i.e., FS of Analyses 1 and 2 are smaller than FS
of Analyses 4 and 5, respectively); or shear strength
values resulting unity FS based on pre-failure slope
surface is higher than those based on current slope sur-
face (i.e., back-calculated shear strength in Analysis 3
is higher than that in Analysis 6). These results are con-
sistent with the fact that the current slope is more stable
than the pre-failure slope, as driving force was reduced
by the soil ‘‘loss’’ and resistant force was increased by
the soil ‘‘gain’’ (Figure 5). 3) The FS values based
on the average BST and lowest RST shear strength
values are very close indicating that the average BST
and RST shear strength values are comparable, and
they are also both close to the back-calculated shear
strength values based on the pre-failure slope sur-
face. 4) The RST shear strength is slightly higher
Figure 5. Slope profile, possible slip surfaces and results than that based on back-calculation suggesting that
of slope stability analyses. the slide may have occurred under less favorable con-
ditions (such as water table higher than the current
one), so that the failure occurred at a higher shear
Table 4. Summary of shear strength values and factors of strength (i.e. the RST shear strength), since the RST
safety. shear strength represent the ultimate shear resistance
of the soil. 5) The FS values are about 1.4 based on
No. Source φ c M-P Bishop Surface
current water table and slope surface conditions indi-
1 BST (average) 21.3 7.8 1.072 1.069 Pre-failure cating that the slope is currently stable if the water
2 RST (low) 24.3 0 1.034 1.029 Pre-failure table is not changed drastically. In fact, no apparent
3 Back-calculated 23.6 0 1.001 0.996 Pre-failure slope movement has been observed since 2004 when
4 BST (average) 21.3 7.8 1.414 1.411 Current the slope was investigated.
5 RST (low) 24.3 0 1.393 1.389 Current
6 Back-calculated 18.0 0 1.002 1.000 Current

Note: M-P = Morgenstern-Price method; Bishop = Bishop

5 CONCLUSIONS
simplified method; Current water table during investigation
was used.
The BST and RST were employed for a landslide
investigation, and the shear strength values obtained
were used in the slope stability analyses. Shear
depth of the slip surface was determined by the critical strength values of the soils were also estimated from
slip surface. back-calculations based on the slope giving unity
The observed ground water table condition was factor of safety. The results show that the RST
used in the slope analyses. Both current and pre-failure shear strength value essentially represented the shear
slope geometry were considered. Back-calculations strength of the soils at the slope failure, which was
were also performed to determine the average shear also comparable with the average BST shear strength
strength values of the soil giving unity factor of safety value and the back-calculated shear strength values.
(FS). Two methods, the Morgenstern-Price method The slide most likely has occurred under a higher water
and the Bishop simplified method, were used for the table. However, the slope is stable under current water
slope stability calculations. The average shear strength table and ground conditions.

528
ACKNOWLEDGEMENTS Bromhead, E.N. 1979. A simple ring shear apparatus. Ground
Engineering, 12 (5), 40–44.
The Iowa Highway Research Board sponsored this Demartincourt, J.P. Bauer, G.E. 1983. The modified bore-
study under contract TR-489. The findings and opin- hole shear device. Geotechnical Testing Journal, ASTM,
ions expressed in this paper are those of the authors 6, 24–29.
Handy, R.L. Fox, N.S. 1967. A soil borehole direct shear
and do not necessarily reflect the views of the sponsors test device. Highway Research News, Transportation
and administrations. Research Record, 27, 42–51.

REFERENCES

ASTM 2002. D6467. Standard test method for torsional ring

shear test to determine drained residual shear strength of
cohesive soils. American Society of Testing and Materials.

529
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

The importance of geological and geotechnical investigations of landslides

occurred at dam reservoirs: Case studies from the Havuzlu and Demirkent
Landslides (Artvin Dam, Turkey)

A.B. Yener
General Directorate of State Hydraulic Works (DSI), Ankara, Turkey

S. Durmaz
General Directorate of Mineral Research and Exploration (MTA), Ankara, Turkey

B.M. Demir
General Directorate of Disaster Affairs (AIGM), Ankara, Turkey

ABSTRACT: Landslides, occurring in dam reservoirs, result in a large number of causalities and huge eco-
nomic losses all over the world. Therefore, regional and moderate scaled landslide investigations around a dam
and in its reservoir provides major contribution to the decrease of losses caused by landslide disasters. In this
study, moderate-scale susceptibility, hazard and risk assessments of landslides in the reservoir area of the Artvin
Dam are introduced with present limitations and review of the precautionary measures. There are two huge
landslides in the reservoir area of the Artvin Dam, called Havuzlu and Demirkent landslides. Based on the field
investigations, laboratory experiments and evaluations on the previous and most recent data, a susceptibility map
of the area was constructed. This map indicated that 88.9% of the existing landslides, including Havuzlu and
Demirkent landslides, have high and very high susceptibilities and it is obvious that a risk should be expected
in the investigated area. This finding suggests that the crest of the Artvin Dam should be reassessed and raised
to compensate possible wave effects. While it is predicted that the reservoir is blocked after a new and huge
landslide, and the lake water overruns over the dam. Besides, driving by-pass tunnels at 565 m asl with a length
of about 2000 m and a diameter of 10 m could be helpful to derive the rising water in the reservoir. On the
other hand, the Yusufeli Dam Power Building and structures located in the upstream area could be damaged
by the flood, and therefore the foundation levels of the buildings and structures should be reassessed and the
construction project should also be revised. Then, possible choices should be assessed by the control office and
the contractor in terms of feasibility and ease of application.

1 INTRODUCTION been readily foreseen if better geological consulting

had been carried out before the construction of the
1.1 Aim of the study dam and reservoir.
This study, discussion of the medium scaled land-
Landslides, that occur in dam reservoirs, have caused
slides susceptibility, hazard and risk assessments
large number of casualties and huge economic losses
which take place in the Artvin Dam Reservoir, intro-
in all over the world. Therefore, regional and mid-
ducing present limitations and review of the precau-
dle scaled landslide investigations around the dam
tionary measures were aimed.
and in its reservoir provides major contribution to the
decrease of losses caused by landslide disasters. For
instance, in 1963, a major rock slide resulted in a death
1.2 Description of the study site
toll of approximately 2600 in Italy. The slide block
moved suddenly into the newly filled reservoir of the The Artvin Dam and its hydroelectrical power plant
Vaiont Dam (Italy), flushing the lake water up and (HEPP) are planned to be constructed on Coruh River,
over the dam. The wall of water was over 200 m high NE Anatolia, Turkey. It is designed as concrete arch
as it swept into nearby villages, wiping out everything dam with a height of 180 m and 332 MW power plant
in its path. The rock slide and the flood could have capacity.

531
 Figure 2. Stratigraphic columnar section of the investigated
Figure 1. Location map of the investigated area. area (Not-to-scale, simplified from Ertunc, 1980).

Different scaled landslides have been identified at

the Artvin Dam reservoir site (Ertunc., 1980, 1991,
Gunay., 1991). In this study, two large-scaled land-
slides, called Havuzlu and Demirkent landslides, were
investigated and evaluated in the reservoir area (Fig. 1).

1.3 The methods employed

Detailed geological surveys were carried out by using
1/25 000 scaled topographic maps and 1/5 000 scaled
geological maps. The landslide characteristics were
also evaluated by measurements and observations.
Laboratory experiments were conducted on the sam-
ples taken from the study area, and a landslide database
was obtained. Finally, the lithological, morpholo-
gical, gradient and aspect maps of the study site were
produced using Arcmap, SPSS and Arcview com-
puter codes and landslide susceptibility maps were
constructed.

2 GEOLOGY

Liassic Yusufeli formation and the Ýkizdere Mag-

matites of Upper Eocene—Oligocene crop out in the
study area (Baydar et al. 1969, Ertunc. 1980, 1991,
Figure 3. Geological map of the investigated area ( adapted
Gunay. 1991). In addition, Quaternary terrace, allu-
from Ertunc 1980, Gunay, 1991).
vium and slope wash deposits with landslide material
widely cover the investigated area (Figure 2).
Liassic Yusufeli formation covering large areas in
the middle of the reservoir consists of basic igneous the main body of a regional batolite which partly
rocks (gabbro, amphibolite), spillite, metabasalt, penetrated into the Yusufeli formation. The unit is
agglomerate, slate and lithic tuff. The formation is composed of granodiorite-tonalite, adamellite, por-
black, dark green and dark gray in color. Due to frac- phyritic microgranite and granite-gneiss. The Ikizdere
turing, a blocky rock mass is observed in the study magmatites are identified with the presence of gran-
area. The strength of the rocks increases with depth. ite outcrops in the study area. Light pink colored
The Ikizdere magmatites, observed at the middle and coarse grained granitic rocks generally have high
part of the reservoir and towards downstream, form strength (Figure 3).

532
 Artvin Dam Site

Yusufeli Dam Site

Figure 4. Landslide susceptibility map of the investigated area.

3 LANDSLIDE INVESTIGATIONS IN THE 3.1 The Havuzlu landslide

ARTVIN DAM RESERVOIR
The Havuzlu landslide is located on the NW of the
To assess landslide susceptibility, hazard and risk Coruh River, approximately 8 km to the upstream.
in the studied area, firstly, landslide locations were It takes place below the Saburta Mountain between
determined through field studies and aerial photo the two parallel hills trending NW-SE. Based on the
interpretations. Seven different scaled landslides were investigations, the bedrocks in the region have been
determined in the reservoir site of the Artvin Dam, identified as phyllite, tuff and gabbro of the Yusufeli
with an area of 36.8 km2 . Areas of landslides vary formation (Baydar et al. 1969, Ertunc.1980, 1991,
between 0.3 km2 and 1.2 km2 . Due to the scales of Gunay. 1991). It is observed that the landslide devel-
the landslides, the Havuzlu and Demirkent landslides oped in the Yusufeli Formation within the loose hill
were investigated in detail. material. Its average gradinet is 22◦ Some flat areas

533
observed suggest a secondary movement. Head of the Landslides, have high and very high susceptibilities
landslide could not be observed and a definite bound- (Fig 4).
ary of the sliding surface could not also be identified.
The thickness of the landslide material varies between
100–150 meters. The laboratory experiments on the 4 CONCLUSIONS
samples taken from the study area indicated that the
landslide materials are generally clayey gravel (GC). Due to the determined geological properties of the
Volume of the landslide is predicted as 86×106 m3 , units that formed the slopes of the landslide areas
however, the calculated volume of the reservoir below and sliding materials, the models and geometries
the Havuzlu landslide is 4 × 106 m3 . The data sup- of landslides, and susceptibility maps, reasons and
plied from field and laboratory studies, susceptibility possible results of landslides are explained. Also,
map and partial movements on the landslide occurred comments and evaluations on the instability of the
in 1988 indicated that the Havuzlu landslide is still area and the possible movements on the investigated
active. For this reason, the reservoir slopes would landslides are presented. In conclusion, with regard
easily be blocked in case of any partial movement on to the field investigations, laboratory studies and
the landslide towards the reservoir. Besides, even a evaluations of existing and new data, susceptibil-
partial sliding could create a wave at 25–30 m heights ity map of the area indicated that 88.9% of existing
where the material slip into the reservoir and 12 m landslides, including Havuzlu and Demirkent land-
heights around the Artvin Dam is calculated. slides, have high and very high susceptibilities and
that a risk of new landslides should be expected in
the investigated area. Meanwhile monitoring of the
3.2 The Demirkent landslide landslides should be carried out, continuously.
The Demirkent landslide, which is located on the It is also predicted that the reservoir will be blocked
SE of the Coruh River approximately 6.5 km to the after a new and huge landslide, causing the overflow
upstream, developed between two parallel hills trend- of lake water over the dam. Therefore, the crest of
ing NE-SW. The volume of the landslide is predicted the Artvin Dam should be reassessed and raised to
to be 57 × 106 m3 . Maximum thickness is about 120 compensate possible wave effects. Besides, driving
meters. Although the landslide has an average gradient by-pass tunnels at 565 m asl with about 2000 m in
of 23◦ , it is topographically flat when compared to its length and 10 m in diameter, could be helpful to derive
surroundings. Diabase and gabbro of the Yusufeli for- the rising water in the reservoir. On the other hand,
mation are observed in the area. The landslide material as the Yusufeli Dam Power Building and structures
is generally composed of fine grained materials such located in the upstream could be damaged because of
as green to gray clay, silt and gravel. No evidence of the flood, foundation levels of the buildings and struc-
instability was encountered during the field studies, tures should be reviewed and the project should be
and no mass movement has recently been reported in revised. Available choices should be determined by
the area. Due to these reasons, once the reservoir is the control and contractor authorities according to the
filled, secondary mass movements are not expected in allowance of the feasibility and the ease of application.
the Demirkent landslide.
REFERENCES
3.3 Landslide susceptibility map of the
investigated area Baydar, O., Erdogan, A., Topcan, A., Kengil, R., Korkmazer,
Landslide database including landslide features was B., Kaynar, A. & Selim, M. 1969. Geology of the region
between Yusufeli, Ogdem, Madenkoy, Tortum Lake and
established by observations and measurements dur- Ersis. Ankara: MTA (in Turkish).
ing the field studies, also the evaluations and data of Ertunc, A. 1980. Engineering geological investigation of pos-
preceding works were used for this purpose. sible dam sites, reservoirs and tunnel routes at the Coruh
Then, the slope unit map of the area was pro- Basin.. Ankara: EIE (in Turkish).
duced by using computer programs such as Arcmap, Ertunc, A. 1991. Effects of landslides on dam projects along
SPSS, etc. Slope, slope aspect, topographical eleva- the Coruh River, Turkey 1. landslide symposium pro-
tion, shape of slope, lithology, water conditions and ceedings, Trabzon, 27–29 November 1991. Trabzon:
vegetation cover were considered as independent input Karadeniz Technical University Press (in Turkish).
parameters in this study. Gunay, S. 1991. Engineering geology report of the Artvin
Dam and HEPP project. Artvin: DSI (in Turkish, unpub-
Finally, the susceptibility map of the reservoir lished).
site of the Artvin Dam and its surroundings was Yilmaz, S.B., Gulibrahimoglu, I., Yazici, E.N., Yaprak, S.,
produced using the slope unit map with logistic regres- Saraloglu, A., Konak, O., Kose, Z., Cuvalci, F. & Tosun,
sion method. Based on this map, 88.9% of the C.Y. 1998. Environmental geology and natural resources
existing landslides, included Havuzlu and Demirkent of Artvin. Ankara: MTA (in Turkish).

534
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

An innovative approach combining geological mapping and drilling process

monitoring for quantitative assessment of natural terrain hazards

Z.Q. Yue, J. Chen & W. Gao

Department of Civil Engineering, The University of Hong Kong, Hong Kong, China

ABSTRACT: This paper presents an innovative approach to characterize geotechnical conditions of soil and
rocks in the ground of hilly natural terrains. The approach combines the conventional geological mapping method
and the recent invented Drilling Process Monitoring (DPM) method. The approach can take advantages of the
two methods for accurate, economical, quick and flexible applications in hilly natural terrains. Consequently, the
approach can offer us quality factual data on the ground conditions for better quantitative assessment of natural
terrain hazards.

1 THE ISSUE Prior to any development at a hillside, a quantitative

assessment of the hilly natural terrain hazards has to
Figure 1 is a satellite image showing the hilly natural be conducted. A site investigation has to be carried
terrain above the Po Shan Road Landslide in 1972 in out accordingly (Chen, 2003; Chen et al., 2005; GEO,
Hong Kong Island. Figure 2 is a site photograph show- 1987, 1988, 1994). However, there is an issue.
ing a front view of the hillside in 2004 when slope The issue is how to safely, accurately, economically
maintenance works were being carried out. The 1972 and environmental-friendly characterize the mecha-
landslide destroyed a 4-story building, collapsed a nical conditions of soil and rocks in hilly natural
12-storey building and killed 67 persons (HKG, 1972). terrains.
The Po Shan Road is about 180 mPD (meter above The conventional methods for site investigation
the mean sea level). The failure scarp was up to include trial pits and trenches, engineering geological
300 mPD. The hillside peak is the Victoria Peak mapping and aerial photograph interpretation. Modern
of 552 mPD. The hillside has variable slope angles digital technologies including geographical informa-
with an average of 33◦ . Since 1972, the hillside sur- tion system (GIS), digital photogrammetry, mobile
rounding the hillside has not been developed due to mapping systems combining GIS and global posi-
great concerns on the natural terrain hazards. tioning system (GPS) can also be used. A natural
terrain landslide inventory for the site in GIS can
also be established using historical records and aerial
photographs.
These methods, however, can reveal only the
ground conditions at or near the hillside surface. They

Figure 1. Steep natural hillside above Po Shan Road in Hong Figure 2. A front view of the steep hillside surrounding the
Kong Island (modified after http://www.google.com). landslide.

535
cannot reveal effectively the ground properties, in par- always specified in ground investigation for natural
ticular, the defined mechanical properties at depth. terrain hazard assessment in Hong Kong. The access
Landslide investigations in Hong Kong and other is difficult. Heavy timber scaffolding is needed to form
regions have evidently shown that the ground proper- safe and stable access and drilling platforms. Some-
ties at depth play a dominant role in the occurrence of times, a helicopter may be mobilized to transport the
slope failures and the extrapolation of ground proper- drilling equipment. The drilling time is long. The nat-
ties from the surface observations can be inaccurate ural terrain can cover a considerably large area with
and unreliable (Chen et al. 2007; Lan et al, 2003; dense vegetation. Other difficult features are possible
Shang et al. 2003; Sheng et al. 2002; Yue et al. 2001; to be encountered in the field.
Yue & Lee, 2002; Zheng et al. 2006). Therefore, innovative approaches for quantitative
The hillside terrain at Po Shan was formed via a long characterization of the ground conditions and proper-
and natural process. It experienced numerous geolo- ties must be developed and used for environmentally
gical and geotectonic movements, physical and chemi- responsible investigations in natural terrains.
cal weathering and deterioration. It has colluvium, The purpose of this paper is therefore to address the
residual soils, weathered soils and rocks, weak zones, issue with an innovative approach. The approach is to
faults and joints. Figure 3 is a site photograph taken combine and integrate the conventional engineering
in 2006 showing the surface condition of the hillside geological mapping and the recent invented drilling
indicated in Figures 1 and 2. Abundant boulders of process monitoring (DPM). Details of the DPM can be
weathered volcanic rocks and heavy vegetation and found in Yue et al. (2001, 2002, 2003, 2004a, 2004b,
trees were present. 2004c, 2006, 2007) and Yue (2004, 2005), Sugawawa
In order to obtain the factual data about ground con- (2003) and Lam & Siu (2006). The combined approach
ditions below the hillside surface (from 0 to 50 m can take advantages of the two methods. As a result,
below the ground), conventional sub-surface explo- it can be applied to quantifying the ground condi-
ration methods may be adopted. The common method tions and particularly the spatial distribution of the
is to drill vertical or inclined boreholes in the ground geomaterial mechanical properties in the 0 to 50 m
using hydraulic rotary core drilling machines to obtain deep ground of hilly natural terrains. The applica-
disturbed or undisturbed soil samples and rock cores. tion can be accurate, economical, quick, flexible and
However, such drilling or coring can be used in a environmental-friendly.
very limited scope. The conventional ground investi- In ensuing, a brief summary of the proposed
gation approach in civil and geotechnical engineering approach is given. A case study is further used to illus-
cannot be used effectively and economically for the trate the proposed approach. Next, further discussions
natural terrain hazard studies. The reasons can be are made on how to use the approach as a solution of
summarized as follows. choice to address the issue.
The hydraulic rotary core drilling machines are usu-
ally heavy and need a drilling lubricant such as water.
The water can destabilize the hillside slope that may
2 THE INNOVATIVE APPROACH
be marginally stable, which is why air foam is nearly
2.1 Basic principle
The basic principle of the proposed approach is to use
the conventional engineering geological mapping to
accurately characterize and record the conditions and
properties on the ground surface of hillsides. Such
information may include exposed soils and rocks,
their physical and mechanical features, discontinu-
ities, seepage, as well as vegetations. The modern
digital technologies can further be used to measure and
record the information exposed on the ground surface
more accurately and extensively.
Based on the mapped information, the DPM is
applied to reveal the mechanical properties and
strength beneath the ground surface at some selected
spots in the hilly natural terrains. The method
uses an air-driven rotary-percussive drilling machine
Figure 3. A corner of the steep natural hillside above with down-the-hole hammer to drill a hole of 100 mm
Po Shan Road in Hong Kong Island with trees and boulders in diameter in the ground and in the meantime uses
in 2006. a drilling process monitor to record the entire drilling

536
process in real time. From the monitored data, the 2.4 Drilling process monitoring
mechanical strength properties and their spatial dis-
It is well understood that forming a hole in the ground
tribution along the hole can be obtained. The drilling
with an air-driven rotary percussive drilling machine
can be carried out quickly on a simple and light steel
equipped with a DTH hammer requires detaching geo-
scaffolding. The hole can have a depth up to 60 m and
material from the cut face with drill bit and removing
a declination or inclination angle variable with respect
the detached geomaterial from the hole with flushing
to the hillside slope.
air, which in fact is a mechanical failure process of the
ground geomaterial.
2.2 Geological mapping The DPM is associated with the drilling machine
in Figure 5. It automatically and digitally monitors the
The conventional geological mapping method has drilling process and quickly characterizes the mecha-
been developed and used for many years. Figure 4 nical and strength properties of the ground soil and
shows an example using a geological hammer to map rocks along the hole. It has a hardware for in-situ auto-
the mechanical properties of soils and rocks exposed matic monitoring and recording of drilling parameters
on the hillside and a digital photograph to record the in real time and a software package for analyzing and
measured spot. Other modern digital technologies can presenting the monitored digital data in time series and
also be used. Details can be found in the relevant liter- along the hole depth.
atures (e.g., GEO, 1987, 1988, 1994; Chen et al. 2005; The drilling machine in Figure 5 was equipped
Zheng 2006). with a DPM hardware. The hardware device was
non-destructively mounted onto the drilling machine.
2.3 Air-driven rotary-percussive drilling The whole drilling process was monitored in a digital
with down-the-hole hammer manner and in real time series.
Figure 6 presents an interface from the DPM soft-
As shown in Figure 5, an air-driven (pneumatic) ware showing a typical original DPM data in time
rotary-percussive drilling machine with a down-the- series associated with the drilling of a single hole.
hole (DTH) hammer was being used to drill a declined The graphs under the columns 1 to 7 represent the
hole of 100 mm in diameter into the hillside shown complete time-histories of the downward and upward
in Figure 4. The platform was a steel scaffolding thrust pressures, the forward and reverse rotation pres-
on the hillside. The flush medium is compressed air sures, the percussion pressure, the bit rotation speed,
flow, which has no adverse effect to the groundwater, and the chuck position, respectively. The 1 to 5 pres-
however, is very noisy and dusty. sures represents the five compressed air flows in the
The machine was operated by two skilled men. five pipes from the controller for pushing the rod
One man was sitting near the controller, operating downward, lifting the rod upward, rotating the rod
and powering the rig by steering the supplies of com- clock-wisely, rotating the rod anti-clock-wisely, and
pressed air flow from the main pipe into five dividing punching the DTH hammer and flushing debris at the
pipes. The other was standing near rig and adding or bit out of the hole, respectively, in real time.
disconnecting the 1 m long drill rods.

Figure 4. Mapping the properties of soils and rocks on Figure 5. Drilling a declined hole in a hillside with an air-
ground surface with geological hammer and digital pho- driven rotary-percussive drilling with down-the-hole ham-
tographs. mer, which was being monitored with a DPM device.

537
 Furthermore, the depth of hammer bit in the hole in The plot in Figure 8 clearly shows the spatial vari-
real time can be derived from the original DPM data ation of the mechanical strength of the weathered
in Figure 6. The upper graph in Figure 7 shows the volcanic rock along the drill hole.
complete time-history of bit depth in the hole.
The drilling of the 35 m deep hole started at
11:30:21 and completed at 13:31:20 and used a total 2.5 The combined approach
of about 2 hours. The advancement of bit into the new The geological mapping at the hillside found that hill-
ground ended at about 13:08:00. The retrieving of bit side surface was made of completely decomposed
from the hole started at about 13:14:00. volcanic (CDV) in dry condition, as shown on Figure 9.
Figure 8 plots the depth of bit with respect to the From Figure 8, it is evident that this CDV soil layer
net drilling time, where the associated percussion extended about 7.624 m into the ground. Then much
pressure, downward thrust pressure and the forward stronger weathered or better graded volcanics were
rotation pressure are not shown. encountered in the zones 2 and 3 for 12.32 m thick.
The curve of the bit depth versus the net drilling Subsequently, weaker zones were present in the zones
time can be divided into 10 linear zones along the hole 4 to 8. Stronger geomaterials were shown again in the
of 35.094 m in total length. Each zone has an aver- bottom two zones 9 and10.
age drilling rate that is listed in Figure 8. The lower Since the drilling and monitoring can be carried
depths of the zones 1 to 10 are 7.624, 15.861, 19.943, out quickly and flexibly, many holes can be drilled
22.421, 23.673, 24.184, 31.644, 32.776, 33.086, and and monitored in the hillside. The spatial distribution
35.094 m, respectively. Their thicknesses are 7.624, of the geomaterial strength along each hole can be
8.237, 4.082, 2.478, 1.252, 0.511, 7.460, 1.132, 0.310, revealed and inter-related.
and 2.008 m, respectively.

Zone 1

Zone 2

Zone 3
Zone 4

Zone 7

Figure 8. DPM identified mechanical strength zones of

constant drilling rates along the hole in the hillside in
Figure 6. Original DPM data showing the time history of Figures 6 & 7.
whole drilling process for forming a hole in the hillside.

100 mm diameter
hole by rotary-
percussive drilling

Steel scaffolding
on hillside slope
Bit advancing Retrieving

Figure 7. The complete time-history of the hammer bit Figure 9. Completely decomposed volcanic soil exposed at
depth in the hole and the chuck position outside the hole. the entrance of the declined hole in Figures 6 to 8.

538
 As a result, the two dimensional geological map
made on the ground surface can be extended and
extrapolated into the hillside for a depth upto the
drill hole lengths. Three-dimensional geological map
of the hillside can be drawn with high accuracy and
reliability.

3 CASE STUDIES

A trial use of the proposed approach was conducted

in Hong Kong in 2005. Figure 10 shows the hill-
side, where both the conventional hydraulic drilling
machine and the air-driven rotary-percussive drilling
machine were used. Details of the two drilling methods
can be observed in Figures 5, 11 and 12.
The hydraulic drilling machine was used to obtain
undisturbed soil samples and rock cores from the hill-
side. The drillhole was declined at an angle of 15 to the
horizontal. The time for setting up the timber access Figure 11. Hydraulic drilling on the hillside in Figure 10.
and platform was about 10 working days. The time for
the drilling of a 40 m depth inclination hole was about
five working days.
The steel platform for the rotary-compressive
drilling was established about one day and the drilling
equipment was mobilized onto the platform by hand.
No special access scaffolding was used. The drilling
with automatic monitoring for a similar 40 m depth
declined hole used about 3 hours.
Figures 4 to 9 and the associated text in Section 2
above are part of the case study.
Furthermore, the rotary-percussive drilling can be
carried out for rock cliff up to 50 m high. Figure 13
shows an example of such drilling work in Hong Kong.
The drilled holes were used for installation of rock
bolts and relief drainage pipes.

4 THE SOLUTION TO THE ISSUE Figure 12. DTH hammer and bit with two flushing holes on
above two sliding channels of the rig steel frame.
From the above case studies and discussions, it
is evident that conventional ground investigation
approaches have various limitations in obtaining clear

Pneumatic drilling
on simple steel
platform

Hydraulic drilling on
heavy timber platform

Figure 10. Hydraulic rotary core and pneumatic rotary- Figure 13. Air-driven rotary percussive drilling in a cage
percussive drilling on a steep hillside in Hong Kong. lifted by a crane to drill a hole in rock cliff in HK.

539
and reliable pictures and understanding of ground of the Hong Kong Special Administrative Region,
conditions in large hilly natural terrains. China (Project No. HKU 7137/03E) and a grant from
Hydraulic rotary core drilling techniques are usu- China Natural Science Foundation (No. 50729904).
ally time-consuming and costly although they can The authors also thank some in-kind supports from the
provide undisturbed soil and rock samples. Only a very Geotechnical Engineering Office of the Government
limited number of rotary boreholes can be carried out, of the Hong Kong Special Administration Region,
which results in limited factual data and could lead to Halcrow China Ltd, Maunsell Geotechnical Services
inaccurate and unreliable ground geotechnical models. Ltd., Fugro (Hong Kong) Ltd., and Soils and Materi-
The recently invented DPM method can acquire als Eng. Co., Ltd. The two coauthors J. Chen and W.
digital data of the drilling parameters for the real-time Gao also thank The University of Hong Kong for pro-
process of drilling a hole in the ground. The DPM data viding financial supports to them during their Ph.D.
are factual data and can identify the variation and dis- studies.
tribution of the geomaterial strengths associated with
the soils and rocks along the hole depth.
The air-driven rotary percussive drilling equipped
REFERENCES
with DPM are portable, flexible, reliable and eco-
nomic in-situ techniques. They can be easily and Chen, Z.Y. 2003. Soil Slope Stability Analysis—Theory,
non-destructively established on natural hillsides with Methods and Programs, China Water Sources and
a simple steel platform. The drilling does not cause Hydropower Publisher, Beijing, 560p.
side effects to the hillside stability and can be con- Chen, Z.Y., Wang, X.G., Yang, J., Jia, Z.X. & Wang, Y.J.
sidered non-destructive. The drilled hole can further 2005. Rock Slope Stability Analysis—Theory, Methods
used for soil nailing, rock bolting and drainage pipe and Programs, China Water Sources and Hydropower
installing so that the hillside can also be strengthened Publisher, Beijing, 890p.
after the ground investigation. Chen, N.Sh., Yue, Z.Q., Cui, P. & Li, Z.L. 2007. Landslide
Therefore, the proposed approach takes advan- induced debris flow and its maximum discharge esti-
mation—illustrated with case studies, Geomorphology.
tages of both the conventional engineering geological 84 (1): 44–58.
mapping method and the drilling process monitor- GEO. 1987. Guide to Site Investigation, Geotechnical Engi-
ing method in association with the air-driven rotary neering Office (GEO), Civil Engineering Department,
percussive drilling equipment. As a result, the pro- HK.
posed approach gives an idealized way for accurate, GEO. 1988. Guide to Rock and Soil Descriptions. Geotechni-
economical, quick and flexible applications in hilly cal Engineering Office (GEO), Civil Engineering Depart-
natural terrains. The approach can offer us rich and ment, HK.
reliable factual data on the ground conditions for better GEO. 1994. Geotechnical Manual for Slopes (Second
quantitative assessment of natural terrain hazards. Edition). Geotechnical Engineering Office (GEO), Civil
Engineering Department, HK.
Returning back to the issue proposed at the begin- GHK. 1972. Commission of Inquiry into the Rainstorm
ning about how to conduct the site investigation on Disasters, 1972. Final Report. Hong Kong. Govt. Printer.
the vegetated steep hillside above Po Shan Road, the Lam, J.S. & Siu, C.K. 2006. Evaluation of application of
proposed approach can be a solution of choice. The drilling process monitoring (DPM) techniques for soil
mapping and DPM can accurately and economically nailing works, GEO Report, No. 189, Geotechnical
and environmental-friendly characterize mechanical Engineering Office, HKSAR Government.
conditions of ground soil and rocks in the hilly natural Lan, H.X., Hu, R.L., Yue, Z.Q., Lee, C.F. & Wang, S.J.
terrain. 2003. Engineering and geological characteristics of gran-
Consequently, the quantitative hazard assessment of ite weathering profiles in South China, Journal of Asian
Earth Sciences, 21 (4): 353–364.
the natural terrain can be made using the rich and reli- Shang, Y.J., Yue, Z.Q., Yang, Z.T., Wang, Y.C. & Liu, D.A.
able factual DPM data in addition to the conventional 2003. Addressing severe slope failure hazards along
factual mapping data on the hillside surfaces. Miti- Sichuan-Tibet highway in southwestern China, Episodes,
gation measures such as check dam, flexible barrier Journal of International Geoscience, 26 (2): 94–104.
and preventive measures such as soil nailing, retaining Sheng, Q., Yue, Z.Q., Lee, C.F., Tham, L.G. & Zhou, H. 2002.
structures, drainage improvement, and boulder stabi- Estimating the excavation disturbed zone in the permanent
lization can be proposed and designed more effectively shiplock slopes of the Three Gorges Project, China, Inter-
and environmental-friendly. national Journal of Rock Mechanics and Mining Sciences.
39: 165–184.
Sugawawa, J., Yue, Z.Q., Tham, L.G., Lee, C.F. & Law, K.T.
2003. Weathered rock characterization using drilling
ACKNOWLEDGEMENTS parameters. Canadian Geotechnical Journal, 40 (3):
661–668.
The work described in this paper was substantially sup- Yue, Z.Q., Lee, C.F. & Pan, X.D. 2001. Landslides and
ported by a grant from the Research Grant Council stabilization measures during construction of a national

540
 expressway in Chongqing, Southwestern China, Proceed- Yue, Z.Q., Lee, C.F. & Tham, L.G. 2004c. Automatic drilling
ings of International Conference on Landslides: Causes, process monitoring for rationalizing soil nail design and
Impacts and Countermeasures, June 17–21, 2001, Davos, construction, Proceedings of the 2004 Annual Seminar of
Switzerland, pp. 625–633. HKIE Geotechnical Division, Hong Kong, China, May 14,
Yue, Z.Q., Lee, C.F., Law, K.T., Tham, L.G. & Sugawara, J. 2004. pp. 217–234.
2001. HKU drilling process monitor and its applications Yue, Z.Q. 2004. Automatic monitoring of drilling process
to slope stabilization, Proceedings of Annual Confer- for optimizing ground anchorage, Proceedings of the 8th
ence—Works Bureau and Ministry of Construction on the Conference of Chinese Association of Rock Mechanics
Development and Co-operation of the Construction Indus- and Engineering, Chengdu, Oct. 2004, Science Press,
try of the Mainland and Hong Kong, 20–21 September, pp. 879–886 (in Chinese).
2001, Vol. 2, pp. II35–II46. Yue, Z.Q. 2005. Automatic drilling process monitoring for
Yue, Z.Q. & Lee, C.F. 2002. A plane slide that occurred dur- soil and rock strengths and their spatial distribution in
ing construction of a national expressway in Chongqing, ground. Proceedings of the 2nd World Forum of Chinese
SW China, Quarterly Journal of Engineering Geology and Scholars in Geotechnical Engineering, Nanjing, China.
Hydrogeology, 35: 309–316. August 22–22, 2005. pp. 85–90 (in Chinese).
Yue, Z.Q., Lee, C.F., Law, K.T. Tham, L.G. & Sugawara, J. Yue, Z.Q., Gao, W., Chen, J. & Lee, C.F. 2006. Drilling
2002. Use of HKU drilling process monitor in soil nailing process monitoring for a wealth of extra factual data
in slope stabilization, Chinese Journal of Rock Mechanics from drillhole site investigation, Proceedings of the 10th
and Engineering, 21 (11): 1685–1690. (in Chinese) international congress of the international association
Yue, Z.Q., Guo, J.Y., Tham, L.G. & Lee, C.F. 2003. of engineering geology, Nottingham, United Kingdom,
Application of HKU DPM in automation of geotechnical 6–10 September 2006. Theme 5—Urban Site Investi-
design and construction, Proceedings of the First National gation, paper number: 5–746, page 1–10 (CD RAM
Congress on Geo-Eng China. Vol. 1, pp. 147–155. softcopy).
China Communications Press, ISBN 7-114-04724-X.Oct. Yue, Z.Q., Chen, J. & Gao, W. 2007. Automatic drilling
22–25, 2003, Beijing, China. process monitoring (DPM) for in-situ characterization of
Yue, Z.Q., Lee, C.F., Law, K.T. & Tham, L.G. 2004a. Auto- weak rock mass strength with depth, Proceedings of the
matic monitoring of rotary-percussivedrilling for ground 1st Canada-US Rock Mechanics Symposium, editors: Erik
characterization—illustrated by a case example in Hong Eberhardt, Doug Stead, & Tom Morrison, Vancouver,
Kong, International Journal of Rock Mechanics & Mining Canada, 27–31 May 2007. Taylor & Franics, London.
Science, 41: 573–612. Vol. 1, pp. 199–206.
Yue, Z.Q., Guo, J.Y., Tham, L.G. & Lee, C.F. 2004b. Drilling Zheng, Y.R., Chen, Z.Y., Wang, G.X. & Ling, T.Q. 2006.
process monitoring for ground characterizations during Engineering Treatment of Slope & Landslide, China
soil nailing in weathered soil slopes, Proceedings of the Communication Press, Beijing.
2nd International Conference on Site Characterization
(ISC-2), Porto, Portugal, September 19–22, 2004. Vol. 2,
1219–1224.

541
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Types of cutslope failures along Shiyan-Manchuanguan expressway through

the Liangyun fracture, Hubei Province

Haiying Zhao & Ren Wang

State Key Laboratory of Geomechanics and Geotechnical Engineering, Institute of Rock and Soil Mechanics, Chinese
Academy of Sciences, Wuhan, China
Institute of Rock and Soil Mechanics the chinese Acadeny of Sciences, Wuhan, China

Jianhai Fan & Wei Lin

Headquarters of Shi-Man Expressway Construction, Shiyan, China

ABSTRACT: The Shiyan-Manchuanguan expressway is an important section of the Yinchuan-Wuhan inter-

provincial Expressway. The road must go through different geological structure units and Liangyun Fracture.
Owing to the unique geological conditions and high-steep cut slopes, slope deformation and failure have so
frequently happened in the course of construction as to build up serious pressure and threaten to the normal
constructing and traffic safety after completing. Through lots of field investigation and study, the slope defor-
mation and failure mainly are two types: landslide and dilapidation. The landslides occur usually in interbed
between soft rock and hard rock. The rainstorm between May and October is the peak of slope deformation and
failure. In order to relieve or prevent the economic losses caused by the geologic hazards, the paper studied the
mechanisms of different landslides.

1 GENERAL SITUATION OF THE PROJECT this sequence of metamorphite strata has developed
penetrative tectonic foliations, ductile-brittle fracture,
The expressway starting from Shiyan in Hubei brittle fracture, and diplogenetic fold, which does
province to Manchuanguan in Shaanxi Province is a great damage to the initial rock solidity (Zhang
an important section of Yinchuan-Wuhan Express- Guo-wei, etc., Che Zi-cheng etc.). Therefore the
way, one of the eight interprovincial expressways rock mass gets greatly fragmented and develops joint,
planned. The expressway is designed to begin with cleavage, and weak intercalation. The destructive
Xujiapeng, the end of Xiangfan-Shiyan Expressway, forms of strata lithology are controlled by differ-
passing through Shiyan, Yun county and Yunxi county, ent scheduled structural time, deformable system,
and end up Manchuanguan town in Shaanxi Province. deformable environment and structural pattern of
The road extends in SE-NW direction and is 107 km Liangyun fault. Due to such complex geognostic and
long. There are 107 bridges and 28 tunnels, the geological conditions, the construction of this section
percentage accounting for 45.71%. Constructed in of reservoir is harsher. There are 409 slopes in all along
November 2004 and completed and opened to traf- the whole line. The height of the excavated slopes in
fic in 2008, the road will play a vital role to ensure the district is generally from 24 m to 50 m, and the
smooth transportation on the highway north and south. tallest one reaches 70 m or so. And the slopes buckling
The land along the expressway is higher in north-west collapse are extraordinarily serious with 42 landslides
than in south-east. With high altitude, its direction is and collapsing disasters having occurred. In addition,
consistent with that of the regional mountains. Most lots of latent infection excavated slopes may result in
sections of the Shiyan-Manchuanguan expressway lie buckling deformation and collapse triggered by cer-
in the Daba mountainous area south of the Qinling tain external factors, rainstorm in particular. All in all,
Mountains where the master stratum is metamorphic landslide and excavated slope treatment has become
rocks. This makes it inevitable to go through differ- one of the most important and most serious bottle-
ent geological structure units and Liangyun Fracture. necks which will seriously influence the construction
Under multi-stage tectonic movement and weathering, schedule of the whole project.

543
2 REGIONAL GEOLOGIC SETTING

2.1 Topography and physiognomy

The Shiyan-Manchuanguan Expressway is situated in
the northwestern mountainous area of Hubei Province,
belonging to the southern piedmont of Qinling Moun-
tains. The terrain slopes eastward with distinct ris-
ing and falling. Its altitude ranges from 140 m
to 1000 m. The Mountains trend northwest, so
does the Shiyan-Manchuanguan Expressway. From
west to east, the physiognomy is tectonic denuda-
tion middling-mountains, low-mountains and hills.
According to the genesis and surface configura-
tion, the morphologic feature can be divided into
three geomorphic units. They are tectonic denuda-
tion midding and low mountains which distribute
over K30∼K58 and K81∼K107, tectonic denudation
low-mountains and hills in K0∼K29 and K59∼K80, Figure 1. A real structure outline map. 1. Mesozoic-
erosion and deposit river valleys mostly on either side Cainozoic, 2. Palaeozoic, 3. Upper Sinian, 4. Lower
of the bank of the Han River and her tributary. Sinian (Yaolinghe Group), 5. Middle-Upper Proterozoic
(Wudang Group), 6. Main fault and coding: ① Qingfeng
Fault, ② Gonglu Fault, ③ Liangyun Fault, ④ Zhunshan Fault,
2.2 Stratum and Lithology 7. Danjiangdou Reservoir.
In terms of chronologic age and lithology of the
stratum, there crop out Proterozoic Wudang Group,
Sinian, Cambrian, Ordovician, Devonian, Cretaceous, Qinling Mountains and at the northwestern margin
Tertiary and Quaternary (Communicational Planning of Wudang rise. In this area geological structure is
and Design of Hubei Province, 2003, Second Express- complex, which appears as a series of territorial frac-
way Survey & Design Institute of China, 2003). The ture and fold trending northwest. The road extends
lithology is complex and can be divided into three on or beside the two sections of Liangyun Fracture
classes. (Figure 1). It is the very Liangyun Fracture that deter-
Schist: It is the dominating strata, primarily com- mines route selection and architectural competition
prising plakite and mica quartz schist of Wudang of the Shiyan-Manchuanguan expressway. It is one
Group and chlorite schist of Sinian Yaolinghe Group, active rift from epipleistocene and has no potentiality
which accompany partly diabase intrusive body. to give rise to biggish and moderately strong earth-
The surficial rock, that is laminar hypo-hard rock, quake. Influenced by the tectogenetic movement, the
has been seriously eroded. In contruction area rock in the neighborhood of the fracture is in the poor
schist is distributing along K0+000∼ K58+700 and state of completeness and fragmented, while the atti-
K75+700∼K89+100 where there are local sections of tude of other rock out of the fracture affected zone
glutenite and quaternary deposit. is stable, extending northward and eastward, and the
Limestone: Distributing along K89+100∼K107+ angle of dip is 40◦ ∼80◦ .
300. Middle bedded and heavy bedded dolomie, lime-
stone with shale band and phyllite band that belong
to laminar hard rock and hypo-hard rock crop out 2.4 Weather and hydrology
along K89+100∼K100+100, whereas phyllite clip-
ping laminal limestone belonging to hypo-hard rock, The research area has abundant atmospheric water.
which is very soft and seriously eroded, crop out along The average rainfall rate per year is 696.1∼900.0 mm.
K100+100∼K107+300. Rainfall is concentrated between July and September
Granulitite: The granulitite of Proterozoic Wudang annually, which accounts for 39∼48% of precipita-
Group with discontinuously distributed pycnophyl- tion all year round. In this region, the rainfall is charac-
lite quartz schist intercalation crops out along terized by continuity, concentration and intensity. And
K58+700∼K75+700. surface runoff is developed. There is a wide variety of
groundwater, which can be classified into loose stuff
intergranular water, clastic rock stuff intergranular
2.3 Geologic composition
water and crack water, basement rock crack water, car-
The region of research interest is located at the south- bonate rock karstic water according to its occurrence
ern margin of Indo-China orogenic belt south of conditions.

544
2.5 Earthquake by different deformation mechanism fault at different
forming stages, the deformation of faulted struc-
This region is situated in the Jianghan seismic belt of
ture reflects obvious zoning, which is schistose zone,
central China’s seismic area where the continental seis-
mylonite zone, cataclastic rock zone, and tectonic
mic activity is weak. Referring to ‘‘Zoning map of the
breccia zone.
ground shock parameters in China’’ (GB18306–2001),
with 10% of exceeding probability during a period of
50 years, the bedrock peak ground accelerations is
0.05 g and the characteristic period is 0.35 s. So the 4 ANALYSIS OF TYPES OF SLOPE
earthquake intensity is VI degree. DEFORMATION AND FAILURE

Because of complex geologic structure and artificial

influence, the main types of slope deformation and
3 DISTRIBUTION CHARACTERISTICS AND
failure include landslide, collapse or rock fall, which
DEFORMATION MECHANISM OF LIANGYU
have a detrimental impact on the highway.
FRACTURE

3.1 Structural feature of the fracture 4.1 Slope engineering geology

From Nanxiang Basin Liangyun Fracture extends in a In the geologic hazard distributing graph of Hubei
NW direction with a length of 250 m. By way of Yun Province, the roadworks lies in the high occurrence
county, Yunxi county, and Manchuanguan county, the zone of landslides and collapse hazard in the north-
fault joins with Shanyang fracture extending in an EW western Hubei Province. All of the cut slopes are
direction. The primary fractured surface trend to NE manual excavation. According to engineering geology
and its angle of dip is 50◦ –70◦ . Only in late meso- investigation from September to November in 2006,
zoic and cainozoic does it change into a SW direction there are 409 cut slopes, of which 15 slopes remain
beside marginal basin. Liangyun fracture is composed uncut and one has been excavated out by a unit in
of a series of parallel or approximately parallel, and charge of construction. The statistics (393 cut slopes
synthetic branch reverse thrust fault with different tec- involved) show that at present the percentage of 227
tonic deformation. On the ground the fault takes a slopes being in the stable condition is 57.76%, the
form of linear clough, fault cliff and fault facet devel- percentage of 83 slopes being in the hypo-stable con-
oped tens of kilometers long. Moreover, in the stretch dition is 21.12%, the percentage of 42 slopes being
zone beside the fault there are a series of concomi- in the critical state is 10.69%, and the percentage of
tant block mountains of the transtorm squeezing zone 4 slopes being in the unstable condition is 10.43%.
and fault basins of the transtorm stretch zone. Finally, Table 1 shows the statistics of damage to all the cut
in the dimension the contemporary basic geomorphic slopes along the expressway.
feature behaves as fault morphology (Zhang Guo-wei
et al. 2001, Chen Shu-jun et al. 2004).
4.2 Landslide
3.2 Analysis of the deformation mechanism 4.2.1 Bedding slip
Within the whole line there are two types of bedding
3.2.1 Tectonic styles of fracture landslide (Table 2), one of which is schist landslide and
Liangyun fracture comprises Xinping-Manchuanguan- the other is red sandstone landslide. The stability of
Yunxi fault and Yunxian-Guanghua fault which are different landslide types has huge distinction because
arrayed in right en-echelon formation. Beside the frac- of dissimilar characteristics of the structural surface
ture and on the top and the lap joint of the fracture, of rock mass.
there are associated with a series of fastigiate fault, The red sandstone belongs to soft rock. The bed-
shear pull-apart red beds basin, and shear-compressive ding surface interpenetration rate generally reaches
fold (Zhang Guo-wei, et al. 2001, Chen Shu-jun et al. 100%. And the bedding surface spacing is less than
2003). 30 cm, with weak intercalation. This type of slope
failure owes to slipping along bedding weak inter-
3.2.2 Deformation mechanism of fracture calation (Chen Zu-yu et al. 2005, Zhang Xian-gong
On the basis of field structure analysis and existing et al. 2000). Because the inclination of bedding weak
information analysis, it is considered that there are at intercalation is low and the layer of it is thick, the
least three stage metamorphoses in Liangyun fracture scale of sliding mass is larger. For instance, No.0728
belt. They are ductile thrust fault at the early stage, slope lies in the left of Yunxi county service area and
friability strike-slip deformation action at metaphase, is about 24 m high. The attitude of rock stratification
ductile abnormal fault at the late stage. Influenced composing the slope is 33˚∠19◦ and its thickness is

545
Table 1. Statistics of all cut slopes failure mode.

Unit/ Unstable
Failure mode Type of slope Terrain and lithology percentage slope Unit

Landslides Bedding slip Clockwise parallel slope Red sandstone 14/3.6 2

Schist 40/10.2 17
Insequent slip Anticlockwise parallel slope Schist 57/14.5 17
Steep dip parallel slope Schist
Wedge slip Clockwise crossing slope Schist 81/20.6 15
Anticlockwise crossing slope Schist 100/25.5 15
Fossil landslide clayey soil 3
Schist 2
Soil (akin soil) landslide Soil slope Clay 35/8.9 7
Rock & soil slope Elluvium and schist 4/1.0 1
Collapse Limestone slope Limestone 18/4.6 2
Crossing slope Schist 10/2.5 0
Diabase slope Diabase 20/5.1 11

Table 2. Distribution range of bedding slip. The slopes of soft and hard interbedding are in the
majority of schistous slopes. The bedded plane trend-
Type of slope Chronologic age Group Section ing the slope is well developed and badly eroded into
quite fragmented rock mass. Owing to filling mud
Red sandstone Cretaceous Paomagang VII
between interlamination, the shear strength of rock
Group
sthist Proterozoic Wudang Group I ∼VIII mass suffers extreme deviation. In the case of the
Sinian Yaolinhe Group VIII, IX structural surface complete interpenetration, bigger
interbedded spacing and filling mud between interlam-
ination, the slope off which the lower footslope is cut to
form free rock mass is easy to come into deformation
in the direction of free face under the action of rain-
storm. When the slope angle is bigger than schistosity
Crack in
quaternary angle, this slope slips along the plane of schistosity
bedding plane and the glide plane is smoothing and flattening (Xu
strata
in strong Bang-dong et al. 2001, Chen Zu-yu et al. 2005). If the
weathered rock terrane is cut off a little, large-scale bedding creep-
External stress cause ing will happen. When the slope angle is smaller than
interpenetrate of schistosity angle, glide and curve will take place and
structural plane at last this slope will be going to its yield limit to fail-
ure. Take No.0917 slope which is located in the left
side of K83+620+860. It has five steps and is 40 m
high. The schistosity attitude of rock composing the
slope is 38◦ ∠ 44◦ . Since the basal slope is destroyed as
Figure 2. Landslide sketch map of No.0728. the slope is excavated, large-scale bedding slip occurs
from peak to foot. The drop in level of the landslide
10∼40 cm. There are three joints whose attitudes are is over 60 m and the primary glide direction is 55◦ .
5◦ ∠87◦ , 65◦ ∠77◦ , 179◦ ∠45◦ respectively. Excavating The vertical drop height at the trailing edge of land-
the cut slope made its foot unshored and at the same slide extends to more than 12 m. The leading edge
time rainfall infiltration along vertical joint sheared off gets over roadbed to pile up at the foot of the opposite
strata, so large-scale landslide is induced (Figure 2). slope (Figure 3 and Figure 4). This destroyed type is
After the landslide took place, the landslide trailing the chief destructive form, such as No.0837, No.0125,
edge generated many vertical cracks and perforated No.0410, and No.0516.
transversal cracks, as wide as 3 m. Many Golgotha
on the sliding mass are pulled apart. The leading edge 4.2.2 Insequent slip
of the landslide advances to the roadbed median line. The countertendency slope and steep dip slope are apt
This is a large-scale landslide, with the landslide mass to insequent slip (Table 3). Under the action of self-
200 m long and 30 m thick. weight and the crack perpendicular to bedding plane,

546
 Altitude(m)
540 Trailing edge
Crack
530

520

510 4
Road
Phyllite centerline
500
Tentative slide plane
490 25˚ ∠ 4
Phyllite
480 0 20 40 60 80

Figure 3. Sectional drawing of No.0917 slope.

Figure 5. Imbricate fissure on the mountaintop of No.0615.

Table 4. Distribution range of wedge landslide.

Chronologic
Type of slope age Group lithology Section

proterozoic Wudang plakite

group Albite I, V∼IX
schist
Quartz
schist
contertendencyl proterozoic Wudang II, IV, V
cross slope group VI, IX

Figure 4. Sketch map of the destroyed No.0917.

and covered the roadbed. After landslide is over, we

Table 3. Distribution range of insequent slip. can see a score of imbricate parallel cracks range along
the natural slope to the mountaintop. The crack extends
Chronologic
as wide as 30 cm and the fracture spacing is 1∼3 m
Type of slope age Group lithology Section
(Figure 5).
Countertendency proterozoic Wudang plakite III∼VIII
parallel slope group Albite 4.2.3 Wedge landslide
schist Clockwise oblique crossing schist slope and counter-
Steep dip proterozoic Wudang II∼IV tendency oblique crossing schist slope are apt to this
parallel slope group kind of collapse mode (Table 4).
The failure mode slide badly along the plane com-
posed by bedding plane and two joint planes. No.0704
the rock mass collapses in the direction of free face slope is one typical wedge landslide, which is situ-
like domino (Chen Zu-yu et al. 2005). As to single for- ated on the left of K59+590∼+690 and is 32 m high.
mation, flexural deflection similar to cantilever beam This slope is composed of thick layer clipping lami-
occurs. On the ground, there are a series of imbri- nar layer albite-quartz phyllite. The schistosity plane
cate parallel cracks taking the shape of ‘‘V’’. When (36◦ ∠62◦ ) and ‘‘X’’ shear joint plane (229◦ ∠44◦ ,
the dump deformation reaches to a certain degree, the 65◦ ∠83◦ ) contribute the sliding wedge mass (Figure
maximum buckling distortion of all the strata will be 6).
broken to form one perforated failure plane to induce
whole slope slipping. No.0615 is a typical slope sit- 4.2.4 Resurgent fossil landslide
uated at the left of K52+520∼+650, which has five Because partial skid resistance of the fossil landslide
steps and is 40 m high. The natural slope angle is is excavated under construction, the fossil landslide
over 50◦ and the rock mass composing the slope is relives or relives partly in the condition of abundant
laminal schist rich in pinal and falls to pieces. As a rainfall. This typical landslide is the left slope at
consequence of continuous rainfall after excavation, K50+745∼K51+103 (Figure 7). As the first step of
the slope slipped from the first step to the third step the slope has been excavated in April 2005, large-scale

547
 Figure 8. Imbricate fissure of soil landslide.

Figure 6. Wedge landslide of No.0704.

Figure 9. Soil landslide.

Figure 7. Resurgent fossil landslide at K50+745∼

K51+103.

deeper layer landslide begins to appear. The landslide

is about 360 m wide, 100 m long, and 15–50 m thick.
The primary slipping direction is NE35◦ . At present,
the leading edge of the fossil landslide has revived
and pushed its way in the roadbed and formed obvi-
ous deformation, which is threatening the ground to
produce the girder. Another example is a revived clay
fossil landslide at K71+880∼+950.

4.2.5 Soil (akin soil) landslide

The formulation of soil (akin soil) landslide is homo- Figure 10. Akin soil landslide.
geneous. After being excavated, the slope is apt to
creeping in the case of rainfall. The glide plane appears
circular arc. Failure mode is to pull behind (Figure 8). width. In the middle of the akin soil slope there are
For instance, the right slope at K43+975∼+875 is slightly weathered rocks and on both sides are fully
8–16 m high and its angle of natural slope is smaller eroded elluvium and sliderock. In the course of exca-
than 10◦ (Figure 9). After the whole slope slumped, the vation the intrinsic stress equilibrium is broken. At this
trailing edge is pulled apart and drop height reaches point rainfall can straightly seep into the slope along
8 m. In addition, the slope surfacing has been upheaved manual excavation surface, which leads the slope to
and there are many fissures and fracture zone, and at buckling collapse. The primary slide face is along the
present the last crack has extended to the top of the contact zone of hard schist and soft schist. The primary
second slope protection located 15 m far away. The slide direction is not only along the aspect of slope but
bulge of the leading edge is 4∼5 m length and is 15 m also along the direction of the slope side (Figure 10).

548
4.3 Collapse ACKNOWLEDGEMENT
Within the rock mass of slopes which are diabase
Supported by Project of National Natural Science
slopes, limestone slopes and crosscut schist slopes
Funds: Catastrophe mechanism of cascade collec-
there do not exist large interpenetrate constructional
tion and selective treatment for local debris flows
surface arising slope slumping. Instead there exist
(4067 2193); Self-organized criticality and balance-
many groups of joint and the joint plant is seriously
deviating mechanism of debris flows collapsed
eroded and filled up with clay, so the whole slope is
deposits (50709035)
stable. In this case, the collapse mode of the slope is
rock failure, and the slope failure is on a small scale.
REFERENCES
5 CONCLUSION AND SUGGESTIONS Zhang Guo-wei, Zhang Ben-ren, Yuan Xue-cheng, et al.
2001. Qinling orogenic belt and continental dynamics.
1. In the region, stratum and geologic structure are Beijing: Science press, 221–261.
simple. Wudang group crops out along the express- Xu Bang-dong. 2001. Landslide analyzing and controlling.
way, which is controlled by Liangyun fault zone. Beijing. China railway publishing house, 140–163.
The schist of Wudang group is rich in anthrophyl- Chen Zu-yu, Wang Xiao-gang, Yang Jian, et al. 2005. Rock
lite and the rock beside fault falls to pieces. In this slope stability analysis. Beijing: China WaterPower Press,
condition there exist landslide and collapse. Land- 51–151.
slide is one of the geologic hazards with the highest Zhang Xian-gong, Wang Si-jing, Zhang Zhuo-yuan, et al.
2000. China engineering geology. Beijing. Science press,
frequency, largest scale, most serious damage and 186–216.
most complex mechanism. Chen Shu-jun, Liu Suo-wang, Yao Yun-sheng, et al. 2004.
2. Listing all the slopes in descending order of insta- Research on structural analysis and quaternary slip rate
bility, they are the consequent parallel schist slopes, of Liangyun fault. Journal of geodesy and geodynamics,
soil slopes, countertendency parallel schist slopes, 24 (3):60–66.
and countertendency crossing schist slopes. Chen Shu-jun, Liu Suo-wang, Huang Guang-si, et al. 2003.
3. The mechanism of all the landslides existing along Report of seismic safety evaluation of Huojuling tunnel,
the expressway is that because the foot of the exca- Erdaoya tunnel and Yunling tunnel under the influence of
vated slopes is free, the original stress equilibrium Liangyun fault. Wuhan. Institute of seismology, CEA.
Communicational Planning and Design of Hubei povince.
is destroyed and the slope slumps along the direc- 2003. The specification of engineering geology advanced
tion of the free face. In addition, rainstorm is the exploration of the section from Shiyan to Hongyanzi in
dominant factor inducing slope deformation and Hubei province.
failure. Most of the slopes are destroyed in the rain- Second Expressway Survey & Design Institute of China.
storm from May to October. Moreover, the compact 2003.The specification of engineering geology explo-
district of landslide is dispersed over the zone of ration of the section from Hongyanzi to Manchuanguan
soft rock and hard rock interbedding. in Hubei province
4. Too much explosive is forbidden to use in the
course of construction in order to prevent rock
mass quality from declining and to protect the
slopes against slump. Presplitting blasting is rec-
ommended. Besides, it should be excavated step
by step while getting done with reinforcement and
waterproof in time.

549
Advances in analytical methods, modeling
and prediction of slope behavior
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Probability limit equilibrium and distinct element modeling of jointed rock

slope at northern abutment of Gotvand dam, Iran

M. Aminpoor, A. Noorzad & A.R. Mahboubi

Power and Water University of Technology, Tehran, Iran

ABSTRACT: Rock slope stability analysis has improved significantly in the last few years utilizing new
advanced numerical techniques. Application of these new methods have made it possible to realize complexities
relating to geometry, material anisotropy, non-linear behavior, in situ stresses and the presence of several coupled
processes (e.g. pore pressure, seismic loading, etc.). It is also provided for researchers to understand how initial
instability mechanisms become followed by or preceded by creep, progressive deformation and extensive internal
disruption of the slope mass. In this paper, it is demonstrated that the numeric distinct element method is useful
to perceive some local instable parts of a rock slope due to earthquake loading that it is not possible to anticipate
by limit equilibrium analysis.

1 INTRODUCTION be complex and not easily allowed for in simple static

analysis (Eberhardt, 2003).
Rock slope stability analyses are routinely performed
and directed towards assessing the safe and functional
design of excavated slopes (e.g. open pit mining, road
2 PROJECT DESCRIPTION: ENGINEERING
cuts, etc.) and/or the equilibrium conditions of natural
AND STRUCTURAL GEOLOGY
slopes.
OF GOTVAND DAM SITE
Nowadays, a vast range of slope stability analysis
tools exist for both rock and mixed rock-soil slopes;
2.1 Dam site geology
these range from simple infinite slope and planar
failure limit equilibrium techniques to sophisticated The Neogene sediments of the Fars Group (Gachsaran,
coupled finite-/distinct-element codes. Mishan and Aghajari Formations) outcrop in the vicin-
Selection of the method of rock slope stability anal- ity of the dam site. They include thick evaporitic
ysis depends on both site conditions and the potential units (Gachsaran), marl, limestone and alternating
mode of failure, with careful consideration being given sandstone, siltstone and claystone. Plio-pleistocene
to the varying strengths, weaknesses and limitations Bakhtyari conglomerates unconformable overlain
pertinent to in each methodology. these fine-grained sediments. The rock foundation of
Given the wide scope of numerical applications the dam consists of Aghajari and Bakhtyari Forma-
available today, it has become essential for the prac- tions.
titioner to fully understand the varying strengths and Figure 1 shows a geological cross-section of the
limitations inherent in each of the different method- dam site and also cross section of northern abutment
ologies. For example, limit equilibrium methods still of dam. The rock at the site comprises deposits of the
remain the most commonly adopted solution method Aghajari formation with interbedded claystone, silt-
in rock slope engineering, even though most failures stone and sandstone (AJn), overlain by conglomerate
involve complex internal deformation and fracturing of the Bakhtiari formation (BKn). The sandstone and
which bears little resemblance to the 2-D rigid block siltstone are indurated and fissured while the clay-
assumptions required by most limit equilibrium back- stone appear as heavily over consolidated hard clay.
analyses. The factors initiating eventual failure may The conglomerate is for the most part indurated and

553
 of this formation. The sedimentation of this forma-
tion is related to the river-flood plane deposits. This
formation belongs to late Miocene-Pliocene and com-
poses limey (approx-70%) and silicic (fundamentally
Chert-30%) grains. The cementation is generally limy.
The thickest and thinnest layers are sandstone and
Claystone respectively.
Lateral variations in AJ formation are apparently
observed when assessing the boreholes; as in lateral
parts, sandstone is changed to siltstone and mudstone.
In general, the dominant lithology of the station is
formed by siltstone. Sandstone layers are 0.20–0.30
up to 5 m thick. Fine, tight lamination is charac-
teristic for the siltstones and siltstone/sandstone or
siltstone/claystone transitions.

2.2 Structural geology: Jointing and disposition

of the strata in DBK-URB
One set is the bedding, dipping towards south with
a dip angle of 30 to 60◦ with 45◦ as a mean (set
‘‘B’’). The bedding planes, with few exceptions, can
be considered as tight.
The second set is perpendicular to the bedding
planes (which is commonly offset), dipping around
Figure 1. A) Schematic geological cross-section of the dam 45◦ to the north. Spacing is between 1 to 3 m. Their
site B) Cross-section of northern abutment showing different surfaces are generally undulating, while the small
strata and formations. scale roughness profiles are variable, from slicken-
side to rough. Frequently filled with clayey and/or
silty material, their width varies between a few mil-
moderately strong but contains lenses and beds with limeters to 30–50 centimeters. Shearing along these
poor cementation. The rock is massive in the left val- discontinuities has been mostly undertaken by the fill-
ley side and rises as a steep cliff some 300 m above ing material. The conglomerate walls show different
the valley floor. conditions, from saw-cut, slightly disturbed to bro-
ken matrix and, where greater displacement occurred,
2.1.1 Bakhtyari formation mylonite.
BK formation is basically formed by conglomerate. Some erratic joints dipping towards north or north-
Massive conglomerate units, which can be as thick as east, are filled with broken conglomerate in a silty
100 meters, are only separated by a few sandstone and clayey matrix.
claystone interlayers. Bedding planes, usually tight,
are mostly marked by sandy/silty interbeds, lenses or,
in some cases, by the orientation of the elongated
pebbles. Pebbles of variable size, from 2 cm up to 3 PROBABILITY LIMIT EQUILIBRIUM
30 cm, generally rounded to sub-rounded, are mainly ANALYSIS
limestone fragments. Chert and sandstone or siltstone
pebbles are subordinate. The cement varies from cal- 3.1 Application of limit equilibrium analysis
careous to siliceous, commonly finely crystallized but in rock slope stability problems
sometimes with a coarser, sandy texture. Variable Limit equilibrium techniques are routinely used in the
degree of consolidation or washout of sandy bound analysis of landslides where translational or rotational
materials throughout the thick sequence resulted movements occur on distinct failure surfaces. Analy-
locally in loosely cemented, friable conglomerates. ses are undertaken to provide either a factor of safety
or, through back-analysis, a range of shear strength
2.1.2 Aghajari formation parameters at failure. In general, these methods are the
This formation is generally formed by a sequence of most commonly adopted solution method in rock slope
brown to gray calcareous sandstone with some interbe- engineering, even though many failures involve com-
ded mudstone, marlstone and some siltstones. Thin plex internal deformation and fracturing which bears
veins of gypsum are spread out in some horizons little resemblance to the 2-D rigid block assumptions

554
required by limit equilibrium analyses. However, limit simulation, almost all input parameters can be
equilibrium analyses may be highly relevant to simple assigned a probability distribution, and Monte Carlo
block failure along discontinuities or rock slopes that scheme is then used to compute a probability distribu-
are heavily fractured or weathered (i.e. behaving like tion of the resulting safety factors.
a soil continuum). As is well known, many natural data sets follow
For very weak rock, where the intact material a bell-shaped distribution and measurement of many
strength is of the same magnitude as the induced random variables appear to come from population fre-
stresses, the structural geology may not control sta- quency distributions that are closely approximated by
bility and failure modes such as those observed in normal probability density function. This is also true
soils may occur. These are generally referred to as cir- for many geotechnical engineering material proper-
cular failures, rotational failures or curvilinear slips ties. In this paper, the input data have been considered
(Eberhardt, 2003). to have a normal probability distribution function.
Considerable advances in commercially available For description of uncertainty in a data set, we can
limit equilibrium computer codes have taken place in use the covariance coefficient that is equal to
recent years. These include:
SD
Cov = (1)
– Integration of 2-D limit equilibrium codes with X
finite-element groundwater flow and stress anal-
where SD is standard deviation and X  is the mean
yses (e.g. Geo-Slope’s SIGMA/W, SEEP/W and
SLOPE/W—Geo-Slope 2004). of data.
– Development of 3-D limit equilibrium methods (e.g. Cov coefficient for input parameters of this inves-
Hungr et al. 1989; Lam & Fredlund 1993 and tigation has been chosen to be 0.1. Therefore, the
SLOPE/W—Geo-Slope 2004). standard deviation of all input parameters has been
– Development of probabilistic limit equilibrium considered to be 10% of mean quantity. In order
techniques (e.g. SWEDGE—Rocscience 2001b; to perform the probability analysis on limit equilib-
ROCPLANE—Rocscience 2001c). rium results, shear strength parameters (C and ϕ) and
– Ability to allow for varied support and reinforce- unit weight of all materials have been entered in the
ment. program whit a uncertainty of 10%.
– Incorporation of unsaturated soil shear strength A factor of safety is really an index indicating the
criteria. relative stability of a slope. It dose not imply the actual
– Greatly improved visualization, and pre- and post- risk level of the slope, due to the variability of input
processing graphics (Stead et al. 2005). parameters. With a probabilistic analysis, two useful
indices are available to quantify the stability or the
risk level of a slope. These tow indices are known as
3.2 The SLOPE/W software probability of failure and the reliability index.
The reliability index (β) is defined in terms of
SLOPE/W is the leading software product for com-
the mean (X ) and the standard deviation (SD) of the
puting the factor of safety of earth and rock slopes.
factors of safety as shown in the following equation:
SLOPE/W, can be used to analyze both simple and
complex problems for a variety of slip surface shapes,  − 1.0)
(X
pore-water pressure conditions, soil properties, analy- β= (2)
sis methods and loading conditions. SD
Using limit equilibrium, SLOPE/W can model het-
erogeneous soil types, complex stratigraphic and slip Table 1. Geotechnical properties of materials of dam site
surface geometry, and variable pore-water pressure and dam body.
conditions using a large selection of soil models.
Analyses can be performed using deterministic or Unit weight Cohesion Friction
probabilistic input parameters. Stresses computed by (kN/m3 ) (MPa) angle (degree)
a finite element stress analysis may be used in addi-
Region Saturated Dry Saturated Dry Saturated Dry
tion to the limit equilibrium computations for the most
complete slope stability analysis available. AJn 25 23 0.540 0.500 26 29
S.AJ 24 17.5 0.440 0.500 23 25
BKn 51 35 0.980 0.690 23 25
3.3 Probability limit equilibrium analysis DBK 40 29 0.390 0.460 23 25
of the rock slope at northern abutment INF 37.7 29 0.832 0.640 24 25
of Gotvand dam BKs 42.9 33 1.300 1.000 24 25
Core of
There is an ever-increasing interest in looking at sta- dam 20 28 0.100 0.000 19.1 20.1
bility from a probabilistic point of view. In SLOPE/W

555
 Table 2. Results of probability limit equilibrium analysis on northern abutment for the time before
dam construction.

Probability of Probability of
Analysis Slope Reliability FS decrease FS decrease
type condition FSmean index less than 1.5 (%) less than 1.0 (%)

Static Natural 2.364 14.173 0 0.00

Saturated 1.131 1.864 100 3.50
Pseudostatic Natural 1.666 13.004 0 0.00

This index can also be considered as a way of exploited in developing the solution procedures. In
normalizating the factor of safety with respect to its contrast, discontinuum analysis permits sliding along
uncertainty. and opening/closure between blocks or particles. The
The input data for different rock masses in the underlying basis of the discrete-element method is
slope region are listed in Table 1. Also the results that the dynamic equation of equilibrium for each
of probability limit equilibrium analysis of the rock block in the system is formulated and repeatedly
slope before dam construction considering both static solved until the boundary conditions and laws of
and earthquake event conditions are summarized in contact and motion are satisfied. The method thus
Table 2. accounts for complex non-linear interaction phenom-
It can be seen from the results that earthquake occur- ena between blocks.
rence will not have a considerable effect on the slope Discontinuum modelling constitutes the most com-
stability. This means that total rock slope stability can monly applied numerical approach to rock slope
be reliable during and after earthquake event. But some analysis. Several variations of the discrete-element
local instable regions may be occurred due to rock methodology exist:
mass jointing that limit equilibrium analysis is not able
to demonstrate. – distinct-element method;
Note that completely saturated condition is a pes- – discontinuous deformation analysis;
simistic state of slope which simulates the intense – particle flow codes.
rainfall and it should be considered as a very improb-
able state because of depth of slope and also high dip
of slope that makes a great runoff. 4.2 The UDEC software (Universal Distinct
Element Code)
The distinct-element method, developed by Cundall
4 DISCONTINUUM MODELING
(1971) and described in detail by Hart (1993), was the
first to treat a discontinuous rock mass as an assembly
4.1 Application of discontinuum modeling
of quasi-rigid, and later deformable, blocks interacting
in rock slope stability problems
through deformable joints of definable stiffness. As
Where a rock slope comprises multiple joint sets, such, the numerical model must represent two types
which control the mechanism of failure, then a discon- of mechanical behavior: that of the discontinuities and
tinuum modeling approach may be considered more that of the solid material.
appropriate. Discontinuum methods treat the problem The dual nature of distinct-element codes, for
domain as an assemblage of distinct, interacting bod- example UDEC (Itasca 2000), make them particu-
ies or blocks that are subjected to external loads and larly well suited to problems that involve jointed rock
are expected to undergo significant motion with time. slopes. On the one hand, they are highly applicable to
This methodology is collectively referred to as the the modeling of discontinuity-controlled instabilities,
discrete-element method (DEM). allowing two-dimensional analysis of translational
The development of discrete-element procedures mechanisms of slope failure and are capable of sim-
represents an important step in the modeling and ulating large displacements due to slip, or opening,
understanding of the mechanical behavior of jointed along discontinuities. On the other hand, they are
rock masses. Although continuum codes can be modi- also capable of modeling the deformation and mate-
fied to accommodate discontinuities, this procedure rial yielding of the joint-bounded intact rock blocks.
is often difficult and time consuming. In addition, This becomes highly relevant for high slopes in weak
any inelastic displacements are further limited to elas- rock, flexural-topples and other complex modes of
tic orders of magnitude by the analytical principles rock slope failure (Eberhardt, 2003).

556
 UDEC is a distinct element program for the 2D assumed. As an example, the Poisson effect for an
modeling of jointed rock subjected to quasi-static or isotropic elastic material is
dynamic loading conditions. It simulates large dis-
placements (slip and opening) along distinct surfaces σxx υ
in discontinuous medium treated as an assemblage = (3)
σyy 1−υ
of discrete (convex or concave) polygonal blocks
with rounded corners. Discontinuities are treated as
boundaries between the blocks. Relative motion along Consider the Poisson effect produced by joints dip-
the discontinuities governed by linear and non-linear ping at various angles. The Poisson effect is a function
force-displacement relations for movement in both the of the orientation and elastic properties of the joints.
normal and shear directions. Consider the special case shown in Figure 2. A rock
mass contains two sets of equally spaced joints dip-
ping at an angle, θ, from the horizontal. The elastic
properties of the joints consist of a normal stiffness,
4.3 Selection of deformable versus rigid blocks kn , and a shear stiffness, ks . The contribution of the
in UDEC simulation of jointed rock masses elastic properties of the intact rock will be examined
An important aspect of a discontinuum analysis is the for the case of θ = 45◦ . the intact rock will be treated
decision to use rigid blocks or deformableblocks to as an isotropic elastic material. The elastic properties
represent the behavior of intact material. The consider- of the rock mass as a whole will be derived by adding
ations for rigid versus deformableblocks are discussed the compliances of the jointing and the intact rock.
in this section.
Early distinct element codes assumed thatblocks
were rigid. However, the importance of including
block deformability is now recognized, particularly for
stability analyses of underground openings and stud-
ies of seismic response of buried structures. One of
the most obvious reasons to include block deforma-
bility in a distinct element analysis is the requirement
to represent the ‘‘Poisson’s ratio effect’’ of a confined
rock mass.
Rock mechanics problems are usually very sensi-
tive to the Poisson’s ratio chosen for a rock mass. This
is because joints and intact rock are pressure-sensitive:
their failure criteria are functions of the confining
stress (e.g., the Mohr-Coulomb criterion). Capturing
the true Poisson behavior of a jointed rock mass is
critical for meaningful numerical modeling.
The effective Poisson’s ratio of a rock mass is com-
prised of two parts: (1) a component due to the jointing; Figure 2. Model for Poisson’s effect in rock with joints
dipping at angle θ from the horizontal, and with spacing S.
and (2) a component due to the elastic properties of the
intact rock. Except at shallow depths or low confin-
ing stress levels, the compressibility of the intact rock
makes a large contribution to the compressibility of
a rock mass as a whole. Thus, the Poisson’s ratio of
the intact rock has a significant effect on the Poisson’s
ratio of a jointed rock mass.
A single Poison’s ratio, ν, is, strictly speaking,
defined only for isotropic elastic materials. However,
there are only a few jointing patterns which lead to
isotropic elastic properties for a rock mass.
Therefore, it is convenient to define a ‘‘Poisson
effect’’ that can be used for discussion of anisotropic
materials.
The Poisson effect will be defined as the ratio of
horizontal-to-vertical stress when a load is applied
in the vertical direction and no strain is allowed in Figure 3. Poisson’s effect for rock with two equally spaced
the horizontal direction. Plane strain conditions are joint sets, with θ = 45◦ (blocks are deformable with ν = 0.2).

557
Table 3. Rock material properties, utilized for distinct element UDEC modeling.

Bulk Shear Friction Tensile

Density modulus modulus Cohesion angle strength
Region (ton/m3 ) (GPa) (GPa) (kPa) (deg) (kPa)

D.BK (intact rock) 2.5 10.8 0.500 10,000 45 2,000

BKn (rock mass) 2.5 3.130 0.500 490 35 0

The compliance matrix due to the two equally Table 4. Discontinuity properties, utilized for distinct ele-
spaced sets of joints dipping at 45◦ is ment UDEC modeling.
  Jkn Jks Friction
1 ks + kn ks − kn
C (jointing) = (4) Discon- (GPa/ (GPa/ Cohesion angle Dip Spacing
2Skn ks ks − kn ks + kn tinuity m) m) (kPa) (deg) (deg) (m)

Thus, the Poisson effect for the rock mass as a Joint

whole is set 1 1.7 0.17 20 30 315 1–3
Joint
σxx [υ(1 + υ)] /E + (kn − ks )/(2Skn ks ) set 2 1.7 0.17 20 30 43 2–4
= Bedding 1.3 0.13 20 30 34 1–3
σyy [(1 + υ)(1 − υ)] /E + (kn + ks )/(2Skn ks )
(5)

Eq. (5) is graphed for several values of the ratio

E/(S.kn ) in Figure 3 for the case of ν = 0.2. Also
plotted are the results of UDEC simulations. For low
values of E/(S.kn ), the Poisson effect of a rock mass
is dominated by the elastic properties of the intact
rock. For high values of E/(S.kn ), the Poisson effect
is dominated by the jointing.

4.4 Distinct element modeling of the rock slope

at northern abutment of Gotvand dam
For discontinuum modeling of the northern abutment
rock slope, UDEC code has been used to simulate
both intact rock and discontinuity behavior under
different conditions.
Table 3 has listed the rock material properties uti-
lized in UDEC model. Also Table 4 has presented
discontinuity properties existing in regions used in
UDEC model.
In this paper, the slope has simulated into two main
regions. First is the disturbed Bakhtyari region and
second is the in situ Bakhtyari formation. Disturbed
Bakhtyari is a huge rock mass that include the main
body of the slope and because of its three main dis-
continuity sets, it has the principle role in the probable Figure 4. Moving blocks along bedding strata due to
instability occurrence that should be modeled. There- earthquake loading, showing displacement vectors.
fore, the model has been made by D.BK region that
includes of multiple discontinuity sets and BKS that is of blocks has no effect on model results, just as which
made of continuous rock masses. can be seen from method described in section 4.3.
Three models have been made of rigid, elastic and Built models in distinct element code indicate
elastic-plastic blocks and it can be seenv that rigidity that slope will have a reliable stability under natural

558
conditions. But earthquake loading can cause some REFERENCES
instable planes in surface places that will move down
through middle and toe of slope. These surface planes Eberhardt, E., Kaiser, P.K. & Stead, D. 2002. Numeri-
consist of blocks above surface bedding strata. Figure 4 cal analysis of progressive failure in natural rock slopes.
illustrates moving blocks under seismic loading. In: da Gama, Dinis, e Sousa, Ribeiro (Eds.), EUROCK
2002—Proc. ISRM Int. Symp. on Rock Eng. for
Mountainous Regions, Funchal, Madeira. Sociedade
Portuguesa de Geotecnia, Lisboa, pp. 145–153.
5 CONCLUSION Eberhardt, E. 2003. Rock Slope Stability Analysis—Utiliza-
tion of Advanced Numerical Techniques Geological Engi-
In this study, it is demonstrated that in complex rock neering/Earth and Ocean Sciences, UBC, 6339 Stores
slopes under different conditions, some local instable Rd:4–5, 8–9 & 23–29.
blocks may be formed that simple continuous methods Geo-Slope, 2000. Geo-Slope Office (Slope/W). Geo-Slope
are not able to analyze. As it is observed, limit equilib- International Ltd., Calgary, Canada.
rium analysis of the current rock slope showed that it Itasca, 2004. Itasca Software Products—UDEC. Itasca Con-
will resist under earthquake loading. But discontinu- sulting Group Inc., Minneapolis.
Mahab Ghodss Consulting Engineers, 2007. Overall Review
ous modeling revealed that surface bedding strata will
of Geological and Geotechnical Characteristics of Got-
start to move due to occurrence of earthquake. vand Dam Site.
Therefore, consideration should be given that limit Stead, D., Eberhardt, E. & Coggan, J.S. 2006. Developments
equilibrium techniques should be used in conjunction in the characterization of complex rock slope deforma-
with numerical modeling to maximize the advantages tion and failure using numerical modeling techniques.
of both. Engineering Geology 83: 217–235.

559
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Rock block sliding analysis of a highway slope in Portugal

P.G.C. Santarém Andrade & A.L. Almeida Saraiva

Earth’s Science Department of Science and Technology Faculty, University of Coimbra, Portugal

ABSTRACT: This work analyses the stability problems that exist on a road slope located in the central part of
Portugal, at km 55 of the highway IP3. A joint-study recorded the main characteristics of the joint-sets at two
separate sections of the slope. To assess the relative stability and potential unstability in the rock slope the authors
used several methods of analysis. A kinematics analysis allowed an identification of the discontinuities which
are likely to cause instability problems. The studied slope showed frequent rock block sliding and sometimes
wedge failures. The SMR (Romana) system was used to define the relative susceptibility of the slope’s rock
unstability. The study of the slope’s unstability behaviour is concluded with the evaluation of the factor of safety
using the Mohr-Coulomb and Barton & Bandis models.

1 INTRODUCTION of quartz. The quartzites show a granoblastic tex-

ture. Other minerals of relative importance are present:
This paper analyses a road slope located in the central white micas (sericite and muscovite), albite, ortho-
part of Portugal, showing unstability problems. The clase, pyrite and iron oxides. The quartzites have pyrite
most frequent or the worst unstability situations that minerals deposits mostly very weathered.
have occurred were identified. It was important to Near to the slope surface the quartzites are more
characterize the different parameters that influenced fractured and more weathered (W4) and have a F3
the rock slope stability. The rock mass classifica- spacing.
tion Slope Mass Rating (SMR) proposed by Romana The physical and mechanical characteristics of
(1985) system was used to identify the susceptibility these schists and quartzites can be found on a pre-
of the rock slope to failure. When using the SMR sys- vious work by Santarém Andrade & Almeida Saraiva
tem, the favourability of joint orientations regarding (2000).
slope orientations must be defined.
For the unstable sectors, the factor of safety (FS)
was determined in accordance with the Mohr- 3 METHODOLOGY
Coulomb and Barton & Bandis models.
In a stability analysis the discontinuities orientation
must be determined initially in order to foresee and
2 GEOLOGY prevent situations of unstability.
A graphic kinematics analysis was carried out for
The slope is located at km 55 of the IP3 (Itinerário Prin- the different stability problem situations. Rock blocks
cipal n◦ 3), it is 150 meters long and about 21 meters slides are frequent and they are related with three
high (Figure 1). The studied slope is a 75◦ road cutting major joints families (Figure 1). The occasional wedge
slope and has a strike of N80◦ W. It is situated in the failures occur at the main joint intersections.
Buçaco Ordovician—Silurian synclinal and is mainly The rock blocks are originated by the intersections
constituted by impure quartzites and also by schists of the main joint sets: stratification joints (N85◦ E;
and carbonaceous schists. The later show a black tonal- 75◦ N), (N85◦ W; 31◦ S and N9◦ E; subvertical).
ity and sometimes present a plastic behaviour. They The rock slope’s stability behaviour was studied
are very to completely weathered (W4 to W5) and using a geomechanical classification. A classification
disintegrate easily. The quartzites are the predomi- adopted from the Romana system (1985, 1996)—Slope
nant rocks, while schists and carbonaceous schists are Mass Rating (SMR) was used to define the slope’s
mainly located at the slope’s toe. unstable sectors. The discontinuities characteristics,
The quartzites are less weathered than the car- slope’s orientation, ground water conditions, rock
bonaceous schists. They show whitish, purplish and material strength and rock blocks dimensions are
greyish colours and are constituted by more than 75% the most important parameters of the SMR system.

561
 where FS = factor of safety; c = apparent cohesion;
A = area of the block base; W = weight of block;
ψ = slope angle and φ = friction angle.
The rock block’s weight is related with its dimen-
sions and its specific weight.
Unstability problems are generally associated with
the presence of water and, in this case, the model
of rupture, based on the Mohr-Coulomb equation
(Equation 3), includes a perpendicular fracture to the
landslide surface, which was verified ‘‘in situ’’. The
presence of water causes an accentuated reduction of
the rock blocks stability and causes sliding.

(cA + (W cos ψ − Uw ) tan φ)

FS = (3)
(Wsenψ + Vw )
Figure 1. Rock blocks slides of the studied slope.
where Uw = force related with the water flowing at
the surface between the block and its base and Vw =
This classification was elaborated from the Bieni- force related with the water filling the subvertical
awski’s Rock Mass Rating (RMR) (1979, 1989). The discontinuity plane.
SMR determination is defined on Equation 1: The studied slope corresponds to a structure located
on the superficial part of the rock masses. The rock’s
SMR = RMR − (F1 × F2 × F3 ) + F4 (1) deformability is not an important aspect.
The friction angle values were obtained through
rock joint shear tests. Some of these tests were per-
where F1 , F2 and F3 = adjustment factors related with formed under low normal stresses (smaller than 0.2
the joint-slope orientation; and F4 = factor related MPa) to reproduce as close as possible the situation
with the slope’s method of excavation. The different that occurs ‘‘in situ’’.
factors allowed the RMR correction. To execute the joint shear test several blocks were
The SMR values vary between 0 and 100 and can carefully removed from the slope using picks and
be classified in five stability groups (Romana, 1985). hammers not to damage the discontinuity surfaces.
The values below 21 can be assumed as completely The FS obtained through the Barton & Bandis
unstable and the values above 80 are classified as com- model (1990) was also determined. This model implies
pletely stable. The SMR system can be executed on a the definition of the JRC (Joint Roughness Coeffi-
previous study phase to identify possible unstability cient) values.
situations and to indicate support measures. The SMR
classification considers planar and toppling failures.
The wedge failures are classified as a special type of
plane failures. 4 RESULTS
The present study only analysed only failures with
dimensions superior to 0,005 m3 . The determination The authors proceeded to evaluate the different sta-
of the rock blocks volumes included the disconti- bility problems that occur in the sections of the slope.
nuity spacing and the angles between the joint sets. The SMR (Romana) values (Table 2) were defined and
The spacing for each discontinuity set was defined as compared with the unstability behavior observed ‘‘in
the modal value of the considered intervals for each situ’’ in the last 6 years (Tables 1–2). From the struc-
family. tural geologic survey and kinematics analysis it was
The analysis of the stability of the slope can be done verified that slope unstability is mainly related with
through deterministic and/or probabilistic methods. rock blocks slides.
In a first stage the factor of safety (FS) of the unsta- The slope was divided in 2 sections 50 and 100
ble sectors was determined in accordance with the meters long in order to use the SMR classification
Mohr-Coulomb model. The block slides were anal- and compare its results with the ‘‘in situ’’ values. The
ysed considering the gravity forces and the factor of geologic-structural survey and the laboratorial and ‘‘in
safety was obtained by the Equation 2: situ’’ tests of the geotechnical characterization allowed
the definition of the SMR classification parameters.
The SMR (‘‘in situ’’) value corresponds to the situa-
(cA + (W cos ψ) tan φ)
FS = (2) tion of unstability or rupture that was registered during
Wsenφ six years.

562
Table 1. Unstability reported situations and SMR (‘‘in situ’’) Table 3. FS results from Equation 1.
values.
Factor of safety Factor of safety
SMR considering c without considering c
Slope section Unstability situations (‘‘in situ’’) Joints
safety c φ = 45.5◦ φ = 34.8◦ φ = 45.5◦ φ = 34.8◦
Slope 1 - a Plane failure, rock falls 55
and small wedge failures Quartzites 3.62–1.54 2.45–1.27 0.27–0.20 0.24–0.19
Slope 2 - b Plane failure, rock falls 45
and small wedge failures,
the unstability situations are
more frequent than in
Table 4. FS results from Equation 3.
Slope 1-a
Factor of safety considering c

Joints φ = 45.5◦ φ = 34.8◦

Table 2. SMR (Romana) and SMR (‘‘in situ’’) values.
Quartzites 2.15–0.69 1.40–0.54
Slope section SMR (Romana) SMR (‘‘in situ’’)

Slope 1 - a 47 55
Slope 1 - b 34 45 Table 5. JRC results of the quartzites discontinuities.

JRC

Joints JRC maximum JRC minimum

The Slope 1 - a was classified as partially stable and
the Slope 1 - b was defined as unstable according to the Quartzites W2 14 9
SMR (Romana) system. The SMR obtained ‘‘in situ’’ Quartzites W3 11 7
for both sections were classified as partially stable.
The differences between SMR (‘‘in situ’’) and SMR
(Romana) are 7 and 8 positive points, confirming
that the SMR (Romana) system values can be supe-
The situations of unstability are associated with the
rior to the observed ‘‘in situ’’ and are conservatives
presence of water. The occurrence of important rock
for the time period considered. These differences can
slope unstability is generally connected with higher
be explained by the SMR values, as they were calcu-
rainfall periods. Considering these situations, the FS
lated for a longer period of time of the slope’s useful
values were also calculated by Equation 3.
life, whereas the situations of unstability analysed in
The failure model of Equation 3 approximately
this work were only observed during six years.
includes a perpendicular fracture to the landslide
The most problematic unstability sites were defined
surface, which is similarly verified ‘‘in situ’’.
through a kinematics analysis: the Markland test and
With the introduction of the influence of water in
its refinement by Hocking (1976); and using a geome-
the calculation of the FS, a clear reduction of these
chanical classification: the SMR (Romana) system.
values was verified, as can be observed in table 4,
The FS values for these sites were then determined by
where some FS results are inferior to 1. The results
Mohr-Coulomb and Barton & Bandis models. The FS
revealed that water action causes an increase of the
was established in order to evaluate the adequacy of
rock blocks unstability.
design.
The JRC values for the quartzites discontinuities of
In table 3 we can observe a great variety of FS results
slope 1 are presented in table 5.
obtained from Equation 1, the φ maximum and the φ
The JRC related FS values were obtained according
minimum obtained from the rock joint shear tests were
to the Barton & Bandis model (1990) (Equation 4):
respectively 45.5◦ and 34.8◦ .
The FS results showed higher values when the  
apparent cohesion was considered. When it was W cos ψ tan(JRC log10 (JCSt/σ  ) + φr )
n
assumed that the joints do not display any apparent FS = (4)
cohesion, an accentuated decrease of the FS (FS < 1) Wsenψ
was verified, with values that assumed immediate
failure. The FS results for the same range of Joint where JCS = joint compressive strength; σn =
Roughness Coefficient (JRC) values tend to diminish normal stress acting on a plane on which the shear
with the increase of the φ. strength is mobilized and φr = residual friction angle.

563
Table 6. FS results by Equations 4 and 5 for quartzites W2
0,6
discontinuities.
0,5
Factor of safety
0,4

FS
Joints JRC = 14 JRC = 9 0,3

Quartzites W2 0.45 0.33 0,2

0,1
5 7 9 11 13 15

Table 7. FS results by Equations 4 and 5 for quartzites W3 JRC

discontinuities.

Factor of safety Figure 2. Relation between the FS defined by the Equations

4 and 5 and the JRC values.
Joints JRC = 11 JRC = 7

Quartzites W3 0.37 0.29

0.5

0.4
The blocks slide occurs in the upper part of the rock FS
mass where normal stresses are very reduced, with
situations where (JCS/σn ) > 100. For these normal 0.3

stresses, the Equation 5 allowed a more conservative

FS determination based on the Barton & Bandis model
(1990). Authors such as Vallejo et al. (2002) consider 0.2
5 6 7 8 9 10 11 12 13 14 15
the Equation 5 very useful to avoid extremely high JRC
friction angles obtained from Equation 4 when low
normal stresses are used.
Figure 3. Relation between the FS defined by the Equations
φ = φr + 1.7JRC (5) 4 and 5 ( - With scale effect; - Without scale effect for a
joint profile of 0.4 m) and the JRC values.
The results of the FS based on Equations 4 and 5
were similar to those obtained using the Mohr-
Coulomb model when the values of the apparent In the present study, it was assumed that the sliding
cohesion were considered null. The results show a rock blocks have approximately 40 cm of length profile
direct proportional relationship between the FS and and the relation [Ln /L0 ] is equal to 4.
the JRC values, which translates the influence of the An adaptation of the Barton & Bandis criterion,
increase of roughness in the reduction of the verified considering the scale effect, is represented in the
displacements. The FS values correspond to situations Equation 8:
of unstability, with values lower than 1.
The JRC and JCS values tend to decrease with the τ = σ tan[JRCn log10 (JCSn /σn ) + φr + ir ] (8)
increase of the discontinuities extension.
To analyse the scale effect on the FS determination,
where ir = angle obtained through the large-scale
the Equations 6 and 7 defined by Bandis et al. (1981)
roughness profiles defined by Hack et al. (2003).
were used:
The ratio between Ln and L0 for the studied joints
was not very high. Consequently the FS values with
JRCn = JRC0 [Ln /L0 ]−0,02JRC0 (6)
scale correction were not very different from those
JCSn = JCS0 [Ln /L0 ]−0, 03JCS0 (7) obtained without scale correction. The Figure 3 shows
a distinction, not very accentuated, between the dif-
where JRCo = joint roughness coefficient for samples ferent results. When the scale effect was considered a
with approximately 10 cm of length profile; JCSo = small increase of the FS for the lower JRC values was
joint compressive strength for samples with approx- verified, while for the higher JRC values a reduction
imately 10 cm of length profile; Ln = joint length of the FS results was observed.
profile considered ‘‘in situ’’ and L0 = sample length In order to verify the scale effect on a discontinuity
profile (10 cm). surface of greater extension (approximately 2 meters),

564
 obtained through the Mohr-Coulomb model when the
apparent cohesion was considered null, to exclusively
0,5
establish the FS derived from the JRC values is not
recommended. The FS values must be confirmed
0,4 whenever it is possible, through the results of joint
FS

shear tests.
The FS values defined through the Equations 4
0,3
and 5 and the Mohr-Coulomb model when the apparent
cohesion was considered null seemed very conser-
0,2
vative. A more suitable approach was provided by
5 6 7 8 9 10 11 12 13 14 15 the Mohr-Coulomb model considering the presence
JRC of water: the FS results varied between 1.40 and 0.54
for the φ minimum.
Figure 4. Relation between the FS defined by the Equations
4 and 5 ( - With scale effect; - Without scale effect for a
REFERENCES
joint profile of 2.0 m) and the JRC values.
Bandis, S.C., Lumsden, A.C. & Barton, N.R. 1981. Experi-
mental studies of scale effects on the shear behaviour of
the FS with and without scale effect was determined rock joints. International Journal of Rock Mechanics and
wile the other parameters values remaining constant Mining Science and Geomech., 1 (18): 1–21.
(Figure 4). Barton, N.R. & Bandis, S.C. 1990. Review of predictive capa-
The results in Figure 4 showed a reduction of the FS bilities of JRC-JCS model in engineering pratice. Proc.
calculated with scale effect for the majority of the JRC International Conference on Rock Joints, Loen, Norway:
values. This reduction is more evident for the higher 603–610. Rotterdam: Balkema.
JRC values. Bieniawski, Z.T. 1979. The geomechanics classification in
rock engineering applications. Proc. 4th Congress Inter-
In most of the studied situations of unstability, it
national on Rock Mechanics, Montreux, Swiss: 51–58.
was verified a reduction of the tangential resistance Rotterdam: Balkema.
and FS values when the relation [Ln /Lo ] was increased Bieniawski, Z.T. 1989. Engineering rock mass classification.
and the scale effect influence was considered. Chichester: John Wiley & Sons.
Hack, R., Price, D. & Rengers, N. 2003. A new approach of
rock slope stability—a probability classification (SSPC).
5 CONCLUSIONS Bulletin of the International Association of Engineering
Geology and the Environment 62 (2):167–184.
The slope sections were defined from partially stable Hocking, G. 1976. A method for distinguishing between
single and double plane sliding of tetrahedral wedges.
to unstable according to the SMR (Romana) system.
International Journal of Rock Mechanics and Mining
The SMR obtained ‘‘in situ’’ was slightly different as Science and Geomech., 9 (13): 225–226.
both sections were considered as partially stable. Romana, M. 1985. New adjustment ratings for application
The SMR system must be applied with some limi- of Bieniawski classification to slopes. Int. Symposium on
tations and it can be used in rock slopes to define the the Role of Rock Mechanics, Zacatecas, México: 49–53.
main types of unstability that can occur as well as International Society of Rock Mechanics.
their degree of intensity. One of its advantages is to Romana, M. 1996. The SMR geomechanical classification
be a relative fast method which prevents the execution for slopes: A critical ten-years review. Proc. 8th Interna-
of an expensive and time consuming characterization. tional Conference and Field Trip on Landslides, Granada,
Spain: 255–267. Rotterdam: Balkema.
The SMR and its possible adaptations do not con-
Santarém Andrade, P.G.C. & Almeida Saraiva, A.L. 2000.
stitute the exclusive methods of rock slopes stability Physical and mechanical characterization of rock mate-
study. Other geomechanical classifications and more rial in excavation slopes. Proc. 7th National Geotechnics
accurate methods of analysis should be used. Congress, Porto, Portugal: 311–318. Lisbon: SPG. Porto:
Despite the similarity of the results, between the FEUP. (in Portuguese).
FS values obtained through Barton & Bandis (1990) Vallejo, G., Ferrer L., Ortunño, M., Oteo, C. 2002. Ingenieria
model (Equations 4 and 5) and Equation 5 and those Gelogica. Madrid: Prentice Hall.

565
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Contribution to the safety evaluation of slopes using long term observation

results

J. Barradas
Laboratório Nacional de Engenharia Civil, LNEC, Lisboa, Portugal

ABSTRACT: The existence of patterns of behaviour in some slopes, relating the accumulated precipitation
with movements along slip surfaces, is illustrated by the results of the long-term observation of two slopes. The
increments of sub-horizontal displacements, along deep slip surfaces, in excavated natural soil slopes, obtained
using inclinometers, along more than ten and twenty years, are directly correlated with the accumulated amounts
of rain by means of polynomial functions. These relations provide the basis for simple methods of evaluation of
the need for improvement of the stability conditions and of the performance of new stabilization works. If they
are associated with methods of climatic forecast, an evaluation of the medium and long term safety conditions
and forecasts of future behaviour can be done as well. These methods require the availability of medium or
long term good quality observation results, but they do not require permanent monitoring or even very frequent
measurements.

1 INTRODUCTION with displacements without any explicit function of

correlation (Picarelli & Russo 2004).
The evaluation of the behaviour and the stability This paper deals with cases of excavated natural
conditions of instrumented slopes after construction soil slopes in which there is no apparent progressive
are frequently difficult when long term movements acceleration of movements, that are instead irregular
are detected. In some cases, the same slip surfaces (of very different amount from one year to the other),
that existed during construction go on being active discontinuous and irreversible). It has been found that
during the lifetime of the infrastructure for which there are correlations between the increments of sub-
the slopes were created or modified. Such move- horizontal displacements, along deep slip surfaces,
ments are usually irregular and irreversible and they obtained using inclinometers, and the accumulated
are highly dependent on the intensity and distribu- amounts of rain in the vicinity of the slopes. These
tion of rainfall in the vicinity of the slope, particu- correlations express a kind of pattern that represents
larly if other causes such as new excavations, other the behaviour of the slope, so long as there are no
works, leakage of pipes, earthquakes, etc. do not changes in the stability conditions or other relevant
intervene. environmental actions (apart from the rainfall).
We have in these cases slow moving active land- The cases of two slopes are presented that show the
slides and many times the most economic solution widespread pattern of behaviour (obtained in a set of
is to live with the movements associated with the inclinometers), the change of pattern when a signifi-
slip surfaces. However this implies a strong capacity cant change of the stability conditions (improvement in
of prediction of future movements and trends which that instance) was achieved and the polynomial func-
usually is not available. tion that best represents the correlation. These slopes
In these cases, models of the same kind as have been monitored along more than ten and more
those used in the design of the geotechnical works, than twenty years.
although useful for the control of safety, are usu- Some applications of this method as well as its
ally not enough for the establishment of criteria to requirements and main limitations are briefly pre-
evaluate the behaviour of the works, in order to sented. These uses include a contribution to the safety
manage properly the infrastructure affected by the evaluation of the slopes and to the decision making
slope. involved in the management of the slope and the infras-
Studies presenting correlations of rainfall with tructure affected by it. From this method, some gains in
the behaviour of slopes deal more frequently with the efficiency of the execution of the slope monitoring
the occurrence of landslides or present comparisons can also be obtained.

567
2 SLOPE IN THE REGION OF LISBON been excavated when the motorway was inaugurated
(at the beginning of the 1960 decade).
2.1 Presentation In figure 1, a typical section of the slope is pre-
sented, with the geometry after the excavation, the
The enlargement of a motorway at the lower part of a
stabilization works executed until 1995 and the incli-
natural slope, 35 m in height, required the execution of
nometric tubes installed in that section (I3, I6, I7).
its excavation, in 1994–95. The same slope had already

Figure 1. Slope in the region of Lisbon. Cross section with stabilization works.

0.00 4
5 Ref. 14
Ref. 10 1995-03-30 21
1994-05-16 14 60.61 26 2.00
61.18 17 2.00 30
22 33
25 37
26 42
59.18 28 4.00 58.61 4.00
31 46
33 4 - 1995-05-22
5- 1994-06-16
57.18 10- 1994-07-14 6.00 56.61 14- 1996-05-20 6.00
14- 1994-08-16 21- 1997-06-11
17- 1994-09-13 26- 1998-06-03
55.18 22- 1994-10-18 8.00 54.61 30- 1999-05-10 8.00
25- 1994-11-24 33- 2000-05-05
Elevation (m)

26- 1994-12-09 37- 2001-05-28

Depth (m)

53.18 28- 1995-01-12 10.00 52.61 42- 2002-05-02 10.00

31- 1995-02-17 46- 2003-05-12
33- 1995-03-16
51.18 12.00 50.61 12.00

49.18 14.00 48.61 14.00

47.18 16.00 46.61 16.00

45.18 18.00 44.61 18.00

43.18 20.00 42.61 20.00

41.72 21.46 41.15 21.46

-1.50 0.00 1.50 3.00 4.50 -1.50 0.00 1.50 3.00 4.50
A2 Tube I3 A1 A2 Tube I3A A1

Figure 2. Profiles of sub-horizontal displacements in inclinometric tubes I3 and I3A.

568
2.2 Observation results and additional works system (construction of sub-horizontal linear drains,
drainage wells and vertical linear drains).
In figure 2, profiles of sub-horizontal displacements,
In the calculation of sub-horizontal displacements,
obtained in inclinometers I3 and I3A, in the direction
a bias error correction (Mikkelsen 2003) was executed.
closer to the steepest direction, are presented. The sec-
ond of these tubes replaced the first when this became
obstructed by excessive distortions.
The displacements in tube I3 were obtained during 2.3 Correlation between the observed behaviour,
the execution of a large part of the excavation and sta- the rainfall and the slope conditions
bilization works and those of tube I3A were measured Figures 4 and 5 show comparisons of the accumulated
after these works. quantities of rainfall in the rainiest three consecu-
Figure 2 shows that the same slip surface was active tive months of those years that were more rainy (of
during the excavation and afterwards. the periods 1995–2003 and 1998–2003, respectively)
Since the end of construction, in almost all the incli- with the increments of sub-horizontal displacements in
nometers, from the beginning of the rainy season of the rainy season measured in the inclinometers. The
the rainiest years, there were increments of the sub- results of precipitation published by the Meteorologi-
horizontal displacements. In most cases these corres- cal Institute, for a station in Lisbon (Gago Coutinho)
ponded to distortions along a slip surface located were used. In these figures, each rainy season is
inside a formation of grey clayey marls (Barradas identified by the year of its end.
2003) which had been already active during the
execution of the excavation (see fig. 1).
Figure 3 shows evolutions in time of the resultants of
sub-horizontal displacements measured in three pairs
of inclinometers (they are in pairs for the same reason
as inclinometers I3 and I3A) and of the load in a typ-
ical anchorage of those that are affected by the deep
ground movements. The similarity of the evolutions is
striking, except, in the case of the anchorage, where
some decrease in tension after each new increment can
be observed.
In 1996 and in 1997, a reinforcement of some
anchored works (beams and abutments) was executed,
with new anchors (in the first year) and (mostly in
1997) there was an improvement of the deep drainage
Figure 4. Accumulated rainfall in the rainiest three consec-
utive months vs. increments of sub-horizontal displacements
in the rainy season.

Figure 3. Inclinometric tubes I3, I3A, I5, I5A, I6, I6A and Figure 5. Accumulated rainfall in the rainiest three consec-
anchorage A37. Evolution of sub-horizontal displacements utive months vs. increments of sub-horizontal displacements
and anchor loads. in the rainy season.

569
 It can be seen that there is a very clear positive Lisbon), some landslides of considerable size occurred
correlation between the above mentioned accumu- when the toe of a natural slope, about 45 m in height,
lated quantities of rainfall and the referred increments was excavated.
of sub-horizontal displacements in all inclinometric The upper part of this slope includes colluvium
tubes presented, which are all that were installed in deposits reaching a considerable depth (20–25 m).
the main part of the slope. On the other hand, the con- They consist mainly of clayey silt and silty sand, with
sequences of the works implemented in 1997 seam some gravel. The weaker materials (clayey silt) had an
to have been remarkable, in reducing the detrimen- estimated ϕ  = 13◦ . Underneath and behind the col-
tal effects of the rainiest periods, according to the luvium there are several ‘‘in situ’’ layers of quaternary
correlations shown in figure 4. and cretacic age.
The results of figures 4 and 5 also suggest: In this case the stabilization works consisted in
the placement of rockfill masses and drainage works,
– a difference in the behaviour pattern between differ-
including wells and linear drains (Barradas 1996).
ent parts of the slope, since inclinometers I2, I6 and
The monitoring system included about 20 inclino-
I6A (which had a different behaviour in the first
metric tubes.
two years) and inclinometers 7A and 8A (which
had a different behaviour later) are in parts of the
slope (lateral and rear part, the two first and lower 3.2 Observation results
part the two last) that are different from the other
Figure 6 shows the evolution in time of sub-horizontal
inclinometers and they had a different response;
displacements, from 1987 to 2006, in one of the incli-
– the apparent non-linearity of the relation accumu-
nometers in which deep seated movements have been
lated rainfall/displacements specially in the case of
observed: tube 20A.
the upper part of the slope.
The deepest level of which results are presented in
figure 6 (11.5 m) is immediately above the slip surface
2.4 Evaluation of the safety conditions
intersection with tube 20A.
Figures 4 and 5 show the improvement of the behaviour In the calculation of sub-horizontal displacements,
pattern of the slope after the works executed in 1997. a bias error correction (Mikkelsen 2003) was executed.
However, due to:
– the irreversible and detrimental character of cer- 3.3 Correlation between the observed behaviour
tain aspects of the behaviour observed in the first and the rainfall
service years of the excavated slope (in which
Figure 7 shows the existing positive correlation bet-
permanent increments of displacement occurred,
ween the accumulated rainfall and the increments
including the upper part of the slope, where an
of sub-horizontal displacements in the rainy season
important water pipeline was buried, and affecting
measured in the inclinometer 20A.
a large number of the anchorages);
The accumulated rainfall (obtained from data
– the continuation of such behaviour, although atten-
of Coimbra meteorological station) is computed in
uated, after the additional works of 1996 and
two ways:
1997;
– and the state of overload that affected already many – quantities of precipitation occurred in the rainiest
anchorages of the slope (probably all of them in three consecutive months of those years that were
certain parts of the slope), more rainy;
the safety conditions of the slope were deemed insuffi-
cient, considering the risks arising from the proximity
of two important infra-structures (the motorway at the
base of the slope and the water pipeline at the top) and
it was decided to study a new solution for the slope
(Barradas 2005). This has already been implemented.
It included a substantial amount of earthwork that
changed the profile of the slope (reducing its height)
and the construction of a by-pass for the pipeline.

3 SLOPE IN THE REGION OF COIMBRA

3.1 Presentation
Figure 6. Evolution of sub-horizontal displacements since
In 1980, during the construction of a section of A1 November 1987 until October 2006, relative to the observa-
Motorway, in Coimbra region (200 km to the North of tion of 1987-11-19.

570
 It is also interesting to remark that, in those years
in which there were two periods of three consecu-
tive months that were rather rainy and in which the
accumulated rainfall values (for three months) did not
differ much from each other (the second value being
still above 300 mm), the respective data do not agree
so well with the general regression function. There
were several of these years before 1986; that is the
reason why this year was chosen for the beginning of
the period considered so far. Apparently there has been
a trend in the region, during the last decades, towards a
certain concentration of the rainiest period (when sev-
eral months are considered). This tendency improves
the applicability of the proposed method.
Although the same trend and pattern of behaviour
can be observed, in general, in the inclinometric tubes
Figure 7. Comparison of accumulated rainfall in the rainiest of this slope in which a slip surface was intersected
three consecutive months and in the period September–March and there are long term movements, the correlation
with the increments of sub-horizontal displacements in the between rainfall and displacements is not always as
rainy season, between 1986 and 2006, in tube 20A. good as in inclinometer 20A. This can be seen in
figure 8, in which a comparison is presented, with the
same criteria as above, but in this case for tubes 20A
– values accumulated from September of each year to and 17A, and for the period 1994–2006, and only for
March of the next year (seven months). the rainiest three consecutive months of those years

In figure 7, the quadratic regression functions Table 1. Comparison of accumulated rainfall with the incre-
for both set of data are presented, as well as the ments of sub-horizontal displacements in the rainy season,
correspondent value of R2 . between 1986 and 2006, in tube 20A. Values of R 2 for several
When accumulated rainfall of the three rainiest con- types of polynomial functions.
secutive months is considered a minimum threshold of
300 mm has been found, for movements in the slope Type of
to be detected. polynomial Third
No correlations similar or comparable to those that function Linear Quadratic degree
are presented in figure 7 can be obtained, in the
Rainiest three
case of this slope, comparing piezometric observation consecutive
(instead of accumulated rainfall) with displacements, months 0.92 0.95 0.98
given the fact that no frequent measurements of water Period
levels were generally available. On the other hand, September–March 0.82 0.86 0.88
even if they existed, they would not allow the same
kind of assessment of the behaviour of the slope, since
the evolution of the piezometric level is already part
of the response of the slope.
In figure 7, it can be seen that:

– The quadratic polynomial function used for the

regression expresses well the correlation when Tube 20A
(1994-2006)
accumulated rainfall of the rainiest three consec- Tube 17A
utive months is considered (R 2 = 0.95, for the (1994-2006)
Poly. (Tube 20A
period 1986–2006). (1994-2006))
– The same type of function expresses a worse corre- Poly. (Tube 17A
(1994-2006))
lation when the values of accumulated rainfall from
September to March of each year are considered
(R 2 = 0.86, for the period 1986–2006).

The results of R2 for several types of polynomial Figure 8. Comparison of accumulated rainfall in the raini-
functions that were used in the regression of the results est three consecutive months with the increments of sub-
of tube 20A, obtained between 1986 and 2006, are horizontal displacements in the rainy season, between 1994
presented in table 1. and 2006, in tube 20A and 17A.

571
that were rainiest. Tube 17A, for which the displace- its stability conditions. The patterns give an over-
ments measured immediately above the slip surface all information related with the real state of the
are used (depth = 12.5 m), is located in the southern slope, incorporating the influence of aspects that
part of the slope, in a cross section that is different exist usually only in a fragmentary and disperse
from the one of tube 20A (both tubes are about 100 m way (such as those related with the performance of
apart). The obtained results of R 2 (see fig. 8) were the drainage system and the hydro-geological con-
0.96 (for tube 20A) and 0.90 (for tube 17A). ditions) or may be even inexistent (such as the loss
The observation results of tube 17A, although of strength of the soil subject to progressive distor-
important for the part of the slope in which it is located, tions along the slip surfaces). The evaluations and
seam to be of a lesser quality than inclinometer 20A assessments that were mentioned will be of great
(as is also apparent comparing the displacement pro- interest in the decision making involved in the man-
files of tube 17A, obtained in successive observations, agement of the slope and the infrastructure affected
which show some irregularities in the part that is by it.
subjected to the largest distortions). 4. The quality of the observation results affects, as it
would be expected, the quality of the correlations
that can be established. For each slope, a choice of
3.4 Evaluation of the safety conditions the more reliable (and representative) monitoring
devices should be done, in order to make use of the
The near permanence of the relations between accu- proposed method.
mulated rainfall and displacements that are illustrated 5. Associating this method with statistical methods of
in figures 7 and 8 do not show a trend towards a wors- climatic forecast (distribution of rain) it will be pos-
ening of the global safety conditions in the period sible to evaluate the medium and long term safety
1986–2006. However they show that the pattern had conditions of the slope and to forecast its future
no trend to an attenuation of movements with time behaviour.
either. So the detrimental effects, such as continuing 6. From the use of this method, some gains in the
damages (although irregular and generally at a slow efficiency of the execution of the slope monitoring
rate) in infra-structures and houses were expected to can also be obtained.
continue. 7. This method requires the availability of medium or
The safety conditions were improved in 2006–2007 long term good quality observation results (includ-
when a new deep drainage system including wells and ing several years in which the accumulated quanti-
automatic pumping was installed. Giving the absence ties of rainfall were larger then average), but they
of anchorages, there is no problem of their overload, do not require permanent monitoring or even very
as was the case in the slope previously presented (in frequent measurements (it can be applied with only
section 2). three or four surveys per year).

4 CONCLUSIONS REFERENCES

From what has been presented, the following conclu- Barradas, J. 1996. Long term behaviour of a natural soil
sions can be drawn: slope excavated at the toe. In Proc. Seventh International
Symposium on Landslides, vol.3. Trondheim, Norway.
1. In cases of excavated natural slopes properly instru- Balkema.
mented and monitored for long periods, such as Barradas, J. 2003. Behaviour and stabilization of a natu-
ral slope excavated at the toe. In Field Measurements
those that were presented, it is possible to estab- in Geomechanics, Proceedings of the 6th International
lish behaviour patterns for the slopes, relating this Symposium FMGM 2003. Oslo, Norway. Balkema.
directly with the accumulated amounts of rainfall Barradas, J. 2005. Comportamento e estabilização da escav-
in the vicinity of the slopes. ação de um talude natural. In 2.as Jornadas Luso-
2. For this purpose, the better way that was found to Espanho-las de Geotecnia. Modelação e segurança em
consider the accumulated rainfall was the rainiest Geotecnia—LNEC. Lisbon, Portugal (in Portuguese).
three consecutive months and the functions used Mikkelsen, P.E. 2003. Advances in inclinometer data anal-
for regression that better express the correlation are ysis. In Field Measurements in Geomechanics, Proceed-
polynomial quadratic. ings of the 6th International Symposium FMGM 2003.
Oslo, Norway. Balkema.
3. The establishment of such patterns, accompanied Picarelli, L. & Russo, C. 2004. Remarks on the mechanics of
by adequate statistical criteria, allows an evalua- slow active landslides and the interaction with man-made
tion, in each new observation year, if there was an works. In Landslides: Evaluation and stabilization. Pro-
improvement or a worsening of the behaviour of ceedings of the 9th International Symposium on Land-
the slope and so, indirectly, also an assessment of slides. Rio de Janeiro, Brazil. Balkema.

572
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Delimitation of safety zones by finite element analysis

J. Bojorque, G. De Roeck & J. Maertens

Dept. of Civil Engineering, K.U. Leuven, Leuven, Belgium

ABSTRACT: This paper deals with the determination of safety zones and local minima by using finite element
analysis. Used is made of finite element method in order not to constraint the analysis by neither the assumptions
in the location of the sliding surface nor in the interslice force function. It is known, that only the most critical
failure mechanism and global minimum are evaluated by the strength reduction method, in such approach local
minima most of the time are unnoticed. Here, it is proposed that the safety zones and the local minima can be
detected by keeping the information generated in the strength reduction process. In addition, the importance
of soil properties in the location of the failure mechanism is highlighted. The methodology is presented in an
artificial case study and in a real natural slope. The safety zones should be considered in landslide stabilization
and remediation.

1 INTRODUCTION The modified shear strength reduction technique

in the framework of Finite Difference Method (Cala
Numerical methods have been recognized as a power- et al. 2004), has been proposed in order to detect
ful tool for practical geotechnical applications. These several sliding surfaces. The technique is based on
techniques have become important for slope stability gradually reducing the strength properties after iden-
analysis, especially when complex stratigraphy and tification of the first sliding surface. The modified
complex soil behaviour are treated. Moreover, the use technique needs extra computations and is not applica-
of the strength reduction method (SRM) to define ble for finite element analysis, causing some numerical
the stability of a slope has shown some advantages problems. In this paper, it is proposed that the strength
over traditional methods (Griffiths & Lane 1999, reduction technique used in the framework of finite
Dawson & Drescher 1999), among others. On the other element method can still given information regarding
hand, one limitation of the strength reduction approach different failure mechanisms by keeping the differ-
is that only the most critical failure mechanism and ent stages generated during the process. Furthermore,
global minimum are evaluated (Cala et al. 2004; Cheng the necessity of detect different failure mechanisms
et al. 2007). Therefore, by SRM, local minima most of that can arise from small changes in soil parameters is
the time are unnoticed. In engineering practice apart highlighted. Used is made of finite element methods
from the global factor of safety and the critical sliding in order not to constraint the analysis by neither the
surface associated with it, it is important to detect the assumptions in the location of the sliding surface nor
local minima. in the interslice force function. Moreover, the advan-
In classical limit equilibrium analysis, local min- tages of finite element slope stability analysis can be
ima are defined by evaluating different sliding surfaces exploited.
associated with different safety factors. A single line Safety zones and local minima are detected in
to characterize the sliding mechanism can mislead the finite element analysis by keeping all the shear zones
implementation of remedial measures. Safety maps defined by using different strength reduction coeffi-
can be generated by using limit equilibrium concepts cients (Bojorque et al. 2007). The safety zones should
to detect safety/unsafety areas. Those maps are repre- be considered for the selection and location of remedial
sented by a series of contour lines along which minimal measures for landslide stabilization. The verification
safety factors are constant (Baker & Leshchinsky of the procedure is presented in an artificial case study
2001, Renaud et al. 2003). However, limit equilib- acquired form literature in which the development of
rium methods need to use key assumptions in other to local minima is indicated. The methodology is imple-
solve the problem. These assumptions and the disre- mented, as well, in a case study of a natural slope
gard of the strain-stress relationship limit their use in located in Ecuador where the presence of different
some extension. failure mechanisms should be evaluated.

573
2 FINITE ELEMENT METHOD—STRENGTH will be higher than the factor of safety and other lower
REDUCTION TECHNIQUE than this value. By retaining and properly visualiza-
tion of the different steps, it is possible to detect other
In the framework of finite element slope stability failure surfaces that can emerge from the computa-
analysis, the Factor of Safety (FoS) is defined as clas- tions. The failure mechanism can be identified by
sical methods. Hence, the factor of safety of a slope the shear strain contour computed from the results
is defined as the factor by which the shear strength of the SRM. Different failure mechanisms associated
parameters must be divided in order to bring the slope with different reduced parameters can be detected and
to failure (Griffiths and Lane 1999). When the same incorporated in safety zones by locating different shear
strength reduction factor (SRF) is used for both, the zones (Bojorque et al. 2007).
cohesion (c) and tangent of the friction angle (tan φ), The finite element program PLAXIS using 15-node
Equation 1 holds. elements is used for the slope stability analysis based
on the strength reduction method (PLAXIS-BV 2004).
c tan φ The soils are modelled as elastic-plastic material
SRF = = (1)
cf (tan φ)f with Mohr-Coulomb failure criterion and zeroten-
sile strength. Vertical and horizontal displacements
where cf and (tan φ)f represent the factored restricted on the base and horizontal displacements
parameters. restricted on the sides are used as boundary conditions
To compute the factor of safety and its associated for the analyses.
failure mechanism, finite element method uses the
strength reduction method (SRM). In this process, the
cohesion and tan φ are gradually reduced or increased 3 NUMERICAL EXAMPLE
until non-convergence or convergence, respectively, in
the plastic solution is found. This reduction or increase A two slope angles example is presented to compute
will depend on both, the initial estimation of the the safety zones. The geometrical configuration and
strength reduction factor and the stability of the slope. soil properties are given in Figure 1. This example is
When small changes in the strength reduction factor taken from the benched slope case presented in (Cala
produce jumps from convergence to non-convergence et al. 2004). The lower slope is inclined 45◦ and the
in the solution, the factor of safety is determined. At upper one is inclined 40◦ .
this limit, Equation 1 can be re-written as,
3.1 Limit equilibrium safety map
SRF at failure = Factor of Safety
A comparative analysis is presented with respect to
available strength
= (2) limit equilibrium calculations in order to validate the
critical strength results. Bishop’s limit equilibrium equations using
circular slip surface is adopted. Figure 2 shows the crit-
It is noticed, based on Equation 2, that if the initial ical sliding surface determined by Bishop’s method.
estimation of the safety factor, represented by SRF, is Besides, the location of the 100 most unsafe slip
higher than the factor of safety, the SRM will reduce surfaces are given.
the soil strength parameters by increasing the esti- It is worth noticing that two well defined failure
mated SRF. Otherwise, if the initial estimation is lower mechanisms are presented. One located in the lower
than the factor of safety, the approach needs to increase slope, in which the critical slip surface is developed
the strength parameters.
Before performing the SRM, the initial state of
stress should be determined. For cases where the slope
is stable, FoS ≥ 1, the initial state of stress is normally
computed by the gravity loading procedure using the
slope own-weight. On the contrary, when the slope is
unstable, FoS < 1, the K0-procedure is adopted, with
a typical value of K0 = 1−sin φ. In this last approach,
the vertical stress (σv ) is determine by the weight of
the slope and the horizontal stress (σh ) is obtained
from the relationship K0 = (σh /σv ). Once the initial
stresses are computed, the determination of the fac-
tor of safety is obtained by the SRM. The SRM will
process different calculations for different strength Figure 1. Geometrical configuration and soil properties of
parameters (strength reduction factor), some of them the two-angle slope example.

574
and has a safety factor, associated with it, equal to 0.94. The darker zones are critical areas where the failure
The other mechanism is located in the upper slope with can occur. This map is generated by using 10, 000 cir-
a minimum safety factor corresponding to 0.99. The cular slip surfaces. From this map the determination of
first 100 sliding surfaces corresponded to factors of safety zones are enhanced by retaining and visualizing
safety from 0.94 to 1.06. A better representation of the all slip surface information.
limit equilibrium results can be given by a safety map
(Baker and Leshchinsky 2001; Renaud et al. 2003).
3.2 Finite element results
The safety map is constructed by dividing the slope
model by a mesh and assigning to each point in a mesh For the two slope angles example the FoS determine
a factor of safety obtained by minimizing the factor by FE-SRM is equal to 0.87. This value was computed
of safety between all the slip surfaces going through by performing 60 steps in the SRM. It is noticed that
this point (Renaud et al. 2003). For this example the after the fourth step, the critical factor of safety has
slope is divided into a rectangular mesh instead of a been reached, the next computations from 5 to 60, are
triangular mesh. For each point in the grid the mini- performed to check the stability of the SRM. From step
mum factor of safety from the nearest slip surface is 4 to the last step 60, the plastic points are constant and
input. Any limit equilibrium method can be used for are located at the lower slope (Fig. 4b). In stage 3, when
the computation of the sliding surfaces, for this exam- the strength reduction factor is equal to 0.93, two fail-
ple Bishop’s method with circular slip surfaces is used. ure mechanisms are detected (Fig. 4a). If only the last
A rectangular mesh spacing of 0.15 m is adopted for result is considered, as it is typically the case, the sec-
the discretization. A filtering value of 1.4 in the safety ond mechanism at the upper slope is unnoticed. This
factor is used to enhance the visualization of the safety can mislead the implementation of remedial measures.
map. Figure 3 shows the safety map for the two slope Therefore, the slope stability performance for others
angles example and the critical slip surface is repre- strength reduction factors is needed.
sented by the white hashed line which correspond a In PLAXIS, the initial SRF is computed automati-
factor of safety equal to 0.94. cally, this can be a drawback if others values are needed

(a)
Figure 2. Critical and first 100 most unsafe slip surfaces
generated by Bishop’s limit equilibrium method. Factor of
safety for the critical slip surface equal to 0.94.

(b)

Figure 4. Plastic and tension points generated during the

Figure 3. Safety map generated from Bishop’s limit equi- strength reduction method. (a) Stage 3, strength reduction
librium method. Critical slip surface (white hashed line) factor equal to 0.93. (b) Stage 4, strength reduction factor
correspond to a factor of safety equal to 0.94. equal to 0.88.

575
 40 degrees. Three main soil units are identified; collu-
vium, weathered tuff material (Tuff) and the volcanic
rock basement. The stratigraphy and cross-section
configuration are shown in Figure 6. Soil mechani-
cal parameters are given in Table 1, in which, γ is the
unit weight, c is the cohesion, φ is the friction angle,
E represents the Young’s modulus, and ν is the Pois-
son’s ratio. For all the analyses zero-tensile strength is
used. Soil parameters were obtained from performing
in situ and laboratory tests.
For the colluvium stratum, in which the potential
Figure 5. Shear strain contour generated during the strength landslide is situated, the cohesion has a value between
reduction method at Stage 4, strength reduction factor equal 9 to 15 kPa and the friction angle varies from 23 to 29◦ .
to 0.88. Even tough, the relative small variation in soil param-
eters, this changes the location of the failure surface.
Figure 7a shows the shear strain contour for the slope
to be check. For this example, the initial SRF value is having a c = 9 kPa and φ = 29◦ , for those param-
1.03, for step 2 the SRF is reduced to 0.98, and in eters the FoS is equal to 1.27 and the failure surface
stage 3 to 0.93. For the example, these SRF values are indicated by the strain contour is located at the upper
adequate to detect different potential failures. part of the colluvium. For the same geometrical con-
When performing a FE-SRM, it is necessary to figuration and by changing the c to 10 kPa and φ to
retain all the different strength reduction computa- 26◦ , a second failure mechanism arises, this second
tions. This information enables to detect potential failure is deeper and it is located at the lower part of
unstable mechanisms. By exporting the strains gen- the stratum (Fig. 7b). For this case, the computed FoS
erated at the Gauss points and drawing the shear strain is equal to 1.19. If now, the c = 15 kPa and φ = 23◦ ,
contour for different limits, others failure mechanisms the upper failure disappears and only the lower fail-
can be detected. Figure 5 shows the shear strain con- ure is detected (Fig. 7c), having a FoS equal to 1.14.
tour computed for stage 4, from this stage and on,
plastic points are not more generated at the upper slope
(Fig. 4b). By the properly vizualization the location of
the two failure zones and the potential unstable zones
are located. By this process it is plausible to detect
different safety zones in which adequate stabilization
measures can be implemented.
Good correspondence between limit equilibrium
results (Figs. 2–3) and SRM (Fig. 5) are produced. The
safety map generated by Bishop’s limit equilibrium
method and the safety zones determined by the shear
strain contour shows the same failure mechanisms.
For simple slope problems, the determination of
safety zones, by shear zones or safety map, is less
critical since the critical sliding mechanism and local
minima are near to each other. Any slope stabilization
Figure 6. Stratigraphy and geometrical configuration for
action done for the most critical failure mechanism the potential instability at Cuenca-Machala motorway (km
will affect the other mechanisms. The local minima 27+000), Ecuador.
will fall inside the shear strain contour and plastic
points developed by FE-SRM. Therefore, it facilitates
the slope stability analysis. Table 1. Geomechanical properties for the different
materials.

Parameter Material
4 INFLUENCE OF SOIL PROPERTIES symbol Units Colluvium tuff Rock

In this second example, the importance of consider- γ kN/m3 16.7 17.6 24.5
ing different soil parameters in the failure location c kPa 9–15 60.0 200
is shown. This example is a slope located at km φ degrees 23–29 33.5 40.0
E MPa 60 100 300
27+000 at the left hand side of the road Cuenca-
ν [-] 0.25 0.25 0.40
Machala in Ecuador. It has an inclination of about

576
 (a)

Figure 8. Delimitation of different failure modes depending

in soil properties combinations.

is equal to FoS1∗ F, and the failure mechanisms is the

same. In which, F is any scalar number. This definition
is valid for cases where no external load or water forces
are presented. Taking this relationship into account, a
plot can be draw indicating some of the combination
of the soil parameters (c and tan φ) that will have the
(b) same failure mechanism. Figure 8 shows the combi-
nation of cohesion, tan φ and φ, which gives the same
failure mechanism and where the factor of safety will
be ratio to each other.
The upper, mixed and lower lines represent the
factored parameters (c and tan φ). All the strength
parameter combinations that fall above the upper fail-
ure line will produce an upper failure. And all the
combinations that fall below the lower failure will
develop a lower failure. At the ‘‘unknown zone’’ (lack
of computations), if the parameters are above the
mixed failure line, the failure can be mixed or upper, if
the parameters fall below the later line the failure will
(c) be mixed or lower. The exact limits between the differ-
ent mechanisms are not yet defined, further research
Figure 7. Shear strain contours generated during the is still going on to define these limits and reduce the
strength reduction method. (a) upper failure, c = 9 kPa and unknown zone.
φ = 29◦ , factor of safety equal to 1.27. (b) mixed failure,
c = 10 kPa and φ = 26◦ , factor of safety equal to 1.19.
(c) lower failure, c = 15 kPa and φ = 23◦ , factor of safety
equal to 1.14. 5 CONCLUSIONS

This study presents a proposal to identify different

This example shows that for small different soil failure surfaces by retaining the information given in
strength parameters very different failure mechanisms the strength reduction method by different strength
can be developed. To construct safety zones the reduction factors. Finite element method is used in
location of the different failures should be delimitated. order not to constraint the analysis by the assumptions
It is worth noticing that by the principal of the employed by others methods regarding the failure. By
strength reduction method, all slopes which have the properly vizualization of the shear strain contour the
same strength parameters (c and tan φ) but are fac- safety zones are detected. The strength reduction fac-
tored by any scalar, will have the same failure mode. tor used to compute the factor of safety gives local
In other words, if a slope has parameters c1 , tan φ1 and minima for different computations. Good agreement
FoS1 , then, if another slope has parameters c2 = c1∗ F is encountered with the safety map generated by limit
and tan φ2 = tan φ1∗ F, then the factor of safety FoS2 equilibrium method.

577
 Small changes in soil properties can cause remark- Bojorque, J., G. De Roeck & J. Maertens. 2007. Comments
able changes in the failure mechanism, thus safety on two-dimensional slope stability analysis by limit equi-
zones should be evaluated by performing different librium and strength reduction methods by Cheng et al.
soil combinations. By appropriately manipulation of doi:10.1016/j.compgeo.2007.04.005 (in press).
the finite element slope stability information, the full Cala, M., J. Flisiak & A. Tajdus. 2004. Slope stability anal-
ysis with modified shear strength reduction technique.
extent of potential slope instability is depicted. This In W. Lacerda, M. Erlich, S. Fontoura, and A. Sayao
will contribute for the correct design and implemen- (Eds.), Proc. 9th intern. symp. on Landslides; Landslides:
tation of the stabilization measures. Evaluation and Stabilization, Rio de Janeiro, Brazil.
Cheng, Y.-M., T. Lansivaara & W.-B. Wei. 2007. Two-
dimensional slope stability analysis by limit equilibrium
ACKNOWLEDGEMENTS and strength reduction methods. Computers and Geotech-
nics 34(3), 137–150.
The first author would like to thank for the financial Dawson, E. & W. Drescher. 1999. Slope stability analysis by
support provided by the K.U. Leuven in the con- strength reduction. Géotechnique 49(6), 835–840.
Griffiths, D. & P. Lane. 1999. Slope stability analysis by
text of the Selective Bilateral Agreement between finite elements. Géotechnique 49(3), 387–403.
K.U. Leuven and Latin America. PLAXIS-BV. 2004. 2d-version 8, finite element code for
soil and rock analysis. Slope Stability Analysis, Delft
University of Technology and Plaxis, The Nederlands.
REFERENCES Renaud, J.-P., M. Anderson, P. Wilkinson, D. Lloyd &
D. Muir-Wood. 2003. The importance of visualization
Baker, R. & D. Leshchinsky. 2001. Spatial distribution of of results from slope stability analysis. Proceedings ICE,
safety factors. Journal of geotechnical and geoenviron- Geotechnical Engineering 156, 27–33.
mental engineering 127(2), 135–145.

578
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Laboratory and numerical modelling of the lateral spreading process

involving the Orvieto hill (Italy)

F. Bozzano, S. Martino & A. Prestininzi

‘‘Sapienza’’ University of Rome, Research Centre for Geological Risks (CERI—Valmontone), Rome, Italy

A. Bretschneider
‘‘Sapienza’’ University of Rome, Department of Earth Sciences, Rome, Italy

ABSTRACT: The results of a study carried out in the cliff rock of the medieval town of Orvieto (Umbria
region, Italy) are here presented. The Orvieto hill is involved in a lateral spreading process as a consequence
of its geological setting, due to the overlaying of a rigid 60 m thick tuff slab on plastic overconsolidated blue
clays. In order to reconstruct the stress path experienced by the Orvieto blue clays during their geological
evolution, a not standard laboratory test was performed, by a triaxial device. The obtained function was used to
perform a sequential FDM numerical modeling, whose findings are consistent with the actual site conditions.
These findings point out the significant contribution of both laboratory tests and numerical modeling in order
to reproduce the present stress-strain conditions of the hill involved in the lateral spreading process, and then to
better define the engineering-geology model of the Orvieto hill.

1 INTRODUCTION The Orvieto hill has been chosen for human set-
tlements beginning from the Bronze Age until the
Lateral spreading processes are widespread in many present, and then there are so many historical build-
sites of central Italy. The Orvieto town, located in the ings, churches and cultural heritage to be preserved.
southern Umbria region (Figure 1), is one of the most
famous examples of medieval towns involved in lateral
spreading process.
2 GEOLOGICAL AND GEOMORPHOLOGICAL
Lot of paintings and pictures testify that since Mid-
EVOLUTION
dle Ages the Orvieto hill was affected by falls on its
cliffs and complex landslides on its slopes; those types
The Orvieto hill is set up by an about 700 m large,
of instabilities, induced by lateral spreading processes,
1500 m long and ∼40 m thick tuff plate that overlies
affect the Orvieto hill till present time.
epiclastic de-posits and blue marine clays 15 m and
some hundred meters thick respectively (Figure 2).
The blue clays (Argille di base, Conversini et al.
1977, Felicioni, et al. 1995) were deposited during the
Middle Pliocene in the marine basin of the Media Valle
del Tevere (i.e. the middle Tiber valley) at a depth
of 100–200 m below sea level (Lembo Fazio et al.
1982). A regional uplift in Upper Pliocene—Lower
Pleistocene caused the end of the clay sedimenta-
tion and, in subaerial condition, the partial erosion
of the clay succession for a total thickness of 150 m
and the gently dipping of the clay top to north-east.
The blue clays are overlain by a fluviolacustrine epi-
clastic deposits about 15 m thick known as the Serie
dell’Albornoz. At the hill top there is the Tufo di
Orvieto tuff (Nappi et al. 1982), an ignimbritic flow
coming from the Vulsini volcanic complex about
Figure 1. Location of the Orvieto town. 336 ky ago. The Tufo di Orvieto tuff is characterised

579
Figure 2. Left, Geological sketch of the Orvieto hill: 1) recent and present alluvia; 2) lacustrine travertines; 3) Tufo di Orvieto
tuff; 4) Serie dell’Albornoz epiclastic deposits; 5) Argille di base clays of Orvieto; 6) cliff edges evolving by falls; 7) landslide
scarp; 8) debris; 9) structural surface; 10) complex landslide debris; 11) gully; 12) downcutting stream; 13) landslide terrace
edge; 14) section trace; 15) sampling point. Right, the A-A geological section, from the top to the bottom, in the four main
evolutionary steps of the Orvieto hill.

by two lithofacies: the first is lithoid with black 3 STRESS-STRAIN EVOLUTION

scoriae in a reddish matrix, the second one is a AND LABORATORY MODELLING
poorly cemented grey pozzolana. In the eastern part
of the hill outcrops a lens of spongy, well-stratified The geological evolution of the Orvieto hill can be
travertines. summarized in four main steps: 1) submarine depo-
During the Quaternary age, the linear erosion of the sition of the blue clays during the Middle Pliocene;
rivers caused the isolation of the Orvieto tuff plate from 2) emersion and erosion of the highest part of the
the Vulsini plateau, and now the Orvieto hill is shaped blue clays during the Lower Pleistocene; 3) deposi-
as a mesa-type relief. Considering the morphological tion of both fluviolacustrine sediments and the tuff
profile, the top of the hill is a quasi-planar surface during the Middle Pleistocene; 4) lateral release due
bounded by sub-vertical cliffs slopes with an average to the river valley erosion from the Middle Pleistocene
height of 40 m; at the base of the tuff plate the slope to the Holocene (Manfredini et al. 1980).
angle rapidly changes from sub-vertical to about 12◦ . The discontinuous geological processes responsible
The gravitational instabilities, both on the cliff walls for the geological setting and the evolutionary history
and on the hill slopes, produced in the past a thick of the Orvieto hill, induced, on the basal blue clay
debris belt all around the Orvieto hill. succession, a complex stress path characterized by
The main factor that causes the cliff instabilities is loading and unloading conditions. These conditions,
the lateral spreading process, due to the overlapping in addition to weathering, caused the geomechanical
of the tuff plate on the blue clays (Lembo Fazio et al. properties of the clay deposits to significantly decay,
1984, Manfredini et al. 1980); falls on the cliff edges as proved by the presence of a 15 m thick superficial
are a consequence of this process and of the absence layer of weathered and softened clays. In the previous
of the horizontal confining pressure due to the river studies many authors distinguished these three main
valley erosion. The instabilities of the clayey slope is, layers along the vertical profile of the clay deposit
instead, mainly due to the rainfall regime (Tommasi starting from ground level: weathered clay, softened
et al. 2006). clay, stiff clay (Conversini et al. 1977, Lembo Fazio

580
et al. 1982, Tommasi et al. 1996), about 10 m and 5 m as experimental conditions did not permit to simulta-
thick respectively along the slope of the Orvieto hill. neously change the σh and σv in the different sections
In order to reconstruct the stress path experienced of the simulated stress path, the same path was approx-
by the Orvieto blue clays during their geological evo- imated by a segmented straight line; on this line, the
lution, a not standard laboratory test was performed, σh were first changed, then the material was let to
by a triaxial de-vice. The stress-path experienced by consolidate and finally the σv were changed in a con-
the Orvieto clays has been planned considering the trolled way. Initially the specimen was isotropically
four main steps of the geo-logical evolution of the hill consolidated at a pressure of 8.6 × 105 Pa and brought
(Tables 1, 2). to a vertical load of 16 × 105 Pa simulating the clay
The test simulated the stress-strain evolution of a deposition (step 1 of Figure 3).
point in the clay slope near the base of the tuff plate, Subsequently, a vertical unloading process simu-
where some cubic clay samples were taken. The target lated the erosion with a vertical unloading until the
of the not standard triaxial test (NSTT), was to char- minimum vertical stress of 1 × 105 Pa (step 2 of
acterize the deform-ability of the blue clays in similar Figure 3). Then the emplacement of the epiclastic
stress-strain conditions experienced during their geo- deposits and of the ignimbritic flow was simulated
logical evolution. The stress path of the NSTT has been with a vertical reloading up to 1 × 106 Pa (step 3 of
planned starting from the values of vertical and hor- Figure 3). Simulation of the last section of the stress
izontal stresses (σv and σh in Table 1), in the sample path (lateral valley erosion, step 4 of Figure 3) was not
point at the four evolutionary steps of the geological feasible. This last step of the test, which was expected
evolution. to simulate zero horizontal load and constant verti-
Then a blue clay specimen was prepared in accor- cal load, would have required an apparatus capable of
dance with standard triaxial test procedures. The test directly measuring the horizontal deformations of the
was performed using a standard triaxial isotropic- specimen.
confinement test apparatus in CID configuration, The NSTT took six weeks for completing. The
using a loading rate of 0.002 mm/min. During the test, NSTT allowed, reproducing a laboratory modelling of
the stress conditions that the clay experienced during
its history, to derive the E-σ3 correlations, both for the
tangent modulus Eti and for the secant modulus Es ,
Table 1. Effective vertical stress calculated in the sample under the tested stress conditions: loading, unloading
point for each evolutionary steps. and reloading. In detail, on the related stress-strain
curves, a first roughly rectilinear segment was used to
Lithologies Thickness γn σv measure the initial tangent modulus (Eti ) and a sec-
m kN/m3 kN/m2 ond segment (comprised between the point where the
stress-strain curve becomes non-linear and the maxi-
step 1 clays 165 20 1686
mum or minimum vertical stress applied for each σ3
step 2 clays 10 20 105
step 3 tuffs 10 20 level) was used to determine the secant modulus (Es ).
epiclastic deposits 15 17 1005 A good linear correlation between the deformability
clays 50 13 and the confining pressure has been identified in the
step 4 tuffs 10 20 first loading step and in the unloading step; otherwise
epiclastic deposits 15 17 1005 in the reloading step no statistically valid correlation
clays 50 13

Table 2. Evolutionary steps of the Orvieto hill. σv =

effective vertical stress; σh = effective horizontal stress;
K0 = lateral stress ratio at rest; p = (σv + σh )/2; q =
(σv − σh )/2.

σv σh p q
Steps kN/m2 kN/m2 K0 kN/m2 kN/m2

I 1686 843 0.5 1264.5 421.5

II 105 63 0.6 84 21
III 1005 603 0.6 804 201
Figure 3. Assessed stress path for not standard triaxial test
IV 1005 0 502.5 502.5
(NSTT).

581
can be established, therefore, a single average value The stress-strain history of the Orvieto cliff was
was selected for the Eti and Es moduli. The obtained numerically modeled through engineering-geological
functions Eti Es vs. (σ3 ) and the values are summarised sections representing the above-described four evolu-
in Table 2. tionary steps as sketched in Figure 2 and in Figure 4;
for the step 4 was adopted the same unloading func-
tion E-σ3 obtained from the step 2 of the NSTT
(Tables 3a, 3b).
4 NUMERICAL MODELLING The results of the numerical modeling, consistent
with the actual site conditions, show in the step 4
In the second part of this study a numerical mod- (Figure 4-step 4) the presence of an about 30 m thick
eling of the Orvieto hill by the FDM code FLAC belt characterized by a G modulus decrease of about
5.0 (Itasca 2005) was performed. The target was to two thirds vs. the initial one. An about 15 m-thick por-
realize a sequential numerical analysis changing the tion of this belt corresponds to the Serie dell’Albornoz
engineering-geological section and the input data in deposits, while the remaining 15 m represent the shal-
each simulated evolutionary step. lowest part of the clays, softened by stress loading and
This was realized inputting in the code, for every reloading according to the data already reported by
step of the sequential numerical model, the geological Conversini et al. 1977, Lembo Fazio et al. 1982 and
section used to plan the NSTT, and the linked E-σ3 Tommasi et al. 1996. Sequential numerical modeling
function obtained from the NSTT (Tables 3a, 3b). In of the stress evolution of the investigated lithotypes
detail the engineering-geological model of the cliff revealed the existence of a softened portion of the
was defined by assigning the physical and mechan- clays which would have not been demonstrated by a
ical parameters obtained from laboratory tests and conventional simulation approach.
integrated with literature data to the investigated Furthermore was also simulated, in a parametrical
lithologies. way, the decay of the properties of the tuffs joints.
Moreover a vertical variability of the investigated The results (Figure 4) show that with the decreasing
parameters was assigned in the 2D numerical model; of the joints properties, new tension cracks can arise
as a con-sequence the parameters Eti and Es was sup- in an area extending as far as 140 m from the cliff
posed to change with depth within the same geotech- edge and that at the cliff edges, stresses cluster along
nical unit either discretely (for predetermined stress well-defined surfaces. These surfaces create a poten-
levels) or continuously (under the law of variation of tial wedge failure mode with concentration of tensile
the parameter) (Tables 3a, 3b). stresses.

Table 3a. Values of parameters used in the numerical simulations.

γn ten c t φ
Deposits kN/m3 Pa Pa Pa ◦ ν

Orvieto Tuffs∗ 13 4.90E+05 9.80E+05 35 0.2

Albornoz Series deposits∗∗ 17 5.00E+04 7.15E+04 35 0.3
Clays 20 4.00E+04 8.20E+04 26 0.3

γn = natural weight density; ten = tension cut-off; c = cohesion; t = resistance to tensile stress; φ = shear
resistance angle; n = Poisson ratio.

Table 3b. Values of parameters used in the numerical simulations.

σ3 Et(σ 3 ) Es(σ 3 )
Deposits kPa Pa Pa

Orvieto Tuffs∗ 2 ∗ 109

Albornoz Series deposits∗∗ 5 ∗ 108
Clays 0.55 ∗ 105 > 45.62(σ3 ) 8 (σ3 )
7.9 ∗ 105 + 2 ∗ 107 + 3 ∗ 106
1 ∗ 105 < 7.9 ∗ 105 132.22 (σ3 ) + 4 ∗ 107
1 ∗ 105 > 6 ∗ 105 6.48 ∗ 107 1.89 ∗ 107

σ3 = confining pressure range in tests; Et = tangent Young’s modulus; Es = secant Young’s

modulus.

582
Figure 4. Results of numerical stress-strain analyses of the Orvieto hill. Detail of north-western margin of the tuff plate on
the assumption of minimum joint strength, contour of shear stress on vertical plane and plasticity state.

its geological history since it summarizes the results

of both the laboratory and numerical modeling.
On the basis of the obtained engineering-geology
model (Figure 5) four different zones can be distin-
guished in the Orvieto hill.
The zone 1 corresponds to the tuffs at the hill-top;
the zone 2 corresponds to the tuff deposits involved in
the horizontal stress reduction as a consequence of the
lateral spreading process and according to the numeri-
cal modeling. The zone 3 corresponds to both the epi-
clastic deposits and the softened layer of the basal clays
as a consequence of the loading and unloading cycles
Figure 5. Engineering-geology model of the Orvieto hill: occurred during the geological evolution and accord-
1) undisturbed tuff plate, 2) released tuffs zone, resulting ing to the numerical modeling; at last, the zone 4 repre-
from the numerical model, 3) softened soils resulted from sents the stiff and undisturbed clay portion of the hill.
the numerical model, 4) stiff and undisturbed clay.

5 ENGINEERING-GEOLOGY MODEL
6 CONCLUSIONS
An engineering-geology model for the Orvieto hill was
obtained by merging the previously described geolog- An engineering-geology model of the Orvieto hill
ical model, laboratory model and numerical model. has been experimentally obtained by reproducing its
This engineering-geology model takes into account the stress-strain evolution via laboratory and numerical
stress variations due to the evolution of the hill during modeling. At this aim a not standard triaxial test

583
was performed and the obtained values for the stiff- Felicioni, G., Martini, E. & Ribaldi, C. 1995. Studio dei
ness parameters were considered as the input for the centri abitati instabili in Umbria. Articolo estratto dalla
sequential numerical modeling, per-formed by a FDM Pubblicazione n. 979 del GNDCI-CNR, U.O. 2.17 Rubet-
approach. The results of the numerical modeling allow tino editore.
to set up an engineering-geology model, that is not ITASCA 2005. FLAC 5.0: User manual. Licence number
213-039-0127–16143 (Sapienza—University of Rome,
only referred to the actual conditions of the site since Earth Science Department).
it takes into account all the geological evolution of Lembo Fazio, A., Manfredini, G., Sciotti, M. & Totani, G.
the hill. As a consequence, a geomechanical zoning, 1982. Caratteristiche geotecniche dell’ argilla di Orvieto.
based on the stress-strain evolution of the Orvieto hill, Quaderni dell’Istituto di Arte Mineraria Giugno 1982.
was obtained, showing the existence of a 15 m thick Lembo Fazio, A., Manfredini, M., Ribacchi, R. & Sciotti, M.
low-stiffness clayey level, just below the Orvieto tuff 1984. Slope failure and cliff instabilities in the Orvieto
plate, which well corresponds to the softened clays hill. In Proceedings of the 4th International Sym-posium
recognized by various Authors all along the Orvieto on Landslides, Toronto, Ont., 16–21 September 1984,
hill slopes. Canadian Geological Society, Rexel, Ont. Vol. 2:
115–120.
This study infers that a coupled laboratory and Nappi, G., Chiodi, M., Rossi, S. & Volponi, E.
numerical modeling approach, which takes into 1982. L’ignimbrite di Orvieto nel quadro dell’evoluzione
account the geological evolution of a specific site, has vulcanotettonica dei vulsini orientali. Caratteristiche
to be regarded as a fundamental tool for better refining geologiche e tecniche. Boll. Soc. Geol. It., 101: 327–342.
an engineering-geology model. Manfredini, G., Martinetti, S., Ribacchi, R. & Sciotti, M.
1980. Problemi di stabilità della Rupe di Orvieto. Asso-
ciazione Geotecnica Italiana— XIV Convegno Nazionale
di Geotecnica. Firenze 28–31 ottobre 1980; Volume I:
REFERENCES 231–246.
Tommasi, P., Pellegrini P., Boldini, D. & Ribacchi, R. 2006.
Barbero, M., Barla, G. & Demarie, G.V. 2004. Applicazione Influence of rainfall regime on hydraulic conditions and
del metodo degli elementi distinti alla dinamica di mezzi movements rates in the overconsolidated clayey slope of
discontinui. Rivista Italiana di Geotecnica 3 (2004): 9–24. the Orvieto hill (central Italy). Can. Geotech. J. 43(2006):
Barla, G., Borli Brunetto, M. & Vai, L. 1990. Un esempio 70–86.
di modellazione matematica in rocce tenere: la Rupe di Tommasi, P., Ribacchi, R. & Sciotti, M. 1996. Geotechnical
Orvieto. Mir ’90, III Ciclo di conferenze di meccanica ed aspects in the preservation of the historical town of Orvi-
ingegneria delle rocce: 14.1–14.12. Torino. eto. In Carlo Viggiani (ed.), Geotechnical Engineering for
Conversini, P., Lupi, S., Martini, E., Pialli, G. & Sabatini, the Preservation of Monuments and Historic Sites; Proc.
P. 1977. Rupe d’Orvieto Indagini geologico-tecniche. Int. Symp., Napoli, Italy, 3–4 October 1996. Rotterdam:
Quaderni della Regione Umbria suppl. al n. 15. Balkema, 849–858.

584
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Albano Lake coastal rock slide (Roma, Italy): Geological constraints

and numerical modelling

F. Bozzano, C. Esposito, S. Martino & P. Mazzanti

‘‘Sapienza’’ University of Rome, Department of Earth Sciences and ‘‘CERI’’ Research Center on Geological Risks,
Rome, Italy

G. Diano
Geologist

ABSTRACT: The aim of this paper is to back analyze the stress-strain conditions that led to an ancient massive
rock slope failure occurred in the inner slope of the Albano lake (Rome, Italy), whose scar and accumulation
areas are respectively subaerial and submerged. Detailed geological and geomorphological field surveys, DEM-
derived topographical reconstructions of the pre-landslide topography and geomechanical data derived from
both in situ and laboratory characterizations, allowed to obtain the engineering-geology model of the slope prior
to the landslide occurrence. A numerical modelling by the FDM code FLAC 5.0 was then performed in order to
analyze the possible landslide trigger; static, hydrostatic and pseudostatic conditions were taken into account.
Static and hydrostatic analyses indicate significant displacements, with the lake level close to the actual one;
if an horizontal acceleration higher than 0.02 g is applied, a failure occurs according to a rock-slide kinematic
mechanism, strongly influenced by the local geologic-structural setting.

1 INTRODUCTION topographic information on the area obtained through

a Digital Elevation Model derived from LiDAR data
The hereby presented case study is framed within a for the subaerial slope and through a Lacustrine Digital
wider research addressed to define the landslide haz- Elevation Model (LDEM) derived from a multi-beam
ard conditions of a peculiar morphologic context such sonar survey (Anzidei et al. 2007, Baiocchi et al. 2007)
as a lacustrine environment. As a matter of fact, the for the submerged one. This kind of data allowed us
assessment of landslide-related hazard and risk con- to reconstruct a geological cross section and a related
ditions of coastal areas has to take into account also engineering-geology section of the slope involved in
the ‘‘secondary’’ effects linked to the slope instabilities the recognized massive rock slope failure. The stress-
such as the possible generation of tsunami waves due to strain analysis on this section was then performed
the impact of a landsliding mass from a subaerial slope via numerical modelling by taking into account dif-
onto the water (Bozzano et al. 2006, Tinti et al. 2005). ferent conditions (static, hydrostatic and pseudostatic
In this context, the recognition of even past large-sized analyses) in order to identify and assess both the
landslides assumes a relevant role, since they can be predisposing and the triggering factors.
considered as geomorphic evidence of the potential
occurrence of this kind of processes. Particularly sig-
nificant is then the comprehension and reconstruction 2 GEOLOGIC AND GEOMORPHOLOGIC
of the stress-strain conditions that led to a massive rock SETTING
slope failure in order to assess the potential of similar
landslides to occur in the present boundary conditions. The study area is located along the slope which bounds
For these purposes an effort has to be done in order in the southern part the Albano lake. The latter cor-
to collect the most significant data for the reconstruc- responds to the inner slope of a large multiple maar
tion of an affordable engineering-geology model of the partially filled with water (Fig. 1). This maar is in its
slope before the landslide occurrence. In the presented turn framed within the Colli Albani volcanic edifice,
case the traditional studies (such as geological and whose origin can be related to back-arc extensional
geomorphological surveys and geomechanical charac- processes which affected the Thyrrenian coast of cen-
terizations) were coupled with the analysis of detailed tral Italy (Funiciello et al. 2003, Giordano et al. 2006).

585
 phenomena, geomorphic evidence of at least one past,
large landslide is present in the southern slope of the
maar.

3 THE ALBANO ROCKSLIDE

As above mentioned one of the most significant

gravity-induced landform was identified along the
southern sector of the slope which bounds the
lake, where a wide, markedly concave slope sector,
whose shape is ascribable to the scar area of a large
landslide is present (Fig. 2). In addition, the obser-
vation of the LDEM allowed us to recognize a huge,
convex positive landform just downslope the previ-
ously mentioned scar area (Fig. 2). This landform
Figure 1. Virtual aerial oblique view of the Albano lake
(from the website http://earth.google.it).

The Albano maar activity is linked to the final

hydromagmatic phase of the Colli Albani volcanic
complex; the corresponding crater is actually the result
of the coalescence of at least three recognizable craters
which mark different stages of activity. As a result, the
geological setting of the inner slopes of the Albano
lake, corresponding with the internal walls of the
multiple maar, is mainly featured by outcrops of hydro-
magmatic deposits, which locally overlay lava and
scoria deposits related to the previous volcanic phases
of the Colli Albani complex. The former are fea-
tured by two typical lithofacies (Giordano et al. 2002).
The first one represents the most part of the Albano
maar deposits and is a plane parallel to low-angle Figure 2. In this figure a photograph and a 3d rendering are
cross-stratified alternation of scoria lapilli beds and combined to give a view of the whole subaerial-submerged
ash-rich layers, generally cemented for the zeolitisa- southern slope of the lake, affected by the slope failure.
tion. The second lithofacies is represented by massive
and chaotic, ash-matrix supported, up to 30 m thick
density current deposits, containing up to block-sized
xenoliths. In addition, an intracrater facies shaped as a
series of tongues that locally drape the internal walls of
the Albano maar from the crater rim down to the lake
level outcrops with a marked dipslope attitude (De Rita
et al. 1986). The subaerial geological frame is com-
pleted by talus slope and shore deposits that cover the
lowest parts of the subaerial inner slopes.
Many gravity-induced landforms are present along
the slopes that bound the lake; based on the geomor-
phologic surveys performed on purpose, active rock
falls (and/or topple) and debris flows with volumes
ranging between 10−1 and 103 m3 are the most diffuse
processes. In addition, the availability of the LDEM
allowed us to observe a large number of subaerial land-
slide detachment areas whose deposit is located below
the lake level. Even if the most diffuse and frequent Figure 3. DEM and topographic profile of the investigated
landslides consist of mainly small and medium sized slope adopted for the landslide volume assessment.

586
Figure 4. a) Geological cross section of the investigated slope; b) Engineering-geology model adopted for the numerical
modeling, based on the hypothesized topography of the slope prior to the failure.

Table 1. Main parameters of the intact rock.

Zone III Zone III

Zone I Zone II (lava) (scoria)
Table 2. Main parameters of the rock mass (discontinu-
υ 0.25 0.25 0.25 – ities plus intact rock) derived from the equivalent continuum
densi (kg/m3 ) 1956 1956 2873 – approach.
φi 28◦ 28◦ 35◦ –
σti (Pa) 1.38 * 106 1.38 * 106 9.58 * 106 – Zone I Zone II Zone III
ci (Pa) 2.76 * 106 2.76 * 106 1.92 * 106 –
Ei (Pa) 1.07 * 1010 1.07 * 1010 8.00 * 1010 1.10 * 109 densj (kg/m3 ) 2358 2358 2703
stdv (Ei ) – – – 4.29 * 108 Ej (Pa) 9.27 * 109 9.27 * 109 7.08 * 1010
CV – – – 0.39 φj 4◦ 4◦ 10◦
σ tj (Pa) 1.09 * 105 1.09 * 105 8.68 * 104
∗ ν is the Poisson ratio, φ is the friction angle, σ t is the tension
cj (Pa) 7.48 * 106 7.48 * 106 4.03 * 107
cut-off, c is the cohesion, E is the Young’s modulus, CV is the
coefficient of variation; the subscript i indicates a parameter ∗ See the caption of Table 1 for the symbols; the subscript j
evaluated for the intact rock. indicates a parameter evaluated for the jointed rock mass.

587
Table 3. Main parameters of the considered joint systems. to better define the above mentioned Albano rock-
slide, the most relevant evidence of a large-sized event
Zone I Zone II Zone III occurred in the study area and then the only constraint
to depict a possible scenario in case of a large landslide
φj 33◦ 33◦ –
occurrence. The mechanism of the rock-slide was sim-
Ij 0◦ ; 17◦ ; 34◦ – –
cj 0 0 – ulated by a stress-strain numerical modelling. A back
RMR 42–44 42–44 52–57 analysis was performed in order to analyse the role of
Q 658 658 317 the geological setting, of the rock mass properties as
well as of the external forces on the rock-slide trigger.
∗ f and c are respectively the friction angle and the cohesion
A 2D numerical model by the FDM code FLAC 5.0
along the joints, I is the inclination of the joints, RMR and was obtained starting from the topographical section
Q are respectively the values of the Bieniawski (1988) and obtained via LiDAR and multibeam surveys (Baiocchi
Barton (1988) classifications. et al. 2007), respectively for the subaerial and for
the submerged slope. A pre-failure slope shape and
geological setting were hypothesized on the basis of
can be interpreted as the corresponding debris accu- the morphological features adjacent to the detachment
mulation deriving from the slope failure occurred in area as well as to the depositional mechanism assumed
the present subaerial slope. An indirect confirmation for the volcanic flow-type deposits. In order to build
of the relationship between the landforms previously the engineering-geology section of the slope, three
described can be found also in the calculation of different ‘‘geomechanical’’ zones were distinguished
their volumes. The estimation of the detached rock in the above mentioned section (Fig.4b) by consid-
volume was performed by means of DEM-derived ering the geology features and the rock mass (intact
topographical reconstructions of the hypothesized pre- rock plus discontinuities) properties derived from in
event morphology and gave as result a volume of about situ surveys, laboratory tests and bibliography data
3 * 106 m3 (Fig. 3). This volume is quite well compara- (Tabs. 1–3).
ble with the volume of about 2.5 * 106 , estimated for A first zone (Zone I) corresponds to the scoria
the hypothesized debris accumulation performed on deposits including lava levels; a second zone (Zone
the DEM of the submerged part, especially if some evi- II) corresponds to the hydromagmatic deposits with
dence of successive landslides that involved the debris a sub-horizontal attitude; a third zone (Zone III) cor-
material are taken into account. responds to the hydromagmatic deposits, showing a
The geological setting of the subaerial part of the slope-dipping attitude. Taking into account both geo-
hereby discussed slope sector is featured by the super- logical properties and geomechanical behaviour of
imposition of hydromagmatic deposits upon a thick the different deposits the following constitutive model
bank of lava lenses and scoria deposits. In particu- were applied to the three zones:
lar, detailed geological surveys pointed out that the
material actually involved in the rock slope failure Zone I: elastic constitutive model;
is constituted by massive and chaotic, ignimbrite Zone II: Mohr Columb elastic-perfectly plastic
deposits belonging to one of the previously mentioned constitutive model;
intracrater facies of the hydromagmatic deposits, char- Zone III: Mohr Coulomb elastic-perfectly plastic
acterized by a marked dip slope attitude of the consti- constitutive model with ubiquitous joints.
tutive layers, quite parallel to the slope angle of the The values of the corresponding parameters were
topography (Fig. 4a). Based on the morphologic evi- evaluated on the basis of a conceptual model of
dence (such as the straight and sharp shape of both the the slope based on field evidences, by consider-
crown and the flanks of the landslide scar, the flat mor- ing heterogeneities, anisotropies and/or weak planes
phology of the topmost part of the debris accumulation within the rock masses. Moreover, the kinematic
and the abrupt slope angle reduction in the lower part of role of the rock mass discontinuities were also taken
the scar area) and on the geological-structural setting, into account. In particular, the following rock mass
this landslide can be preliminarily classified as a com- features were attributed to the simulated geological
pound translational rockslide—structurally controlled units:
of type E (block slide with toe breakout), according to
Hungr and Evans 2004. Zone I: heterogeneous (scoria and lava levels) and
isotropic material, for the scoria deposits a Gauss-
normal distribution of the mechanical parameter
values was applied;
3.1 Numerical modelling
Zone II: homogeneous and isotropic material (no
On the basis of the collected geological-geomorpholo- possible effects of the actual discontinuities on the rock
gical data, another phase of activity was addressed slide);

588
Figure 5. Results of the numerical modeling under pseudostatic conditions: plasticity indicators show the envelope of the
failing rock mass that fits well with the actual observed features of the scar area.

589
Zone III: homogeneous and anisotropic material shows sliding along joints within the rock mass,
(possible effects of the actual discontinuities on the without failures of the intact rock, thus highligh-
rock slide). ting the relevant kinematic role of the slope-dipping
anisotropies and thus the relevant constraint due to the
For both Zone II and III an equivalent continuum local geological setting of the slope; ii) critical role
approach (Sitharam et al., 2001) was adopted in order of the lake level in predisposing the landslide phe-
to quantify the geomechanical parameter of the rock nomenon; iii) possible role of the seismic input in the
mass; the obtained values were derived from both landslide trigger.
intact rock and joints properties (Esposito et al., 2007). Moreover, a back-analysis of the post-failure
The iso-oriented discontinuities within the rock mass dynamic was performed for this landslide via the
were considered as weak planes for the anisotropic DAN-W software (Hungr 1995), by using a simple
constitutive model. basal frictional rheology and considering the sliding
The back analysis of the rock-slide was performed mass as a rigid block. In spite of the simple model
under three different conditions: i) static analysis: only applied the values of basal friction (about 25◦ ) and
the weight force of the slope rock mass was applied; reached maximum velocity (about 15 m/s) are compa-
ii) hydrostatic analysis: the weight force of both the rable with the available literature data for this kind of
slope rock mass and the lake water column were slope failure.
applied; iii) pseudostatic analysis: equivalent horizon- The above mentioned results can be considered
tal accelerations due to the expected earthquakes were the basis on which successive studies, such as the
applied. In particular, the hydrostatic analysis was numerical modelling of the landslide-generated wave
performed by supposing different levels of the lake, propagation, should be developed in order to better
close to the actual one, in the range 323–268 m a.s.l.; define the possible scenarios deriving from the impact
the pseudostatic analysis was performed by applying of the landslide mass on the lake and the related hazard
different horizontal incremental accelerations up to and risk conditions for the nearshore villages.
0.175 g, in agreement with the maximum values for
the statistical recurrence time of 475 years (INGV,
2005). The obtained results point out that equilibrium
is reached in all the static and hydrostatic conditions; REFERENCES
nevertheless, in the hydrostatic conditions, the anal-
ysis of both the displacement field and the plasticity Anzidei, M., Esposito, A. & De Giosa, F. 2007. The dark
state within the slope points out significant displace- side of the Albano crater lake. Annals of Geophysics, 49:
ments, mainly in the detachment area of the rock-slide, 1275–1287.
if the lake level is assumed to be very close to the actual Baiocchi, V., Anzidei, M., Esposito, A., Fabiani, U.,
one. If an horizontal acceleration higher than 0.02 g Pietrantonio, G. & Riguzzi, F. 2007. Integrer bathymetrie
is applied, a sliding failure occur within the slope- et lidar. Geomatique Expert, 55: 32–35.
dipping phreatomagmatic deposits at an average depth Barton, N. 1988. Rock Mass Classsification and Tunnel
of about 30 m b.g.l. Reinforcement Selection Using the Q-System. American
Society of Testing and Materials, Philadelphia, 59–88.
The crown area is marked by tension cracks while Bieniawski, Z.T. 1988. The Rock Mass Rating (RMR) Sys-
the break out zone closely corresponds to the morpho- tem (geomechanics classification) in engineering pratice.
logical plain which can be actually observed at the In: Louis Kirkaldie, (Ed.), Rock Classification Systems
bottom of the subaerial slope. As a consequence, for Engineering Purposes, pp 17–34.
the resulting landslide mechanism is very consis- Bozzano, F., Chiocci, F.L., Mazzanti, P., Bosman, C.,
tent with the observed rock-slide scar; moreover, the Casalbore, D., Giuliani, R., Martino, S., Prestininzi, A. &
obtained plasticity state shows sliding along joints Scarascia Mugnozza, G. 2006. Subaerial and subma-
within the rock mass, without failures of the intact rine characterisation of the landslide responsible for the
rock. The latter evidence justifies the kinematic role of 1783 Scilla tsunami. EGU 2006, Geophysical Research
Abstracts 8, 10422.
the slope-dipping anisotropies within the hydromag- De Rita, D., Funiciello, R. & Pantosti, D. 1986. Dynamics
matic deposits (Fig. 5). and evolution of the Albano crater, south of Rome. Proc.
IAVCEI Int. Conf., Kagoshima: 502–505.
Esposito, C., Martino, S. & Scarascia Mugnozza, G. 2007.
4 CONCLUDING REMARKS Mountain slope deformations along thrust fronts in jointed
limestone: an equivalent continuum modelling approach.
Geomorphology, 90: 55–72.
The main findings from the numerical back analysis of Freda, C., Gaeta, M., Karner, D.B., Marra, F., Renne, P.R.,
the Alban rock-slide can be summarised in the follo- Taddeucci, J., Scarlato, P., Christensen, J.N. & Dallai, L.
wing points: i) significant constraints due to the geo- 2006. Eruptive history and petrologic evolution of the
logical setting and, in particular, to the anisotropic Albano multiple maar (Alban Hills, Central Italy). Bull.
conditions of the rock mass, since the plasticity state Volcanol. 68: 567–591.

590
Funiciello, R., Giordano, G. & De Rita, D. 2003. The Albano Hungr, O. & Evans, S.G. 2004. The occurrence and clas-
maar lake (Colli Albani Volcano, Italy): recent volcanic sification of massive rock slope failure. Felsbau 22:
activity and evidence of pre-Roman Age catastrophic lahar 16–23.
events. Journal of Volcanology and Geothermal Research, INGV 2005. Carta dei valori di pericolosità sismica del terri-
123: 43–61. torio nazionale (OPCM 28/04/06 n. 3519), http://zonesis
Giordano, G., De Rita, D., Cas, R. & Rodani, S. 2002. Valley miche.mi.ingv.it/mappa_ps_apr04/griglia002/lazio.html.
pond and ignimbrite veneer deposits in the small-volume Itasca 2005. FLAC, Fast Lagrangian Analysis of Continua,
phreatomagmatic ‘Peperino Albano’ basic ignimbrite, Version 5.0. Itasca Consulting Group, license: DST—‘‘La
Lago Albano maar, Colli Albani volcano, Italy: infuence Sapienza’’, Roma—serial number: 213–039–0127–16143.
of topography. Journal of Volcanology and Geothermal Sitharam, T.G., Sridevi, J., Shimizu, N. 2001. Practical equiv-
Research, 118: 131–144. alent continuum characterization of jointed rock masses.
Giordano, G., De Benedetti, A.A., Diana, A., Diano, G., Int. J. of Rock Mech. and Min. Sciences 38: 437–448.
Gaudioso, F., Marasco, F., Miceli, M., Mollo, S., Cas, Tinti, S., Manucci, A., Pagnoni, G., Armigliato, A. &
R.A.F. & Funiciello, R. 2006. The Colli Albani mafic Zaniboni, F., 2005. The 30 December 2002 landslide-
caldera (Roma, Italy): Stratigraphy, structure and petrol- induced tsunamis in Stromboli: sequence of the events
ogy. Journal of Volcanology and Geothermal Research, reconstructed from the eyewitness accounts. Natural
155: 49–80. Hazards and Earth System Sciences, 5: 763–775.
Hungr, O. 1995. A model for the runout analysis of rapid
flow slides, Debris flows, and avalanches. Can. Geotech.
J., 32: 610–623

591
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Superposition principle for stability analysis of reinforced slopes

and its FE validation

F. Cai & K. Ugai

Department of Civil and Environmental Engineering, Gunma University, Kiryu, Japan

ABSTRACT: This paper proposes to use principle of superposition for stability analysis of slopes reinforced
with anchors, geosynthetics, or nails. The principle of superposition is implicitly used in Fellenius’ method. This
paper proposes to use the principle of superposition for Bishop’s simplified method. The safety factor calculated
using the principle of superposition of anchor-reinforced slopes was compared with the results calculated using
finite element method with shear reduction technique (SSR-FEM). The comparison shows that the safety factor
calculated using the principle of superposition for stability analysis of anchor-reinforced slopes was consistent
well with that calculated with SSR-FEM, a different approach. The principle of superposition can be easily
extended to other slope stability analysis methods.

1 INTRODUCTION described in detail in the following section, the effect

of the axial tension on the safety factor is incorrectly
The end type of anchorage, where the tendon is grouted considered in some design tools widely used in prac-
below the potential slip surface, has been used to tice (e.g. Geo-Slope International Ltd. 2000, Oasys
stabilize dangerous slopes to a specified safety fac- Ltd. 2001).
tor because of its significant technical advantages This paper proposes to use principle of superposi-
resulting in substantial cost savings and reduced con- tion for stability analysis of anchor-reinforced slopes.
struction period. The basic design concept is transfer- The principle of superposition is implicitly used in
ring the resisting tensile forces generated in anchors Fellenius’ method. This paper proposes to use the prin-
into the ground through the friction mobilized at ciple of superposition for Bishop’s simplified method.
their interfaces. The loads are usually developed by The safety factor calculated using the principle of
the anchorage of the tendon within the soil mass superposition of anchor-reinforced slopes was com-
and tensioning at the surface against a bearing plate, pared with the results calculated using finite element
concrete pad, or tensioned geosynthetics (Koerner method with shear reduction technique (SSR-FEM).
1984, Greenwood 1985). Some successful applica- The comparison shows that the safety factor calculated
tions of this technique to stabilizing slopes have using the principle of superposition for stability anal-
been reported (Tan et al. 1985, Hashimoto 1986, ysis of anchor-reinforced slopes was consistent well
Corona 1996). with that of SSR-FEM, a different approach.
Limit equilibrium method, such as Bishop’s sim-
plified method, has been commonly used to assess
the safety factor of slopes reinforced with anchors,
2 PRINCIPLE OF SUPERPOSITION
assuming that an anchor only supplies an axial ten-
sion to stabilize a slope. The axial tension has been
2.1 Principle of superposition
assumed to act at different positions: at the base of
the slice where the slip surface intersects the anchor For an anchor-reinforced slope shown in Figure 1, if
(e.g. Geo-Slope International Ltd. 2000), or at the the concept of Fellenius’ method was used for the
ground surface where the anchor head is located (e.g. safety factor, the normal force at the base of the slice
Oasys Ltd. 2001, Yamakawa 1995). Additionally, as where the slip surface intersects the anchor is given

593
 b b

W W=0

S P + Sa P

N Na

(a) Anchor-reinforced slope of a weightless soil

Figure 1. Anchor-reinforced slope. b

by Equation 1 based on the force equilibrium in the

direction perpendicular to the base of the slice.
W

N = W cos α + P sin (α + θ ) (1) S

Ns
Calculating the shear strength st the base of the (b) Slope considering the weight of the soil. The shearing
slice with the Mohr-Coulomb equation, and using the resistance at the base of the slice, S, is Ss+Sa.
equilibrium of the resisting moment and the driving
moment, we can obtain the safety factor of anchor- Figure 2. Anchor-reinforced slope superposed an anchor-
reinforced slopes for Fellenius’ method as follows. reinforced slope of weightless soil and a slope considering
the weight of the soil.
 
n (cl + W cos α) tan φ + m P sin (α + θ) tan φ
FS =  
n W sin α − m P cos (α + θ) For the slope with the weight of the soil, the normal
(2) force on the base of the slice is given by

where n is total number of slices, and m is total number Ns = W cos α (5)

of rows of anchors.
The safety factor of Fellenius’ method can also be
obtained by principle of superposition, as shown in Thus the total mobilized shear strength at the base
Figure 2. The normal force at the base of the slice of the slice can be expressed using the principle of
for the anchor-reinforced slope of a weightless soil is superposition as follows.
given by
1
S = Ss + Sa = (cl + Ns tan φ) + Sa (6)
Na = P sin (α + θ ) (3) Fs

This normal force allows an additional frictional Equilibrium of the resisting moment and the driving
shearing resistance to be mobilized. moment is expressed by

   
tan φ tan φ Ss R+ Sa R = WR sin α− P cos (α+θ ) R
Sa = Na = P sin (α + θ ) (4) n m n m
Fs Fs
(7)

The tangential component, i.e., shear force in the

anchor, P cos(α + θ ), only supplied a resisting mo- Combining Equations 6 and 7, and solving it, we
ment or resisting force to sliding of the slope; how- can obtain the same equation as Equation 2 for the
ever, it had no effect to the normal force and shearing safety factor of anchor-reinforced slopes for Fellnius’
resistance at the base of the slice. method.

594
2.2 Bishop’s simplified method 1.0
For Bishop’s simplified method, the normal force at 0.9
the base of the slice for the anchor-reinforced slope
of a weightless soil is also given by Equation 3, and 0.8 m = 15˚
the total mobilized shear strength along a slice base
plane is given by Equation 6 but Ns is obtained as 0.7 20˚
follows. 25˚
For the slope considering the weight of the soil, 0.6
30˚
shown in Figure 2(b), the equilibrium of forces in the
0.5 35˚
vertical direction is given by
0.4
NS cos α + S sin α = W (8)
0.3
Equations 8 and 6 give the normal force and 20 30 40 50 60 70
shear strength at the base of the slice for the slope
considering the weight of the soil.
Figure 3. Ratio of additional resisting moment of Equa-
Fs W − cl sin α − P sin (α + θ ) sin α tan φ tions 13 to that of Equation 14 for a row of anchor.
NS = (9)
Fs mα
larger than that given by Equation 13. Equation 14 did
cb + W tan φ + P sin (α + θ ) cos α tan φ
S= (10) not consider the influence of the shearing resistance,
Fs mα induced by the normal force of the anchor force, on the
normal force on the base of the slice for the slope with
where the weight of the soil, as shown in Equation 9. For
tan φ a row of anchor, the ratio of the additional resisting
mα = cos α + sin α (11) moment of Equation 13 to that of Equation 14 for a
Fs slope reinforced with one row of anchor is shown in
Figure 3.
The safety factor of Bishop’s simplified method for Figure 3 shows that α+ϕm was larger, the additional
the anchor-reinforced slopes can be obtained using resisting moment calculated by Equation 14 was larger
equilibrium of the resisting moment and the driving than that by Equation 13. If α + ϕm > 90◦ , the ratio
moment, and it can be expressed by of the additional resisting moment of Equation 13 to
Equation 14, 1/(1 + tan α tan ϕ/Fs ) < 0.5.
  For Fellenius’ method, comparing the equations
n (cb + W tan φ)/mα + m P sin (α + θ) cos α tan φ/mα
FS =   of the safety factor for anchor-reinforced slopes and
n W sin α − m P cos (α + θ) slopes with no anchor, we know that Equation 14 gave
(12) the same additional resisting moment as that in Equa-
tion 2 because the shear strength had not influence on
the normal force on the base of the slice.
Equation 12 implies that an additional resisting If the normal force at the base of the slice was cal-
moment of the anchor force can be expressed as culated directly using the equilibrium of force in the
vertical direction, we can obtain
 PR sin (α + θ ) tan φ
MRF = (13) N cos α + S sin α = W + P sin θ (15)
m
1 + tan α tan φ/Fs
Combining Equation 15 and the shear strength
The additional resisting moment of the anchor force expressed with the Mohr-Coulomb equation, we can
in BS 8081 (1989) is given by obtain the normal force and shear strength. Thus we
can obtain the safety factor of Bishop’s simplified
 method. The details can be found in Cai & Ugai

MRF = PR sin (α + θ) tan φ (14)
(2003a).
m
 
Because the angle α of the base of the slice located n (cb + W tan φ)/mα + m P sin θ tan φ/mα
FS =  
by anchors is usually larger than zero, the additional n W sin α − m P cos (α + θ)
resisting moment of the anchor force in BS 8081 is (16)

595
 For Equation 16, the additional resisting moment 6m 8m
can be given by

 PR sin θ tan φ

8m

MRF = (17) Rigid plate
m
cos α + sin α tan φ/Fs
D1/2
The additional resisting moment in Equation 13 was
identical to that given by Cai & Ugai (2003a). The cal-

10m
culated safety factor of anchor-reinforced slopes is

2
1:

D1
consistent with that of SSR-FEM, a different approach.
Cai & Ugai (2003a) have compared the safety factor of

D1
Equations 12 and 16. In the following section, we com- 16m
pare the safety factor for the Equation 14 and Equation
13 or 12 using the data reported by Cai & Ugai (2003a,
2003b). Figure 4. Anchor-reinforced model slope.

Table 1. Material parameters.

3 SLOPE REINFORCED BY A ROW ANCHOR
Soil-tendon Soil-grout
Figure 4 shows a model slope reinforced by a row of Parameter Soil interface interface
anchors with a height of 8 m; a slope of 1V:1H; and
a ground thickness of 10 m and the three-dimensional Young’s modulus, E(MPa) 200 200 200
finite element mesh used for SSR-FEM (Cai & Ugai Poisson’s ratio, ν(-) 0.25 0.25 0.25
2003). Two symmetric boundaries, one vertically pass- Unit weight, γ (kN/m3 ) 20.0
Cohesion, c(kPa) 12.0 0.01 12.0
ing through the anchor centerline and the other through
Friction angle, ϕ(◦ ) 20.0 0.0 20.0
the soil midway between the anchors, were used, so Dilatancy angle, ψ(◦ ) 0.0 0.0 0.0
that the slope analyzed was really stabilized by one row
of anchors with the same spacing. A small thickness
mesh was used around the anchor to reduce the com-
putational ill-conditioning of the interface elements.
The slope was reinforced with only one row anchor For the idealized slope shown in Figure 4 with no
for simplicity. anchor, SSRFEM give a safety factor of 1.09, which
An anchor, consisting by a tendon 6 m long and compares well with a value of 1.084 of Bishop’s sim-
32 mm in diameter and a grouted body 6 m long and 90 plified method. The calculated results indicate that the
mm in diameter, was connected to a rigid circular plate failure mechanism of SSRFEM agrees with the criti-
30 cm in diameter at the slope surface. The anchors cal slip surface of Bishop’s simplified method (Cai &
were installed with a horizontal distance between the Ugai 2003).
slope toe and the anchor head Lx = 4 m; an orienta-
tion indicated by an angle between the anchor and a
horizontal line θ = 15◦ ; and a center-to-center spac-
4 RESULTS AND DISCUSSIONS
ing D1 = 1.5 m; unless otherwise stated. The anchor
was modeled by finite elements in the axial direction
4.1 Effect of anchor orientation
and six wedge elements in the circumferential direc-
tion that form a semicircular shape. The tendon was Figure 5 shows the influence of the anchor orientation
assumed to contact smoothly with the soil. The shear on the safety factor in the case of Lx = 4.0 m and
strength of the soil-grouted body interface was the D1 = 1.5 m. There was an optimal range of the angle
same as that of the soil unless otherwise stated. Table 1 θ = 7.5◦ to 22.5◦ for the safety factor of the slope
lists the material parameters of the soil, soil-tendon considered herein. The safety factor is rather insensi-
interface, and soil-grouted body interface. Tendon tive to the angle θ near the optimum; this is similar
and grouted body was modeled as elastic body with to the results for the effect of anchors on the safety
Young’s modulus of 2.1 × 105 , 2.4 × 104 MPa, and factor of infinite slopes reported by Hryciw (1991).
Poisson’s ratio of 0.3 and 0.2 respectively. Although the difference between the safety factor of
The anchor force calculated by SSR-FEM was SSR-FEM and Equation 16 and BS 8081was smaller
used for limit equilibrium analyses. Consequently, than 3% for the slope reinforced with only one row
the safety factor of SSR-FEM and limit equilibrium of anchors, the safety factor of Equation 12 was more
analyses can be compared with each other. consistent with the safety factor of SSR-FEM with

596
regards to the optimal range of the angle θ . The criti- 4.3 Effect of anchor spacing
cal slip surfaces of the vertical and normal approaches
Figure 7 shows the influence of the spacing between
were almost identical to one another.
anchors D1 on the safety factor in the case of Lx =
4.0 m and θ = 15◦ . The results once again indicate
that the safety factor of Equation 12 was more consis-
4.2 Effect of anchor position
tent with the safety factor of SSR-FEM than those of
Figure 6 shows the influence of the anchor position Equation 16 and BS 8081: The maximum axial tension
Lx on the safety factor in the case of θ = 15◦ and and the failure mechanism were almost the same for
D1 = 1.5 m. The safety factor of SSR-FEM agreed the range of the spacing considered herein (Cai & Ugai
with those of limit equilibrium methods, and gen- 2003a), so the safety factor increased with the decrease
erally the agreement was better between the safety in the spacing. The spacing of anchors used in practice
factor of SSR-FEM and Equation 12. The safety fac- is usually larger than 1.5 m (JGS 4101-2000, 2000).
tor of Equation 12 was more consistent with the safety
factor of SSR-FEM than those of Equation 16 and
BS 8081. 5 SLOPE REINFORCED BY THREE ROWS
OF ANCHORS

When the slope Figure 4 is reinforced with three rows

1.25 of anchors, we compared the safety factors calculated
by two-dimensional SSR-FEM analysis, Equation 12

1.20 1.25
Fs

1.20
SSR-FEM
1.15 Eq.16
Eq.12
BS8081 1.15
Fs

SSR-FEM
Eq.16
1.10 Eq.12
1.10
0 5 10 15 20 25 30 BS8081
(degree)
1.0 5
Figure 5. Safety factor versus direction angle (Lx = 4.0 m, 1.0 1.5 2.0 2.5 3.0 3.5
D1 = 1.5). D1 (m)

Figure 7. Safety factor versus direction angle (Lx = 4.0 m,

1.30 D1 = 1.5).

1.25

1.20
Fs

1.15 SSR-FEM
Eq.16
Eq.12
1.10 BS8081

1.0 5
0 2 4 6 8
Lx (m) Figure 8. Two-dimensional finite element mesh (Rigid
bearing plates of anchors are simulated by assuming that the
Figure 6. Safety factor versus anchor position (D1 = 1.5 m, displacement of such nodes within the extent of the solid
θ = 15◦ ). black lines on the slope surface is the same).

597
 1.50 compared with the results of SSR-FEM. The com-
SSR-FEM parison shows that the safety factor calculated using
1.45
Eq.16 the principle of superposition for stability analysis
Eq.12
BS8081
of anchor-reinforced slopes was more consistent with
that calculated with SSR-FEM, a different approach
1.40
than the conventional equation and BS 8081.
The principle of superposition can be easily
1.35 extended to other slope stability analysis methods and
slopes reinforced with nails and geosynthetics.
1.30

1.25
REFERENCES
0 5 10 15 20 25 30
BS 8081.1989. Ground anchorages. British Standards Insti-
tution.
Cai, F. & Ugai, K. 2003a. Reinforcing mechanism of anchors
Figure 9. Safety factor of a slope reinforced with three rows in slopes: a numerical comparison of results of LEM and
of anchors versus angle θ . FEM. Int. J. Numer. Anal. Meth. Geomech., 27: 549–564.
Cai, F. & Ugai, K. 2003b. Numerical comparisons of stabiliz-
ing effects of anchors in slopes between limit equilibrium
(Cai & Ugai 2003b), and Equation 16 and BS 8081. methods and elasto-plastic finite element method. Journal
The position of the three rows of anchors is Lx = 2, of the Japan Landslide Society, 40(4): 8–14.
4, and 6 m, and anchor force is 30 kN/m. The two- Corona, E. P. 1996. Stabilization of excavated slopes in basalt.
Proc. 7th Intern. Symp. Landslides, vol. 3, Senneset K
dimensional finite element mesh, as shown Figure 8,
(ed.). A.A. Balkema: Rotterdam, 1771–1776.
is used for SSR-FEM analysis. Geo-Slope International Ltd. 2000. User’s guide for
Figure 9 shows the influence of the anchor orienta- SLOPE/W, Version 4. Geo-Slope International Ltd.:
tion on the safety factor of the anchor-reinforced slope. Calgary.
For the slope reinforced with three rows of anchors, the Greenwood, J.R. 1985. Geogrids and anchors for slope
safety factor of Equation 12 was more consistent with stabilization-simple method of analysis. Proc. 11th Inter.
the safety factor of SSR-FEM than those of Equation Conf. Soil Mech. Found. Eng., Vol. 5, San Francisco:
16 and BS 8081. 2770–2771.
Hashimoto, I., Kawasaki, K. & Kodera, H. 1986. Cases
of landslide control by dead-anchors. Proc. 21st Japan
National Conf. Soil Mech. Found. Eng., Sapporo:
6 CONCLUSIONS 1483–1486 (in Japanese).
Hryciw, R.D. 1991. Anchor design for slope stabilization
This paper proposes to use principle of superposition by surface loading. Journal of Geotechnical Engineering
for stability analysis of anchor-reinforced slopes. The ASCE; 117 (8):1260–1274.
principle of superposition is implicitly used in Felle- JGS 4101–2000. 2000. Standard for Design and Construction
nius’ method. This paper proposes to use the principle of Grouted Anchors. The Japanese Geotechnical Society:
of superposition for Bishop’s simplified method. As Tokyo (in Japanese).
Jewell, R.A. & Wroth, C.P. 1987. Direct shear tests on
a result, the additional shearing resistance induced by reinforced sand, Geotechnique, 37(1): 53–68.
the normal component of the anchor force only was Koerner, R.M. 1984. In-situ soil slope stabilization using
considered in the equilibrium condition for the normal anchored nets. Proc. Conf. Low Cost and Energy Sav-
force at the base of the slice. This was different from ing Construction Materials. Envo Publishing Company:
that in conventional equation (Equation 16) in which Lehigh vally, Pennsylvania: 465–478.
the anchor force is directly used in the equilibrium Oasys Ltd. 2001. SLOPE 17, GEO suite for Windows,
condition in the vertical direction for Bishop’s simpli- Version 17.8.0. Oasys Ltd., Part of the ARUP Group:
fied method, and also this is different from BS 8081 in London.
which the influence of the additional shear resistance Tan, S.B., Tan, S.L., Yang, K.S. & Chin, Y.K. 1985. Soil
improvement methods in Singapore. Third Inter. Geotech.
induced by the normal component of the anchor force Seminar, Nanyang Technological Institute, Singapore:
was not considered in the equilibrium condition for the 249–272.
normal force at the base of the slice. Yamakawa O. 1995. Simplified design method of slope
The safety factor calculated using the principle reinforced with piles and anchors. Ph.D. Thesis, The
of superposition of anchor-reinforced slopes was University of Tokushima (in Japanese).

598
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Soil suction modelling in weathered gneiss affected by landsliding

M. Calvello, L. Cascini & G. Sorbino

Department of Civil Engineering, University of Salerno, Italy

G. Gullà
CNR-IRPI, Rende-CS, Italy

ABSTRACT: The paper presents a numerical procedure to estimate, at a test site, the hydraulic properties and
the net rainfall flux across the ground surface, by modelling the suction regime observed by tensiometers. The
test site is located in an area affected by landsliding that involves highly heterogenous gneissic soils. Within this
area, suction modelling is a fundamental issue as the movements of the landslides are strictly related to transient
perched water tables directly linked to rainfall. The numerical procedure adopts an inverse analysis to model the
observed suction regime and highlights the role played by the heterogeneity of the soil and by the net flux across
the ground surface on the unsaturated flow characteristics at the slope scale.

1 INTRODUCTION the adopted numerical analysis is illustrated and the

obtained results are discussed.
In natural and man-made slopes, soil suction is cer-
tainly one of the most important physical variables
governing the transfer of energy and water (i.e. infil- 2 GEOLITHOLOGICAL CONTEXT
tration, evaporation and transpiration) between the AND LANDSLIDING
atmosphere and the soil through the slope surface.
Indeed, understanding the mechanisms that control The test site is inside a study area of about 7.5 km2 in
soil suction variations, induced by changes in bound- the Western Sila Massif (Southern Italy), where several
ary conditions, and appropriately modelling the soil geological and geotechnical investigations have been
suction regime are crucial issues for the prediction of developed during the time in order to characterise the
the slope behaviour. landsliding affecting the area and the involved soils
In many natural slopes, modelling the soil suc- (Cascini et al. 1992a, 1992b, 1994). As for the soils,
tion regime may be a quite difficult task, especially weathered gneiss is largely diffused in the area. The
when the hydraulic properties of the soils forming the grade of weathering of the gneiss was defined follow-
slopes are strongly heterogeneous. In such cases, using ing the procedures developed for Hong Kong (GCO
the appropriate values of both the hydraulic proper- 1988) on similar rocks. A simplified version of the
ties and the boundary condition at the ground surface adopted classification system (Cascini et al. 1992a,
(i.e. rainfall intensity, evapotraspiration rate) is cer- Gullà & Matano 1997) is shown in Table 1.
tainly a crucial point that often calls for experimental
and numerical efforts. Among these, non-conventional
approaches, analysing the soil suction regime at the Table 1. Weathering grade of gneiss (from Cascini et al.
slope scale, can be profitably used. 1992a).
Referring to a particularly complex geolithological
context located in Southern Italy, where the land- Class Description
slides movements are strictly related to rainfall, the
paper presents an innovative procedure to cope with VI Residual and Colluvial soils
V Completely weathered rock (Saprolitic soil)
this issue. Particularly, an inverse numerical anal-
IV Highly weathered rock
ysis is carried out aimed at modelling the suction III Moderately weathered rock
regime observed by tensiometers at a test site. In II Slightly weathered rock
the paper, after the description of the geolithological I Fresh rock
context, the landsliding of the area and the test site,

599
 This classification, together with detailed geo-
morphological analyses, allowed the creation of a
landslides inventory map where landslide distribution
can be fully interpreted according to morphology, tec-
tonics and weathering grade of the outcropping gneiss.
The landslides inventory map reveals that the most
widespread types of instability phenomena involve
heterogenous residual, colluvial and saprolitic soils
(classes VI and V). These are characterised by an
impulsive kinematism, with lengthy periods of total
inactivity, followed by brief phases of sudden reac-
tivations triggered by remarkable increments in pore
water pressures induced by rainfall.
Similarly to landsliding that affects the territories of
Brasil and Hong Kong (Lacerda 2004, Brand 1984),
in-situ measurements revealed that the pore water pres-
sures in the landslide bodies are strictly related to
transient perched water tables. These last are directly
linked to the intensity and duration of rainfall events
(Cascini et al. 2006).
Therefore, landslides characterisation calls for the
definition of relationships among rainfall, pore water
pressure increments in the perched water tables and the
triggering phases of landsliding movements. To this
end, an adequate knowledge of the suction regime and
of the unsaturated flow characteristics is absolutely
necessary. An innovative approach to achieve this goal
is discussed in the following sections.

3 THE TEST SITE

The test site concerns a small landslide, with a max-

imum depth of 6 m, in a slope which is part of a
large ancient landslide (Fig. 1). The small landslide Figure 1. Test site: 1) landslide scarp; 2) limit of filled hol-
involves gneiss of classes VI and V, resting on a base- low; 3) terrace; 4) colluvial deposits; 5) gneiss of classes
VI and V; 6) gneiss of class V; 7) gneiss of class IV;
ment formed by less weathered gneiss of class IV 8) gneiss of class IV-III; 9) gneiss of class III-II; 10) borehole
and class III-II (Cascini et al. 1992b, Sorbino 1995, with piezometer; 11) tensiometers’ location; (from Gullà &
Gullà & Sorbino 1996). As for the landslide debris, the Sorbino 1996).
particle size distribution with depth (Gullà & Sorbino
1996) reveals an uneven sequence of soils belonging
to classes VI and V (Fig. 2).
The conspicuous number of piezometers installed
in the test site—in some cases up to five piezome- within the landslide debris by means of five ‘‘Jet-
ters per borehole—show, in the landslide debris, the fill’’ tensiometers (Gullà & Sorbino 1996). They were
presence of a perched water table with a marked tran- installed, along the same vertical (S6-T6 in Fig. 1), at
sient behaviour related to the seasonal meteoric events. depths of 0.81 m, 1.45 m, 2.03 m, 2.88 m and 3.58 m
On the other hand, piezometers located in the base- below ground surface (Fig. 2). Soil suction measure-
ment give remarkably lower piezometric heads with ments from the installation date (May 1993) to May
annual or multiannual regime (Fig. 1). According 1995 are shown in Figure 3 together with the daily
to the data acquired in other sites of the study area rainfall (Gullà & Sorbino 1996). Readings at the ten-
(Cascini et al. 2006), this circumstance clearly reveals siometers were taken with an average time interval of
the presence of two different groundwater regimes in about one week.
the subsoil: the first in the landslide debris, the other As for the hydraulic properties of the soils form-
in the basement. ing the landslide debris, they were provided by in-situ
As for the relationship between the perched water and laboratory tests. Saturated hydraulic properties
table and rainfall, soil suction data were collected were estimated by means of in-situ piezometer tests

600
 Suction (kPa)
0 10 20 30 40 50 60 70 80 90 100
1.E+00

1.E-01

1.E-02

k/ks
1.E-03

1.E-04
Class VI
1.E-05
Class V
1.E-06

Suction (kPa)
0 10 20 30 40 50 60 70 80 90 100
1.00

0.90

0.80

/ s
0.70

0.60
Class VI
0.50
Class V
0.40
Figure 2. Location of tensiometers, soil profile and grain-
size distribution in the borehole S6 (modified after Gullà &
Sorbino 1996). Figure 4. Experimental soil water characteristic curves,
normalized by saturated volumetric water content, and
hydraulic functions, normalized by saturated coefficient of
60 permeability, for class VI and V soils (modified after Sorbino
Daily rainfall (mm/day)

50
1995).
40
30
20
10
0
regime, a transient seepage analysis was performed.
01/05/93 31/10/93 02/05/94 01/11/94 03/05/95
-40 The domain of the analysis concerns the vertical col-
Tensiometer T1 Tensiometer T2
umn of soil where the 5 tensiometers were installed
0 (Fig. 5).
Suction (kPa)

To this aim, two different schemes were adopted.

40
With Scheme 1 the hydraulic properties of the three
different soil layers were determined by best fitting
80 Tensiometer T3

Tensiometer T5 Tensiometer T4
the numerical results with the suction measurements
120
at tensiometers T2 to T4. Scheme 2 was used to esti-
mate a reliable range of evapotraspiration rates (i.e.
Figure 3. Daily rainfall and soil suction measured by net rainfall at the ground surface) by best fitting the
tensiometers (modified after Gullà & Sorbino 1996). results of the calibrated model with the measurements
at tensiometer T5.
The analyses were carried out using the Finite
and by permeameter and oedometric tests in labora- Element code SEEP/W (GeoSlope 2004), which is
tory (Sorbino 1995). As for the unsaturated hydraulic able to integrate the well known Richards’ differential
properties, they were determined in laboratory on equation governing the transient saturated—unsatu-
homogenous undisturbed soil samples collected in the rated water flow in soils. The one-dimensional domain
gneiss layers belonging to classes VI and V. Figure 4 was discretised using quadrilateral elements with sec-
shows the obtained experimental values of volumet- ondary nodes (Fig. 5). Three different soil layers were
ric water content (θ ) and hydraulic conductivity (K) considered, which refer to the saprolitic (A), residual
against suction, normalized, respectively, by the val- (B) and upper colluvial (C) layers.
ues of the saturated volumetric water content (θs ) and The time-dependent boundary conditions of the two
of the saturated coefficient of permeability (Ks ). schemes are different. In the first one, pressure bound-
ary conditions were set at the locations of the low-
ermost and uppermost tensiometers (T1 and T5) and
were assumed equal to their record of measurements.
4 NUMERICAL ANALYSIS
In the second scheme, the lower pressure boundary
condition was set at the location of tensio-meter T4;
4.1 Model definition
while a transient flux condition, corresponding to esti-
In order to address the role played, at the slope scale, mates of the net daily rainfall intensities based on
by the heterogeneity of the soil and by the net flux monthly evapotraspiration rates, was applied at the
across the ground surface on the observed suction ground surface (i.e. upper boundary).

601
 q(t) where: s = suction; a, n, m, A, N and M = curve
fitting parameters. If these last three parameters are
assumed dependent from the previous ones through the
relations A = 1/a; N = n; M = 1 − 1/n, equations (1)

Scheme 2
and (2) are completely defined by only five parameters
(namely a, n, m, θs and Ks ).
u(t)

4.2 Scheme 1
Scheme 1 in Figure 5 was used to calibrate the five
u(t)
unknown parameters defining equations (1) and (2)
for each of the three soil layers.
The five independent parameters characterizing
each soil layer were determined using an inverse anal-
Scheme 1

ysis algorithm to minimize the error between the

measured suctions at tensiometers T2, T3 and T4
and the corresponding computed results. Figure 6
shows a flowchart of the inverse analysis procedure.
A regression analysis was performed to minimize an
objective function, which quantifies the fit between
computed results and observations. The minimiza-
tion was attained by the optimization of the input
parameters needed to perform the numerical model.
If the model fit was not ‘‘optimal’’, the procedure was
repeated until the model is optimized. For details on
u(t) the procedure used and on how to choose the relevant
parameters to optimize by inverse analysis see Calvello
Figure 5. Transient seepage Finite Element model of the & Finno (2002, 2004).
vertical column where tensiometers T1 to T5 are installed: Figure 7 shows the model boundary conditions and
mesh and boundary conditions adopted for Scheme 1 and the comparison between the results of the calibrated
Scheme 2. model and the measured suction values at tensiometers
T2, T3 and T4. The results clearly indicate that the
model well reproduces the recorded behaviour both at
For both schemes the transient numerical analysis low and high suction values.
was carried out with daily time steps covering the two-
year long monitoring period of Figure 3. The assumed
initial suction distribution was given by a steady-state
analysis in which the boundary conditions refer to the START END
first available monitoring data.
For the Scheme 1, the hydraulic properties of the Initial input parameters Optimized input parameters
three layers were assumed unknowns, but defined by
two analytical relationships. Particularly, the relation-
Iterative process

YES
ship proposed by Van Genuchten (1980) was used for Input parameters
NO Model
the variation of the volumetric water content against optimized?

suction, while for the hydraulic conductivity func-

tion the relationship proposed by Mualem (1976) Numerical Model Updated input parameters
was assumed. Both functions are represented by the
following:
Regression
Computed results
(objective function minimization)
θs
θ(s) =  s
n m (1)
1+ a Observations
 2
1 − (As)N −1 (1 + (As)N )−M
K(s) = Ks  M /2 (2) Figure 6. Flow chart of inverse analysis (from Calvello &
1 + (As)N Finno 2004).

602
 Date Only 6 parameters were calibrated by the inverse
01/05/93 31/10/93 02/05/94 01/11/94 03/05/95
-40 analysis. The results show some significant differ-
ences among the initial and final estimates of these
Suction (kPa)

40
parameters. In particular, the values of the calibrated
80

120
Measures at T5 saturated conductivities for soil layers 2 and 3 are not
equal as initially assumed and are higher than their
-40
initial estimates. The calibration also lead to a differ-
0
entiation of the shape of the curves for the three soil
Suction (kPa)

40
layers, as indicated by the different values of the first
80
Measures at T4 Computed (after calibration by inverse analysis) two curve fitting parameters, a and n.
120
The comparison between the experimental curves
-40
of Figure 4 and the calibrated curves for each soil
0
layer is shown in Figure 8. The results indicate that
Suction (kPa)

40
the calibrated hydraulic conductivity functions well
80
Measures at T3
Computed (after calibration by inverse analysis)
compare with the experimental results, while the cali-
120
brated volumetric water content curves show a more
-40 rapid desaturation of the soils with suction than the
0 measured ones. This could depend by the presence of
Suction (kPa)

40
Computed (after calibration by inverse analysis)
largest in-situ pore networks that are not adequately
80
Measures at T2
represented by the small dimensions of the specimens
120
used in the laboratory tests.
-40

0
Suction (kPa)

40
4.3 Scheme 2
80
Measures at T1
120 Scheme 2 defines a domain bounded by the ground
surface and the location of tensiometer T4 (Fig. 5).
Figure 7. Boundary conditions (tensiometers T1 and T5) The hydraulic properties of the two soil layers of the
and comparison between measured suction values at inner model (i.e. layers B and C) were assumed equal to the
tensiometers (T2, T3 and T4) and computed results after corresponding calibrated pair of curves in Figure 8. At
calibration by inverse analysis. the lower boundary, a time-dependent pressure condi-
tion based on the values measured at T4 was used.
At the upper boundary a flux condition was applied,
Table 2. Parameters of volumetric water content curve and
hydraulic conductivity function before and after calibration.
Suction (kPa)
Initial estimates After calibration 0 10 20 30 40 50 60 70 80 90 100
1,E+00
Layer
Hydraulic conductivity ratio, k/ks

1,E-01
Parameter A B C A B C 1,E-02

θs
1,E-03
0.32 0.32 0.35 0.32 0.32 0.35
1,E-04 C soil
Ks (m/d) 0.11 0.04 0.04 0.11 0.70 0.24 B soil
a (kPa) 20 (range 1–29) 6 10 20 1,E-05
A soil
n 1.5 (range 1.01–2.13) 1.5 1.35 1.1 1,E-06

m 0.33 (range 0.18–0.57) 0.33 0.33 0.33 Suction (kPa)

0 10 20 30 40 50 60 70 80 90 100
1,00
∗ in bold parameters calibrated by the inverse analysis.
Volumetric water content ratio

0,90
0,80
0,70
0,60
0,50
0,40
C soil
0,30
The calibrated values of the 15 independent model 0,20
B soil
A soil
parameters are reported in Table 2 together with their 0,10
0,00
initial estimates. The initial saturated volumetric water
content and coefficient of permeability were set equal Figure 8. Dimensionless comparison among the experi-
to the mean values reported by Sorbino (1995). Initial mental soil water characteristic curves and hydraulic func-
values of the parameters a, n and m were estimated tions (dashed lines) and the calibrated curves for soil layers
by best fitting each experimental curve (see Figure 4) A (lower solid line), B (middle solid line) and C (upper solid
with the functions defined by equations (1) and (2). line).

603
based on daily rainfall measures and on the estimates 01/05/93 31/10/93
Date
02/05/94 01/11/94 03/05/95
of the evapotraspiration rate. -40
Computed (EVT Scenario 1)
To this end, the procedure suggested by Tarantino 0
et al. (2002) was adopted. This is based on a compar-

Suction (kPa)
ison between the potential evapotraspiration (Ep ) (i.e. 40

the maximum evaporation rate in the case of water 80

availability) and the water limiting evapotraspiration 120

(Elim ), which is related to the capacity of the soil to Measures at T5 Computed (EVT Scenario 2)
160
transmit water to the atmosphere. The value of daily
evaporation rates for the definition of the flux bound- Figure 10. Comparison between measured suction values
ary is then fixed, in the numerical analyses, equal to the and computed results at tensiometer T5 for two different
lowest value between Ep and Elim . A detailed descrip- estimates of the net rainfall infiltration rate.
tion of the procedures used for the evaluation of Ep
and Elim is reported in Sorbino (2005).
The estimation of Ep was carried out adopting
the Thornthwaite’s empirical equation (Thornthwaite 5 CONCLUSIONS
1954) and using air temperatures recorded at the test
site. The limiting evapotraspiration, Elim , was esti- In the paper, the in-situ soil suction regime is modelled
mated using the steady-state flow solution proposed at a test site, to be considered representative of the
by Gardner (1958) for a homogenous soil, which is subsoil conditions for landsliding that affects a large
a function of the saturated and unsaturated hydraulic area where weathered gneiss prevails.
properties of the soil and of the depth of the water The comparison between the model results and a
table. two-year long record of in-situ measurements, rela-
Figure 9 shows the cumulative total rainfall and the tive to 5 tensiometers installed along the same vertical,
cumulative net flux at the ground surface for two shows that the soil heterogeneity and the net flux across
different evapotranspiration (EVT) scenarios. The the ground surface play a relevant role in the soil
two scenarios refer to two different estimates of suction regime.
the hydraulic properties of the soil between T4 As for the heterogeneity, the laboratory characteri-
and the ground surface, assumed as homogeneous in zation of the hydraulic properties of these soils, albeit
the Gardner solution, leading to two different estimates accurate, may not be sufficient to define reliable values
of the limiting evapotranspiration. for the unsaturated flow analysis at the slope scale. To
In Figure 10 the measured suction values at ten- this aim, accurate in-situ tensiometric measures may
siometer T5 and the computed results are compared instead be profitably used. As for the net flux across the
for the two different evapotranspiration scenarios. The ground surface, the results underline how the estima-
Figure clearly shows that, when the soil is highly tion of the evapotranspiration rate is equally relevant.
heterogeneous, the estimation of the evapotranspi- In this case, the tensiometric measurements may be
ration rate can lead to inadequate definition of the used to define a range of possible evapotranspiration
boundary conditions at the ground surface. On the rates.
contrary, the reliability of its estimation can be highly Tensiometric measurements and the presented pro-
improved when adequate experimental data on suction cedure may also be useful to address further goals.
are available. Among these, forecasting the recharge of perched
water tables at site-scale and/or predicting soil suc-
tion regime for the landslides involving class VI and
2.0 V soils that have been mapped over the whole study
area of the Western Sila Massif.
Cumulative total rainfall EVT
Cumulative flux at ground surface (m)

1.5
scenario 1

1.0

0.5 EVT REFERENCES

scenario 2
0.0
Brand, E.W. 1984. Landslides in Southeast Asia: A State-
-0.5
of-the-Art Report. Proc. 4th ISL. Toronto, Canada, pp.
01/05/93 31/10/93 02/05/94 01/11/94 03/05/95 17–59.
Date Calvello, M. & Finno, R.J. 2002. Calibration of soil models by
inverse analysis. Proc. NUMOG VIII, Rome, pp. 107–113.
Figure 9. Cumulative total rainfall and net flux at the Calvello, M. & Finno, R.J. 2004. Selecting parameters to opti-
ground surface for two different estimates of the net rainfall mize in model calibration by inverse analysis. Computers
infiltration rate (EVT scenarios 1 and 2). and Geotechnics 31 (5): 411–425.

604
Cascini, L. Gullà, G. & Sorbino, G. 2006. Groundwa- Gullà, G. & Matano, F. 1997. Surveys of weathering profile
ter modelling of a weathered gneissic cover. Canadian on gneiss cutslopes in Northern Calabria, Italy. Proc. 8th
Geotechnical Journal 43 (11): 1153–1166. IAEG, Athens, pp. 133–138.
Cascini, L. Critelli, S. Di Nocera, S. & Gullà, G. 1992a. A Lacerda, W.A., 2004. The behavior of colluvial slopes in a
methodological approach to landslide hazard assessment: tropical environment. Proc. 9th ISL. Rio de Janeiro, pp.
A case history. Proc. 6th ISL, Christchurch, 2: 899–904. 1315–1342.
Cascini, L. Critelli, S. Di Nocera, S. & Gullà, G. 1994. Mualem, Y. 1976. A new model for predicting the
Weathering and landsliding in Sila Grande Massif gneiss hydraulic conductivity of unsaturated porous media.
(Northern Calabria, Italy). Proc. 7th IAEG, Lisbon, pp. Water Resources Res. 12: 513–522.
1613–1622. Sorbino, G. 1995. Unsaturated hydraulic characteristics
Cascini, L. Critelli, S. Di Nocera, S. Gullà, G. & Matano, of weathered gneissic soils. Proc. 10th Panam. Conf.
F. 1992b. Grado di alterazione e franosità negli gneiss Guadalajara, pp. 25–35.
del Massiccio Silano: L’Area di S. Pietro in Guarano. Sorbino, G. 2005. Numerical modelling of soil suction mea-
Geologia Applicata ed Idrogeologia 27: 49–76. surements in pyroclastic soils. Proc. Conf. on Advanced
Gardner, W.R. 1958. Some steady-state solutions of the unsat- Experimental Unsaturated Soil Mechanics, Trento, pp.
urated moisture flow equation with application to evapo- 541–547.
ration from a water table. Soil Science (85): 228–232. Tarantino, A. Mongiovì, L. & Mc Dougall, L.R. 2002. Analy-
GCO 1988. Geoguide 3: guide to rock and soil descriptions. sis of hydrological effects of vegetation on slope stability.
Geotechnical Control Office (GCO), Civil Engineering Proc. 3rd Int. Conf. on Unsat. Soils, Recife (2): 749–754.
Services Department, Hong Kong. Thornthwaite, C.W. 1954. A re-examination of the con-
GEO-SLOPE International Ltd. 2004. Seepage modelling cept and measurement of potential transpiration. In The
with SEEP/W, user’s guide version 6.16. GEO-SLOPE measurement of potential evapo-transpiration (ed. J.R.
International Ltd., Calgary, Alberta. Mather), 200–209.
Gullà, G. & Sorbino, G. 1996. Soil suction measurements in Van Genuchten, M.Th. 1980. A closed-form equation for
a landslide involving weathered gneiss. Proc. of the 7th pre-dicting the hydraulic conductivity of unsaturated soil.
ISL, Trondheim, Norway 2: 749–754. Soil Science Society American Journal 44: 892–898.

605
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Modelling the transient groundwater regime for the displacements analysis

of slow-moving active landslides

L. Cascini, M. Calvello & G.M. Grimaldi

Department of Civil Engineering, University of Salerno, Italy

ABSTRACT: Active slow-moving landslides in clayey soils exhibit continuous movements generally related to
changing ground water levels. This paper highlights the importance of groundwater modelling for the prediction
of these types of movements through the analysis of a well-documented case history: the Porta Cassia landslide
on the Orvieto hill in central Italy (e.g. Tommasi et al. 2006). The analysis uses recorded rainfall and pore
pressure data to define a reliable model of the transient groundwater regime in the slope, which is then used to
derive the time-dependent shear stresses along the main slip surfaces. The displacements at selected points along
the slip surface are then computed using a relationship between the local factor of safety and the displacement
rate at those points.

1 INTRODUCTION phenomenological relationships proposed by Vulliet

& Hutter (1988), which relates the displacement rate
According to the general framework proposed by to the time-dependent local factor of safety (i.e. ratio of
Leroueil et al. (1996) for the geotechnical character- the average shear strength to the average shear stress).
isation of slope movements, active landslides move
along one or several pre-existing slip surfaces, where
the mobilised shear strength corresponds to residual 2 THE PORTA CASSIA EARTH SLIDE
conditions. When most of the landslide shear defor-
mations occur in a narrow shear zone above the main 2.1 Site description
slip surfaces, many authors consider the nature of the
deformations as viscous (e.g. Vulliet & Hutter 1988). Orvieto is an ancient town in central Italy resting on
In this case, the displacements rate is related to chang- a tufaceous slab overlying an overconsolidated clayey
ing groundwater levels that affect the shear stress level hill. This town has long been affected by many instabil-
along the slip surface. ity phenomena (Tommasi et al. 1986). Among these, a
The international literature offers different models sudden reactivation occurred in November 1900, when
to compute landslide displacements from measured an earth slide movement, successively named Porta
pore pressure data (Angeli et al. 1996; Corominas Cassia landslide (Figure 1), affected the northern slope
et al. 2005; Mandolini & Urciuoli 1999; Maugeri et al. of the hill causing lots of damages and interrupting a
2006; Van Asch 2005; Van Asch et al. 2007; Vulliet & major road and the Rome-Florence railway (Vinassa
Hutter 1988). The main limitation of such models is the de Regny, 1904). The landslide is still active.
need for reliable pore pressure measurements along the With the aim of collecting data for a quantitative
slip surfaces. To overcome this limitation, a procedure confirmation of the mechanism governing the slope
is proposed which relates the displacement rates along evolution (Tommasi et al. 1997), two monitoring pro-
the slip surfaces to the pore pressures computed by a grammes started in 1982 and 1996–1998 and thus a
physically-based analysis of the groundwater regime. comprehensive data set of pore pressure and displace-
The analysis refers to an active translational land- ments measurements is now available. Figure 1 shows
slide in stiff clays on the Orvieto hill, in central Italy the map of the Orvieto hill, the boundaries of the Porta
(Tommasi et al. 2006). The analysis uses recorded Cassia landslide on the northern flank of the slope and
rainfall and pore pressure data to define a reliable the location of the main monitoring stations.
model of the transient groundwater regime in the slope,
which is then used to derive the time-dependent shear
2.2 Analysis of monitoring data
stresses along the main slip surfaces. The rate of
displacement in a given point along the slip surface Figure 2 shows the stratigraphy of a representative
at a given time is then computed using one of the cross-section of the landslide (cross-section 1–1’ from

607
 0

-5

g.w.l. [m] b.g.s.

-10 OV OM O4
OR
-15

-20

-25
Jan-96 Dec-96 Dec-97 Dec-98 Dec-99 Dec-00

Figure 3. Measured groundwater levels of available

piezometers (modified after Tommasi et al. 2006).

140

horizontal displacements [mm]

120
Figure 1. Boundaries of the Porta Cassia landslide on the
100 inclinometer OR
northern flank of the Orvieto hill and location of the moni-
toring stations along cross-section 1–1’ (from Tommasi et al. 80 inclinometer O4
2006). 60
40
20
Debris Cover Piezometers 0
250 Softened Clay Inclinometers
OM Jan-96 Dec-96 Dec-97 Dec-98 Dec-99 Dec-00
elevation [m a.s.l.]

Intact Clay

200 OR
OV
150
O4 Figure 4. Cumulative horizontal displacements measured
by inclinometers OR and O4 along the slip surface (modified
100 after Tommasi et al. 2006).
0 100 200 300 400 500 600
distance [m]

0.2
horizontal velocity [mm/d]

Figure 2. Representative cross-section and location of the measured 0.32

monitored piezometers and inclinometers. velocity
0.15 O4
annual average
velocity
0.1
extremely slow- very slow
Figure 1). The three layers geotechnical section has threshold
OR
been drawn using reported stratigraphic data and 0.05
geotechnical properties and includes: a clayey debris
cover, varying in depth from a few meters to 20 0
Jan-96 Dec-96 Dec-97 Dec-98 Dec-99 Dec-00
meters; a softened clay stratum, whose thickness varies
between two and ten meters; an intact overconsolidated
Figure 5. Measured and annual average velocities for the
clayey bedrock. verticals OR and O4 showing two different displacement
The Figure also shows the location of the instru- rates.
mented boreholes for which piezometric and incli-
nometric measurements are available (Tommasi et al. lower strata. Indeed, the deeper piezometer, OR, only
2006). As for the piezometers, OR was located in the records modest variations of the groundwater level
lower stratum, at a depth of 31 m b.g.s., and has been (less than 1 m); the other three, all installed within the
regularly monitored since 1988; while OV, OM and O4 upper strata, show seasonal changes of the groundwa-
were installed within the upper two layers, respectively ter level, with differences of a few meters between the
at 6, 4.5 and 6 m b.g.s., and have been monitored since dry and the wet season. The available data suggest that
1997, 1996 and 1998. As for the inclinometers, OR changes of pore water pressures in the debris cover
has been installed in the central part of the slope in are not significantly influenced by the groundwater
1982, and was since regularly monitored, while OM, regime in the deeper clayey bedrock.
OV and O4 were installed in 1996–1998, respectively Figure 4 reports the horizontal displacements
in the upper, middle and lower part of the slope. recorded, between 1996 and 2000, by two of the
Figure 3 reports the groundwater levels recorded, 4 inclinometrs shown in Figure 2. It is clear that
between 1996 and 2000, by the 4 piezometers shown two movements with two different displacement rates
in Figure 2. The measurements highlight the existence exist: a slower one in the upper part of the slope, as
of two different ground water regimes in the upper and recorded by OR at about 20 m b.g.s.; and an order

608
of magnitude faster movement in the lower part of On the basis of this hypothesis, the lower boundary
the slope, as recorded by O4 at about 7 m b.g.s.. As of the mesh was assumed to be impermeable. As for
shown in Figure 5, which reports both the measured the left boundary, the mesh ends at the location of the
and the annual average velocities of the two move- monitored vertical OM. Along this vertical, hydro-
ments, O4 records a ‘‘very slow’’ phenomenon while static conditions were assumed with the stationary
OR an ‘‘extremely slow’’ movement, according to the groundwater level set at the ground surface. These are
classification by Cruden & Varnes (1996). justified by the presence of a spring at that location
(Lembo-Fazio et al. 1984). As for the right bound-
ary, the model section ends at the foot of the slope.
2.3 Landslide classification The hydrostatic conditions with the ground water level
The in-situ investigation and the monitoring data indi- at the ground surface are justified by the presence
cate that the Porta Cassia landslide can be classified of a River (Paglia River) at that location. As for the
as an active (Leroueil et al. 1996) roto-translational top boundary, rainfall and temperature measures have
earth slide in clayey material (Varnes, 1978) with very been used to compute the net rainfall flux to apply
slow maximum measured velocity (Cruden & Varnes, at the ground surface. The net rainfall was com-
1996). Within this definition, the term active is here puted subtracting the evapotraspiration term to the
used to mean a phenomenon sliding along one or total monthly recorded rainfall, using the Thornth-
several pre-existing shear surfaces, where the mobi- waite method (Thornthwaite 1948). The frame on the
lized shear strength corresponds to residual conditions, upper right corner of Figure 6 shows the total and net
without any references to the time and/or the duration monthly rainfall during the five years considered in
of the movements. the analysis (1996–2000).
As for the hydraulic properties of the two layers,
the lower one was assumed to be saturated at all times,
3 GROUNDWATER MODEL while the upper clayey debris cover was assumed to be
in partially saturated conditions, in accordance with
3.1 Model description and hypotheses experimental evidence by Cafaro et al. (2005). Figure 7
shows the hydraulic conductivity function and the soil
Figure 6 shows the cross-section and the mesh used in water characteristic curve for this upper layer. In the
the finite elements model of the transient groundwater graphs, both the hydraulic conductivity and the vol-
flow in the slope. The analysis was performed using the umetric water content are divided by their saturated
commercial finite element code SEEP/W (GEO-Slope values. The shape of both curves is very close to the
2004). The mesh only considers the debris cover and shape of the laboratory tests curves reported by Cafaro
the softened clay layers, not explicitly considering the et al. (2005).
bedrock layer. According to Tommasi et al. (2006) and
consistently with the pore pressure measures shown 3.2 Results of the analysis
in Figure 3, the flow regime characterizing the stiff
The results of the groundwater model are reported with
clay at depth was assumed not influencing the sea-
reference to the two phases of the analysis: the ini-
sonal changes of pore water pressures inside the slope
tial stationary conditions and the successive transient
formed by the softened clay formation and the clayey
debris cover.
10
effective rainfall total rainfall
8
rainfall [mm/day]

250
6
Elevation [m] a.s.l.

OM
4

Rainfall 2
200
Hydrostatic OR
0
Pressure OV Jan-96 Jan-97 Jan-98 Jan-99 Jan-00 Jan-01

150 O4

Debris Cover Hydrostatic

Pressure
100 Softened Clay

0 100 200 300 400 500 600

Distance [m]

Figure 6. Section of the finite element transient groundwater model of the Porta Cassia landslide and chart of the total and
effective rainfall between 1996 and 2000.

609
analysis. For the former, the calibration of the As for the transient analysis, a one-year long
hydraulic conductivities of the two layers was con- calibration stage was defined using the values of
ducted by minimizing the error between the groundwa- pore pressures measured at piezometer OV between
ter level measured and computed at piezometer OV in November 1996 and October 1997. The ground water
November 1997. For the latter, a calibration stage and levels measured at O4 and all the other measure-
a validation stage have been defined when comparing ments at OV were instead used to validate the results
the numerical results with the piezometric measures of the calibrated model. Figures 9 and 10 show the
of the pore water pressure. main results of the analysis with reference to points
The stationary boundary conditions of the model along the monitored verticals OV, O4 and OR. The
are the ones described in the previous paragraph for extremely satisfactory results of the analysis are con-
the left, bottom and right boundaries of the mesh firmed by the values of the coefficients of correlation
(Figure 6), while a null value of flux is assumed at the between computed and measured ground water lev-
ground surface. Figure 8 shows the ratio between the els at piezometer OV, respectively equal to 0.981 and
saturated hydraulic conductivity of the clayey debris 0.874 for the calibration and validation stages. As
cover (k1) and the softened clay (k2) plotted against the for the point on the slip surface along vertical OR
difference between measured and computed ground- (Figure 10), a comparison with measured data is not
water levels at OV ( H). The best results, i.e. H null, shown because piezometer OR is installed at 31 m
are obtained for a ratio of about 5.2, indicating that b.g.s. and thus well below the slip surface at that
the permeability of the debris cover is about five times location. The numerical results of the calibrated and
larger than the one of the softened clay layer. Tommasi validated groundwater analysis are, therefore, even
et al. (2006) report differences of about one order of
magnitude between the two hydraulic conductivities.
Jan-96 Dec-96 Dec-97 Dec-98 Dec-99 Dec-00
0
1.E+00
GWL [m] b.g.s.

-2
1.E-01
1.E-02
k/ks

-4
1.E-03 OV computed
1.E-04 -6 OV measured
Calibrated
1.E-05 Validated
0.01 0.1 1 10 100 1000 -8
suction [kPa]
1 Jan-96 Dec-96 Dec-97 Dec-98 Dec-99 Dec-00
0.8 0
0.6
s

GWL [m] b.g.s.

0.4 -2
0.2
0 -4
O4 computed
0.01 0.1 1 10 100 1000 O4 measured
suction [kPa] -6
Validated
-8
Figure 7. Hydraulic properties of the clayey debris cover:
shape of the hydraulic conductivity function and of the soil
Figure 9. Results of the transient analysis: comparison
water characteristic curve.
between computed and measured groundwater levels at
piezometers OV and O4.
8
k1 = debris cover conductivity
6 k2 = softened clay conductivity Jan-96 Dec-96 Dec-97 Dec-98 Dec-99 Dec-00
k1/k2 [-]

H = measured GWL minus 0

4 computed GWL
GWL [m] b.g.s.

2 -4

0 -8
-1 0 1 2 3 4 5 6
H [m] -12
OR computed
Figure 8. Stationary analysis: ratio between the saturated -16
hydraulic conductivities of the debris cover and of the soft-
ened clay vs. difference between computed and measured Figure 10. Results of the transient analysis: computed
groundwater levels at OV. groundwater level on the slip surface along the vertical OR.

610
Table 1. Values of the calibrated saturated hydraulic con- real profile schematic profile h(t)
ductivities for the two layers. Pw(t)
d d

Layer Hydraulic conductivity dOX

t
[-] [m/s]
zOX F
F>1
Clayey debris cover 7·10−6 z z
F=1

Softened clay 1.16·10−6 F<1

t

(a) (b)

more significant at this location, for which measures Figure 11. Schematic representation of the movement along
of the displacements rate along the slip surface are an infinitesimal slip surface and correlation between the
computed pore water pressure and ‘‘local’’ factor of safety.
available.
Table 1 reports the values of the calibrated satu-
rated hydraulic conductivities for the two layers. For
both the debris cover and the softened clay layers, the Pw (t), as the only component of the local factor of
values of the calibrated saturated conductivities are safety varying with time (Figure 11b).
significantly higher than the values reported by Tom- The displacements along the slip surface, dOX , are
masi et al. (2006). These are respectively equal to computed introducing a threshold level of shear stress
10−10 and 10−9 m/s and were derived from falling (i.e. the residual shear strength in the Mohr-Coulomb
head tests in Casagrande piezometers. A possible rea- plane) below which there are no movements. Above the
son for that is the presence, in the softened clay layer, threshold, a phenomenological relationship is defined
of oxidized joints (Lembo-Fazio et al. 1984; Tom- relating the velocity of the movement to the local factor
masi et al. 1986) acting as preferential flow paths. of safety. The following relationship, first proposed by
Another factor to consider could be that at times, as Vulliet and Hutter (1988), was used:
reported by Senneset and Sandven (1987), the dis-
turbance induced in the soil by the installation of the  n
piezometers may cause a reduction of the measured 1 
vOX = B · ⇒ dOX = vOX · t (1)
coefficients of permeability. F (t)

where B and n are phenomenological constants rep-

4 DISPLACEMENT MODEL resenting, respectively, the typical velocity of the
phenomenon and the variation of viscosity with the
4.1 Model description and results excess shear stress (Van Asch et al. 2007).
The kinematic model, used to compute the time- The displacement monitoring data from inclinome-
dependent displacements of the Porta Cassia landslide, ter O4 (Figure 4) were used both for calibrating the
employs a phenomenological relationship between the input parameters of the model and for validating the
velocity of the slide and the shear stress level act- results of the calibrated model. Particularly, the mea-
ing along the slip surface. The latter, in turn, is sures relative to the period July 1998—November
a function of the time-dependent pore water pres- 1999 were used for calibrating the model, while the
sures computed by the transient groundwater analysis. rest of measurements were used to validate it. This
A detailed description of the model is presented in was done in order to use some of the observations for
Calvello et al. (2007). Herein, only the main assump- evaluating the ‘‘predictive capability’’ of the model.
tions of the model and the most significant results are Figure 12 shows the results of the model with refer-
discussed. ence to both the calibration and the validation stages.
Figure 11 shows that the landslide body is assumed The fit between the numerical results and the obser-
to move as a rigid block along a well defined infinites- vations is almost perfect for the calibration stage and
imal slip surface, at a depth equal to the average depth remains extremely positive for the validation stage.
of the deep creep zone, zOX (Figure 11a). The model Indeed, the coefficient of correlation between com-
is defined as ‘‘local’’ in that the analysis is carried out puted and measured displacements is 0.995 for the first
with reference to the stress state at a single point on stage and a 0.953 for the second one. The values of the
the slip surface. The state of stress at the selected point parameters of the calibrated model are summarized in
along the slip surface is computed using the infinite- Table 2, where the creep threshold was set equal to
slope scheme, assuming the slip surface parallel to the the measured residual friction angle, as determined by
ground surface and the computed pore water pressure, direct shear tests by Lembo Fazio et al. (1984).

611
 140 REFERENCES
Calibrated Validated
120
horiz. displacements [mm]

Angeli, M.G., Gasparetto, P., Menotti, R.M., Pasuto, A. &

100
O4 measured
Silvano, S. 1996. A visco-plastic model for slope anal-
80 ysis applied to a mudslide in Cortina d’Ampezzo, Italy.
O4 computed
60 Quarterly Journal of Engineering Geology 29: 233–240.
40 Cafaro, F., Boldini, D. & Tommasi, P. 2005. Drying behaviour
of the Orvieto overconsolidated clay. International Con-
20
ference on Problematic Soils, Bilsel and Nalbantoglu eds,
0 Famagusta 1: 95–102.
Jan-98 Jan-99 Jan-00 Dec-00 Calvello, M., Cascini, L. & Grimaldi, G.M. 2007. Displace-
ments scenarios at site scale of rainfall-controlled slow
Figure 12. Measured and computer horizontal displace- moving active slides in stiff clays. Proc. ISSGSR 2007, Int.
ments for the selected point on the slip surface along the Symp. on Geotecnical Safety and Risk, Shanghai, China
vertical O4. In Press.
Corominas, J., Moya, J., Ledesma, A., Lloret, A. & Gili, J.A.
2005. Prediction of ground displacements and velocities
Table 2. Values of the creep threshold and of the coeffi- from groundwater level changes at the Vallcebre landslide
cients B and n calibrated with the kinematic model for the (Eastern Pyrenees, Spain). Landslides 2: 83–96.
vertical O4. Cruden, D.M. & Varnes, D.J. 1996. Landslide types and pro-
cesses. Landslides: investigation and mitigation. Trans-
Vertical Creep threshold Coefficient B Coefficient n portation Research Board, Special Report No. 247,
[-] [◦ ] [m/d] [-] National Research Council, National Academy Press,
Washington DC, USA 36–75.
O4 11.3 1.08 e-6 10 GEO-SLOPE 2004. Seepage modelling with SEEP/W, User’s
Guide (version 6.16). GEO-SLOPE International Ltd.
Calgary, Alberta, Canada.
Lembo-Fazio, A., Manfredini, G., Ribacchi, R. & Sciotti,
M. 1984. Slope Failure and Cliff Instability in the Orvieto
5 CONCLUSIONS Hill. Proceedings of the 4th International Symposium on
Landslides, Toronto, Canada 115–120.
This paper underlines the importance of groundwa- Leroueil, S., Locat, J., Vaunat, J., Picarelli, L., Lee, H. &
ter modelling for the analysis of active slow-moving Faure, R. 1996. Geotechnical characterization of slope
slides in clayey materials. Key factors of this analysis movements. Proceedings 7th International Symposium
are rainfall and pore pressure data which can be used Landslides, Trondheim, Norway 1: 53–74.
to define a reliable model of the transient groundwater Mandolini, A. & Urciuoli, G. 1999. Previsione dell’evoluz-
ione cinematica dei pendii mediante un procedimento
regime in the slope. The obtained results can then be di simulazione statistica. Rivista Italiana di Geotecnica
used to derive the time-dependent shear stresses along 35–42.
the slip surface. The displacements at selected points Maugeri, M., Motta, E. & Raciti, E. 2006. Mathematical
along the slip surface can finally be computed using a modelling of the landslide occurred at Gagliano Castelfer-
relationship between the inverse of the time dependent rato (Italy). Natural Hazards and Earth System Sciences
local factor of safety and the displacement rate at those 6: 133–143.
points. Senneset, K. & Sandven, R. 1987. Field and laboratory
This model has been applied to the Porta Cas- testing—General report. Proceedings IX European Con-
sia slide, a well-monitored slow-moving active slide ference on Soil Mechanics and Foundation Engineering,
Dublin, Ireland.
in Central Italy. The comparison between the moni- Thornthwaite, C.W. 1948. An approach toward a rational
toring data and the corresponding computed results classification of climate. Geographical Review 38: 55–94.
is extremely satisfactory for both the groundwater Tommasi, P., Ribacchi, R. & Sciotti, M. 1986. Analisi stor-
analysis and the displacement model. ica dei dissesti e degli interventi sulla rupe di Orvieto:
The proposed model could be used as a predictive un ausilio allo studio del’evoluzione della stabilità del
tool to obtain possible displacements scenarios inter- centro abitato. Geologia Applicata ed Idrogeologia XXI:
preting the evolution of a landslide when rainfall sce- 99–153.
narios, based on statistical analysis, are defined. This Tommasi, P., Ribacchi, R. & Sciotti, M. 1997. Slow Move-
aspect is particularly relevant when the mechanism ments along the slip surface of the 1900 Porta Cassia
landslide in the clayey slope of the Orvieto Hill. Rivista
governing the landslide evolution is not straightfor- Italiana di geotecnica 2: 49–58.
ward. For instance, for the Porta Cassia landslide, one Tommasi, P., Pellegrini, P., Boldini, D. & Ribacchi, R. 2006.
may evaluate whether the sudden reactivation occurred Influence of rainfall regime on hydraulic conditions and
in 1900 was related to exceptional rainfall (i.e. rainfall movement rates in the overconsolidated clayey slope of the
with a high return period) or if other factors need to Orvieto hill (central Italy). Canadian geotechnical journal
be taken into account. 43: 70–86.

612
Van Asch, T. 2005. Modelling the hysteresis in the velocity Board Special Report No. 176, National Academy of
pattern of slow-moving earth flows: the role of excess Sciences, Washington DC, USA 11–33.
pore pressure. Earth Surface Processes and Landforms Vinassa de Regny, P. 1904. Le frane di Orvieto. Giornale di
30: 403–411. Geologia Pratica 1: 110–130.
Van Asch, Th.W.J., Van Beek, L.P.H. & Bogaard, T.A. Vulliet, L. & Hutter, K. 1988. Viscous-type sliding laws for
2007. Problems in predicting the mobility of slow-moving landslides. Canadian Geotechnical Journal 25: 467–477.
landslides. Engineering Geology 91: 46–55.
Varnes, D.J. 1978. Slope movement types and processes.
Landslide Analysis and Control. Transportation Research

613
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Numerical modelling of the thermo-mechanical behaviour of soils

in catastrophic landslides

F. Cecinato & A. Zervos

School of Civil Engineering & Environment, University of Southampton, UK

E. Veveakis & I. Vardoulakis

Faculty of Applied Science, National Technical University of Athens, Greece

ABSTRACT: A new landslide model is proposed by improving on an existing one, which is able to interpret
using a simple 1-D mechanism the post-failure sliding regime of catastrophic landslides and rockslides consisting
of a coherent mass sliding on a thin clayey layer. The model takes into account frictional heating and subsequent
pore pressure build-up, leading to the vanishing of shear resistance and unconstrained acceleration. First, an
existing thermo-elasto-plastic constitutive model for clays is discussed, and modified by re-formulating it in a
general stress space and taking into account thermal softening. The soil constitutive model is then employed into
an existing landslide model. The resulting model equations are shown to be well-posed, and then are discretised
and integrated numerically to back-analyse the final stage of the well-documented case history of Vajont that
occurred in Italy in 1963. Finally, the results are used to highlight the possible importance of thermal softening
in the development of catastrophic failure.

1 INTRODUCTION Jiufengershan rock and soil avalanche triggered by

the Chi-Chi (Taiwan) 1999 earthquake (Chang et al.,
The Vajont landslide of October 9, 1963, has been 2005a, 2005b).
the subject of numerous geological and geomechani- Vardoulakis (2000, 2002) analyzed the pressuriza-
cal investigations, due both to its potential contribution tion phase of the Vajont slide, when thermal pres-
to slope stability analysis and to the social and legal surization sets in, during which the slide accelerates
implications of the disaster. The landslide moved rapidly. He proposed a one-degree-of-freedom, fric-
approximately 2.7 × 108 m3 of rock into an artificial tional pendulum model, employing a Mohr-Coulomb
reservoir of about 1.5×108 m3 , impounding the Vajont constitutive model for the soil and assuming that
deep gorge. The slide moved an 120 m thick (on aver- frictional heating triggered pore water pressurization
age) compact rock mass over a front of 1850 m for inside a shear band of the order of 1 mm. This analysis
a maximum slip of 450–500 m (Hendron and Patton, showed that the catastrophic pressurization phase of
1985) and at a final slip rate of about 25–30 m/s. The the Vajont slide should not have taken more than a few
abrupt filling of the reservoir with debris produced seconds to develop in full.
a giant wave (4.8 × 107 m3 ) that propagated up and In this paper we extend the above study by
down the valley, overflowing the dam and wiping out using a more general thermo-elasto-plastic constitu-
the village of Longarone, located 2 km west. tive model, based on the one recently proposed by
Habib (1975) proposed that the high slip veloc- Laloui et al. (2005). Furthermore we investigate the
ity achieved by the Vajont landslide was due to the impact of thermal softening, which some clays exhibit,
conversion of mechanical energy into heat during in the development of the catastrophic mechanism.
frictional sliding, which should lead to the ‘‘vapor- In the following, we present in section 2 the land-
ization’’ of pore water and hence to a cushion of slide model developed by Vardoulakis (2000, 2002),
zero friction. Temperature increase in the slipping and in section 3 the new thermo-elasto-plastic consti-
zone may also have led to pressurization of pore tutive model. Section 4 deals with the modification
water with the same effect on the shear strength of of the landslide model to include this new constitutive
the slope (Anderson, 1980; Voight and Faust, 1982; law. Finally, in section 5 some computational results
Vardoulakis, 2000, 2002). Total loss of strength by are presented and discussed and conclusions are drawn
thermal pressurization has also been claimed for the in section 6.

615
2 LANDSLIDE MODEL 3 THERMOPLASTIC CONSTITUTIVE MODEL

A thermo-poro-mechanical landslide model has been 3.1 Problem formulation

derived by Vardoulakis (2000, 2002) from first prin-
A simple constitutive model for clays taking into
ciples, and successfully employed to back-analyse the
account thermo-plastic behaviour has been recently
catastrophic sliding phase of the Vajont slide. In land-
developed by Laloui and co-workers (Laloui 2001,
slides of that type, all deformation is assumed to
Laloui & Cekerevac 2003, Laloui et al., 2005). It was
be concentrated on a thin clayey layer with thick-
formulated in ‘triaxial’ stress space and it was used to
ness of the order of 1 mm (Tika & Hutchinson 1999,
simulate isothermal isotropic compression for differ-
Vardoulakis 2002), overlain by a coherent mass sliding
ent temperatures and thermal loading paths (Laloui &
as a rigid block.
Cekerevac 2003).
The model consists of a set of coupled par-
Laloui et al. (2005) added to the isotropic ther-
tial differential equations describing the time evo-
mal law a deviatoric mechanism based on original
lution of temperature, pore pressure and velocity
Cam-Clay within the framework of multi-mechanism
within the shear-band. As the rigid mass starts slid-
plasticity, which implied the use of two separate plastic
ing, strain and strain-rate softening occur within the
multipliers. A simple linear law of thermal softening
shear-band, alongside frictional heating. As soon
was also proposed, as experimental evidence sum-
as the critical temperature for thermoplastic col-
marised by Laloui et al. (2001) shows that the friction
lapse of the soil skeleton is reached, pore pressure
coefficient of some soils decreases with increasing
build-up takes place leading to catastrophic acceler-
temperature (Hicher 1974, Despax 1976, Hueckel
ation.
et al. 1989).
The equation describing the movement of the rigid
The above thermoplastic model can be further
block is obtained through moment balance, using a
developed by adopting the isotropic hardening law
modified slip-circle method. The resulting velocity
describing the evolution of apparent preconsolidation
is the upper boundary condition for the shear-band
stress (Laloui & Cekerevac 2003) and the linear ther-
velocity profile, thus providing the coupling between
mal softening law (Laloui et al., 2005). With these
shearband and sliding mass.
ingredients we can re-derive the constitutive equa-
The model consists of the following equations:
tions from a standard plasticity framework, with a
general stress space formulation in order to facili-
⎧ tate future generalisation of the model in 2- and 3-D.
⎪ ∂θ ∂ 2θ D
⎪
⎪ = k + Furthermore, Modified Cam-Clay will be adopted
⎪
⎪ ∂t
m
∂z 2 j(ρC)
⎪
⎪ m for the deviatoric behaviour, as this is a widely used
⎪
⎨ ∂u  model with clear advantages when it comes to numer-
∂ ∂u ∂θ
= cv + λm (1.1) ical implementation. The resulting shape of the yield
⎪
⎪ ∂t ∂z ∂z ∂t
⎪
⎪ surface in 3 dimensions is shown in Figure 1.
⎪
⎪ 
⎪
⎪ dv ud
⎩ d = Rω02 A(μm ) + .
dt pc (μm f ) 3.1.1 Thermo-elasticity
The thermoelastic law is derived from the classic
Duhamel-Neumann relations (e.g. see Mase, 1970):
The first equation describes the evolution of tem-
perature. The first term on the right-hand side is
a diffusion term, where km is a constant diffusivity M=M 0

coefficient, and the second one is a heat generation

term, where D is the dissipation and is taken equal
to the plastic work. The second equation describes
the diffusion of excess pore pressure u, where cv a M=M T

temperature-dependent consolidation coefficient and

λm a temperature- and pressure-dependent coefficient
determining the generation of excess pore pressure.
It is zero until the temperature reaches the criti-
cal value for thermal collapse of the soil skeleton.
Finally, the third equation describes the time evolu-
tion of the slide velocity and is composed of two
terms that depend on mobilised friction and pore Figure 1. Thermo-plastic yield locus, demonstrating tem-
pressure. perature dependence and thermal softening.

616
σ̇ = Dme ε̇ me + Dte θ̇ (1.2)

where Dme is the stress-dependent elasticity tensor,

Dte = −βs Kδ the thermal tensor, K is the bulk modu-
lus, βs the volumetric thermal expansion coefficient, δ
the Kronecker delta, ε the strain and θ the temperature.

3.1.2 Thermo-plasticity
At plastic yielding, employing standard techniques,
the above can be rewritten in terms of total strain and
temperature rates as:
⎧
⎨ σ̇ = D ε̇ + D θ̇
mep tep
⎪

⎪ fεp fp Figure 2. Numerical simulation of some thermo-

⎩ σ̇c = v · λ̇ mechanical isotropic paths. The dashed line represents iso-
fσc
thermal loading, the solid line thermo-mechanical loading,
the dash-dotted path was obtained by applying a higher
where σc is the preconsolidation stress, p the mean temperature rate during the first 5% of total loading,
p
effective pressure, and f = f (σij , εv , θ ) the yield sur- followed by mechanical loading only.
face derived by substituting the thermal hardening law
of Laloui et al. (2003) into the standard expression
of the Modified Cam Clay yield locus. Associativity
is assumed, and partial derivatives of the yield func-
tion with respect to other model variables are denoted
by subscripts. In contrast to Laloui et al. (2005),
where two different plastic multipliers were defined
for isotropic and deviatoric deformation respectively,
here we use a single plastic multiplier:

fσ Dme ε̇ + (fθ + 2Dte fσ )θ̇

λ̇ =
fσ Dme gσ − fεvp gp

derived from the consistency condition df = 0.

Figure 3. Simulation of isothermal loading and combined
thermo-mechanical loading compared to experimental data of
3.2 Numerical results isothermal isotropic compression at different temperatures.
The above constitutive relations were discretised using The black line simulates thermo-mechanical loading until
a refined explicit scheme with automatic error con- temperature reaches 60◦ C, followed by mechanical loading
trol (Sloan et al., 2001), controlling independently the only. It can be seen how the state path moves between the two
sets of data corresponding to isothermal compression at the
applied total strains and the temperature. The response respective temperatures.
of a soil element to different strain and temperature
paths is demonstrated in Figure 2. It is shown how the
apparent preconsolidation stress decreases as tempera- between two isothermal loading data sets. Good agree-
ture increases, and how the slope of the plastic loading ment with experimental data points is achieved in both
path changes if temperature and mechanical strains are cases.
applied at the same time.
In Figure 3 numerical results are compared to some
experimental data on three isothermal isotropic com-
4 MODIFIED LANDSLIDE MODEL
pression paths carried out at different temperatures
(Laloui & Cekerevac 2003). The dashed line simulates
4.1 Constitutive modifications
an isothermal loading path at ambient temperature,
while the solid line represents a combined thermo- The system of equations (1.1) provides a 1-D analy-
mechanical loading path to simulate the transition sis of the phenomenon, employing a Mohr-Coulomb

617
yield locus as soil constitutive model. To provide a diffusivity varies non-linearly with temperature and
more realistic constitutive assumption, which can be pore pressure as
more easily generalised for applications to a wider
range of soils, we modify the equations to include the km
thermoplastic model described above. Di = 2[M (θ)(p0 −u)]2 F1 (θ,u)
.
The soil is assumed to be at critical state since our 1− j(ρC)m
analysis starts at incipient failure. Plane strain is also
assumed due to the problem geometry. Thus, it is Such an expression can, theoretically at least,
possible to express the dissipation term as assume negative values, which would imply mathe-
matical ill-posedness of the equation and inability to
D = 2λ̇q2 = 2λ̇(Mp )2 solve it. To ensure that this does not happen and that the
problem remains well-posed, the sign of the diffusiv-
ity coefficient is calculated for a wide range of values
where M = M0 − g(θ − θ0 ) is linearly dependent of its parameters and the temperature range 0 < θ <
on temperature g is the thermal sensitivity and M0 1000◦ C. Diffusivity proved to be always positive for all
the critical state parameter at reference temperature. values examined of the parameters involved, showing
The mean effective pressure p can be expressed as the very little variation around the value of 10−7 m2 /sec.
difference between the initial effective stress acting on The parameter ranges used are: 0.25 ≤ M0 ≤ 0.85 and
the landslide base p0 and the excess pore pressure u: 0.1 ≤ σc0 ≤ 10 MPa for the critical state parameter
and the preconsolidation stress respectively at ambient
p = p0 − u(z, t)
p
temperature, 0.003 ≤ εv ≤ 0.5 for the accumulated
plastic volumetric strain, 5 ≤ β ≤ 20 for the plastic
Thus, compressibility, 10−3 ≤ k ≤ 5 · 10−2 for the slope
of the elastic compression line, 0.2 ≤ ν ≤ 0.45 for
the Poisson’s ratio, 0.2 ≤ voidr ≤ 1.5 for the void
D = 2λ̇(θ̇ , ε̇q ){M (θ )[p0 − u(z, t)]}2 ratio and 0.05 ≤ γ ≤ 0 for the parameter-γ of the
thermoplastic model (Laloui et al., 2003).
and
4.3 Numerical implementation
λ̇ = F1 θ̇ + F2 ε̇q
Assuming a linear velocity profile within the shear-
band we obtain the following system of governing
where: equations:
(fθ − 2Kβs fp ) ⎧ 
F1 = , ⎪
⎪ ∂u ∂ ∂u ∂θ
fp2 K + 3Gfq2 − fp fεvp ⎪
⎪ = cv (θ) + λm (θ, u)
⎪
⎪ ∂t ∂z ∂z ∂t
⎪
⎪
3fq G ⎪
⎪ ∂θ ∂ θ
2
F2 = ⎪
⎪ vd
fp2 K + 3Gfq2 − fp fεvp ⎪
⎪ = Di (θ, u) 2 + Fi (θ, u)
⎪
⎨ ∂t ∂z d
  
⎪ dvd = Rω2 A 1 − M (θd )
and G and Kdenote the standard stress-dependent ⎪
⎪ 
⎪
⎪ dt 0 0
tan ϕF 3 − M (θd )2
elastic moduli. ⎪
⎪
⎪
⎪ 
By substituting the dissipation expression in the ⎪
⎪
⎪
⎪ ud
heat equations, we finally obtain ⎪
⎪ +
⎩ pc (θd )
∂θ km ∂ 2θ
= 2[M (p0 −u)]2 F1 The second equation describing the evolution of
∂t 1− ∂z 2
j(ρC)m temperature within the shearband is, unlike its cor-
2[M (p0 − u)]2 F2 responding one of Section 2, nonlinear, since both
+ ε̇q . (1.3) the diffusivity term and the heat generation term are
j(ρC)m − 2[M (p0 − u)]2 F1 temperature- and pore pressure-dependent. The third
equation describing the movement of the rigid block is
modified by the implementation of thermal softening,
4.2 Well-posedness of heat equation
and coupled to the previous two through the values of
Expression (1.3) is a diffusion-generation equation for temperature, excess pore pressure and velocity vd at
temperature, where, unlike Vardoulakis (2002), the the interface of the shearband with the rigid block.

618
Figure 4. Temperature and pore pressure profile within the shearband and its surroundings, for thermal softening.

Figure 5. Velocity and displacement profile if thermal softening is included.

The first two nonlinear partial differential equations were chosen equal to those established by Vardoulakis
were discretised using a Finite-Time Centered-Space (2002).
explicit scheme with a very small time-step dt =
10−4 sec, which was found through numerical exper-
imentation to yield stable numerical results. The last
equation, which is an ordinary differential equation, 5 RESULTS AND DISCUSSION
was discretised with a standard 4th order Runge-Kutta
scheme using the same time-step. In Figure 4 the isochrones of temperature and pore
Calculations were performed for a time window pressure are shown for the case of g = 10−2 , which
of 10 seconds from the initiation of sliding, using corresponds to the average value for the rate of ther-
the physical parameters of ‘section 5’ of Vajont land- mal softening found in the literature. It can be seen
slide (Hendron & Patton 1985, Vardoulakis 2002). that temperature inside the shearband reaches 120◦ C
Initial conditions for temperature and excess pore pres- after 10 seconds. This is well below the water vapori-
sure are the steady state values, equal to ambient sation threshold at the given pressure. The overburden
temperature (12◦ C) and zero respectively. corresponds to an initial effective stress of 2.38 MPa,
The 1-D spatial domain is here set to be 10 times which is the maximum value that excess pore pressure
the thickness of the shearband (taken 1.4 mm, after can reach. It can also be seen that full pressurisation is
Vardoulakis 2002) and assumed to be uniform in reached after t = 7 s, corresponding to the vanishing
hydraulic and geotechnical properties. On the other of the shear resistance.
hand, the shearband is the only area where shear strain- In Figure 5 the corresponding computed slide veloc-
ing, i.e. the mechanism driving energy dissipation and ity and displacement are plotted. The velocity profile
therefore heat production, occurs. Temperature and shows negligible increase up to t = 2 s, followed by
pore-pressure were computed at each grid-point of a sudden kick at t = 3 s which corresponds with the
the spatial domain and isochrones through the domain start of pressurisation due to overtopping of the crit-
were produced at significant time values. The value of ical temperature (c.f. Figure 4). These values predict
the slope of the thermal softening law g representing a similar behaviour and magnitude of results to those
the ‘thermal sensitivity’ of the soil proved to be cru- observed in Vajont, and are in accordance with those
cial to the order of magnitude of catastrophic evolution calculated by the strain and strain-rate softening model
of the phenomenon. Other model parameters values of Vardoulakis (2002).

619
Figure 6. Temperature and pore pressure profile within the shearband and its surroundings, for zero softening.

Figure 7. Velocity and displacement profile for zero thermal softening.

In Figure 6 and Figure 7 results are plotted for the observations, as well as past calculations, shows that
case of no thermal softening (g = 0). The drastic dif- the contribution of thermal softening can be signifi-
ference in the computed values shows how sensitive cant in the destabilisation of a slide, and should be
is the timescale of the phenomenon to thermal soften- taken into account in addition to strain and strain rate
ing. After 10 seconds the velocity is still of the order of softening.
mm/sec, temperature has increased by a few degrees
Celsius and pressurisation has not taken place. A much
longer time would be needed for catastrophic sliding REFERENCES
to be triggered in this case.
The results show that thermal softening acts as a Anderson, D.L. 1980. An earthquake induced heat mech-
destabilising factor in slides and that its effect can be anism to explain the loss of strength of large rock and
comparable in magnitude to that of strain and strain- earth slides. Proceedings of the International Conference
rate softening. on Engineering for Protection From Natural Disasters,
pp. 569–580, John Wiley, New York.
Chang, K.J., Taboada, A., Lin, M.-L. & Chen, R.-F. 2005a.
6 CONCLUSIONS Analysis of landsliding by earthquake shaking using a
block-on-slope thermomechanical model: Example of
Jiufengershan landslide, central Taiwan. Eng. Geol., 80,
A new catastrophic landslide model has been pro- 151–163.
posed, able to interpret the frictional heating and Chang, K.J., Taboada, A. & Chan, Y.C. 2005b. Geo-
pressurisation phase in catastrophic landslides. The logical and morphological study of the Jiufengershan
model accounts for thermal softening instead of the landslide triggered by the Chi-Chi Taiwan earthquake,
formerly proposed strain- and strain-rate softening Geomorphology. 71, 293–309.
mechanism. The model uses a thermoplastic consti- Despax, D. 1976. Influence de la température sur les pro-
tutive model based on Modified Cam-Clay, allowing priétés mécaniques des argyles saturées. Doctoral thesis,
future generalisation to a wider range of soils and Ecole Centrale Paris.
Habib, P. 1975. Production of gaseous pore pressure during
extension to 2-D and 3-D calculations. rock slides, Rock Mech., 7, 193–197.
The model was discretised and integrated numer- Hendron, A.J. & Patton, F.D. 1985. The Vajont slide, a
ically to back-analyse the behaviour of the well- geotechnical analysis based on new geologic observations
documented case of Vajont landslide. The good of the failure surface, Tech. Rep. GL-85–5, U.S. Army
accordance of the predicted values with available Corps of Eng., Washington, D.C.

620
Hicher, P.Y. 1974. Etude des proprietes mecaniques des Sloan, S.W., Abbo, A.J. & Sheng, D. 2001. Refined
argiles a l’ aide d’essais triaxiaux, in uence de la vitesse explicit integration of elastoplastic models with automatic
et de la temperature. Report, Soil Mech. Lab., Ecole Cent. error control. Engineering Computations, 18 (1–2):
de Paris, Paris. 121–154.
Hueckel, T. & Pellegrini, R. 1989. Modeling of thermal Tika, T.E. & Hutchinson, J.N. 1999. Ring shear tests on soil
failure of saturated clays. Numerical models in geome- from the Vajont landslide slip surface. Geotechnique, 49,
chanics, 81–90, Elsevier. 59–74.
Laloui, L. 2001. Thermo-mechanical behaviour of soils. Vardoulakis, I. 2000. Catastrophic landslides due to fric-
Environmental geomechanics, 5: 809–843. tional heating of the failure plane. Mechanics of
Laloui, L. & Cekerevac, C. 2003. Thermo-plasticity of Cohesive-Frictional Materials, 5 (6): 443–467.
clays: An isotropic yield mechanism. Computers and Vardoulakis, I. 2002. Dynamic thermo-poro-mechanical
Geotechnics, 30 (8): 649–660. analysis of catastrophic landslides. Geotechnique, 52 (3):
Laloui, L., Cekerevac, C. & Francois, B. 2005. Constitutive 157–171.
modelling of the thermo-plastic behaviour of soils. Revue Voight, B. & Faust, C. 1982. Frictional heat and strength
Europeenne de genie civil, 9 (5–6): 635–650. loss in some rapid landslides. Geotechnique, 32,
Mase, G.E. 1970. Theory and problems of Continuum 43–54.
Mechanics. Schaum’s outline series, McGraw-Hill.

621
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Some notes on the upper-bound and Sarma’s methods

with inclined slices for stability analysis

Z.Y. Chen
China Institute of Water Resources and Hydropower Research, Benjing, China

ABSTRACT: In 1997, Donald and Chen proposed a slope stability analysis method that divides the failure
mass into a series of slices with inclined interfaces and solves the factor of safety based on the upper-bound
theory of plasticity. This approach is capable of producing accurate results provided by the classical slip-line
field method. It has also been demonstrated that the method is identical to Sarma’s method that has been widely
used for rock slope analysis. This paper gives some further examples demonstrating its accuracies and discusses
two issues involved in the use of the upper-bound or Sarma’s method: (1) the requirement for identifying
alternative directions of relative movement between two contiguous slices; (2) treatments when tension occurs
in the calculations.

1 INTRODUCTION 2 THE UPPER-BOUND METHODS

Among a variety of limit equilibrium approaches, 2.1 Background

Sarma’s method that divides the failure mass into a
The method suggested by Sarma (1979) assumes that
number of slices with inclined interfaces has been
both on the interfaces and base of a slice the limit equi-
commended by Hoek (1983, 1987) for rock slope
librium state applies. The factor of safety F is defined
stability analysis as rock mass normally contains at
to be a coefficient that reduces the available effective
least one set of sub-vertical discontinuities that can
shear strength parameters c and φ to new values of ce
be best modeled by the inclined interfaces. Sarma
and φe in order to bring the structure to failure,
assumes that limit equilibrium conditions prevail on
both the base and interfaces and solve the factor of
safety by the force equilibrium conditions. Donald ce = c/F (1)
and Chen (1997) presented a new approach that estab-
lished the governing equation based on the virtual tan φe = tan φe /F (2)
work principle. In addition to its elegant analyti-
cal formulations, the method enjoys a sound the- In the following presentations, the subscription ‘e’
oretical background supported by the upper-bound appeared for all variables would invariably mean that
theory of Plasticity. A series of testing problems the related c and φ values are reduced by Eqs. (1)
has demonstrated its ability of producing closed- and (2).
form solutions provided by the slip-line field method In order to explain the equivalence between Sarma
using its numerical procedures. In this Paper, we and the upper bound methods, Chen (1999) presented
will present an analytical example demonstrating its a slope that is composed of two blocks as shown in
accuracy. Figure 1. The factor of safety can be readily solved by
Practical applications of Sarma’s method have establishing force equilibrium equations for the left
encountered the following two problems that will be and right blocks, according to Sarma’s concept. How-
discussed in this Paper: (1) the requirement for iden- ever, the problem can be solved in a more efficient way
tifying alternative directions of relative movement by the Virtual Work Principle.
between two contiguous slices; and (2) treatment when In this problem, the normal force N and its contribu-
tension occurs in the calculations. tion of shear force N tan φ on each of the faces forms

623
 Donald and Chen (1997) argued that there are two
possible directions of relative movement: the left slice
moves upward relative to the right one as shown in
Figure 2(a) or downward, in Figure 2(b). Detailed
discussions will be given in Section 3.

2.2 Formulations of the upper-bound methods

The governing equation for calculating the factor of
safety has been given by Donald and Chen (1997),
which is
 xn  
Figure 1. A two-block slope explaining the equivalency dW dTy
between Sarma and the upper-bound methods. (ce cos φe − u sin φe ) sec α − +
x0 dx dx
 
dW dTx
× sin(α − φe ) − η + cos(α − φe )
dx dx
 xn
× E(x)dx − (cej cos φej − u j sin φej )
x0

dα
× L csc(α − φe − θj ) E(x)dx + Ki = 0 (6)
dx
where
Figure 2. Determining V r and V j based on V l . (a) The   x 
dα
left slice moves upward relative to the right one; (b) The left E(x) = κ exp − cot(α − φe − θj ) dζ (7)
slice moves downward relative to the right one. x0 dζ

a resultant P that inclines at an angle φe to the normal 

i j
sin(αil − φeil − θi )
of the failure plane. If we assign a set of virtual dis- Ki = j
(8)
sin(αir − φeir − θi )
placements Vl , Vr , Vj , each inclined at an angle of φe j=1
to the shear surface and allow all forces applied on the
slope to do work on them. The work done by Pl , Pr , Pj Ki is s a coefficient accounting for possible discon-
on Vl , Vr , Vj respectively is thus zero. Pl , Pr , Pj , as tinuities in α, φe and ce , defined as
unknowns, disappear in the work and energy balance
equation, which is 
n

Ki = − Cej cos φej − uj sin φej i Li

Al cl cos φel Vl + Ar cr cos φer Vr + Aj cj cos φej Vj i=1

= Wl Vl cos ρl + Wr Vr cos ρr (3) × csc α r − φer − θj sin( α − φe )i E l (xi )

where ρ is the angle between the weight vector and V . (9)

Eq. (3) remains only one unknown F which is implied
in the subscripts ‘e’ and is readily obtainable. Tx and Ty are horizontal and vertical surface
The values Vr , Vj can be expressed as a linear load respectively, and η refers the horizontal seismic
function of Vl (Refer to Figure 2), coefficient.
The first term of the left-hand side of Eq. (6) refers
sin(θl − θj ) to the work done by the external loads and the energy
Vr = Vl (4) dissipation on the slip surface, while the second term
sin(θr − θj )
is the energy dissipation on the interfaces between two
sin(θr − θl ) contiguous slices.
Vj = Vl (5)
sin(θr − θj )
2.3 Theoretical verifications
where θ is the angle of the velocity vector measured A number of test examples that have closed-form solu-
from the positive x axis. The superscripts j refers to tions by the slip-line field method (Sokolovski, 1960)
the variable on the interfaces, and l and r refer to the have been performed, confirming that the optimiza-
left and right sides of the interfaces. tion process of the upper-bound method is capable of

624
 π φ
creating the critical failure mode of the plastic zone μ= − (11)
and the minimum factor of safety identical to the the- 4 2
oretical solutions. This paper gives a further example
that demonstrates the equivalence between Eq. (6) and The distance between a point on the log-spiral BC and
the solution given by Sokolovski in an analytical form. the origin O, as represented by OE with a length L, can
On a separate Paper (Chen and Ugai, 2008), we gave be obtained by the equation,
a demonstration of the equivalence by the numerical
approaches of the upper-bound method. L = Lb exp[−(δb − δ) tan φ] (12)
Figure 4 shows a uniform slope subjected to a
vertical surface load q. The weight of the soil mass where δ is the angle between OE and the y axis. Lb and
is neglected. The closed-form solution for the ulti- Lc are the lengths of OB and OC respectively.
mate vertical surface load is provided by Sokolovski We have
(1960) as:
  Lc = Lb /s (13)
1 + sin φ
q = c cot φ exp[(π − 2γ  ) tan φ] − 1
1 − sin φ where
(10)
 π  
 s = exp − γ  tan φ (14)
where γ is the inclination of the slope surface. 2
The slip-line filed includes a series of failure
surface, each of them consists of two straight lines (e.g, At any point on BC, we have
AB and CD) connected by a log-spiral (e.g. BEC).
AB and CD incline at μ to the slop surface and crest α − φ = −δ (15)
respectively. BO and CD incline at μ to OA and OD
respectively. and

dx = −L sec φ cos αdδ (16)

For a slope at a state of limiting equilibrium, F = 1.

Calculating the terms in Eq. (6) with the particular
conditions defined by equations (11) to (16), as shown
in Table 1, we have

q(1 − sin φ) = (s2 + 1)c cos φ + (s2 –1)c cot φ

(17)

Figure 3. Sketch for the analyses by the energy approach It is not difficult to demonstrate that Eq. (10) is
of Sarma’s method. identical to Eq. (17). This example validates Eq. (6),
which is original in our discipline.

3 ON THE ALTERNATIVE DIRECTIONS

OF THE INTERSLICE FORCES

3.1 The argument

When establishing the velocity field, one must exam-
ine whether the left slice moves in an upward (Case 1)
or downward direction (Case 2), with respect to the
right one, as shown respectively in Figure 2(a) and (b).
Failure to do so will cause negative values of Vr or
Vj . when Eqs. (4) and (5) are employed. Substituting
these negative values into the work and energy balance
Figure 4. Verifications of Eq. (6) compared to the theo- equation means a violation of Drucker’s Postulate. As a
retical solution provided by the slip-line field method. consequence, the calculation will lead to absurd results

625
Table 1. Integrals of various terms on Eq. (9) for example.

 l (c j cos φ j − u j sin φ j )
l (c cos φ− u sin φ)  ×Lcosec(α − φ − θj )
dTy
Section E(x) sec αE(x)dx l dx sin(α − φ)E(x)dx × dα
dx E(x)dx

AB l Lc sc cos φ 0 0
exp[−(δb − δ) tan φ] 1 1
BC cLc cot φ(s − s−1 ) 0 cLc cot φ(s − s−1 )
= exp[−(α − αb ) tan φ] 2 2
CD s−1 Lc s−1 c cos φ qLc (1 − sin φ)s−1 0

indicating that the bigger the cohesion value at the equilibrium equations. Hoek (1987) noticed that on
shear surface where the negative velocity develops, some occasions, these values may be negative. This
the smaller the factor safety. An example of explaining means that tensile forces are actually exerted on these
this absurd event was given by Chen (1999). faces. This is contradictory to the fundamental prin-
ciple we have employed in Sarma or the upper-bound
methods, which assumes that the frictional shear force
3.2 The criteria
is proportional to the normal force on the failure sur-
The criteria for identifying Case 1 and Case 2 and their face, a phenomenon only valid under compressive
associated parameters are given by Donald and Chen conditions. In his program SARMA, Hoek (1987)
(1996). They can be summarized as follows. warned that the results may not be applicable. How-
Case 1. The left slice moves in an upward direction ever no further instruction was offered to handle this
with respect to the right one, i.e. θl > θr situation.
In this case, we have

θl − θj > 0 (18) 4.2 A brief review of Lajtai’s failure criteria

for rock bridges
and θj is defined as It has been understood that the rock mass can more
or less provide tensile resistance due to the existence
π
θj = − δ + φej (19) of rock bridges. Lajtai (1969) carried out a theoretical
2 study and a series experiments for rock bridges under
enforced shear and confirmed that a rock bridge would
Case 2. The left slice moves in a downward
exhibit the following three failure modes at different
direction with respect to the right one. i.e. θl < θr
values of normal stress σn .
In this case, we have
 1. Failure in tension
θl − θj < 0 At low value of σn , tension will develop and
(20)
θr − θj > −π result in failure, at which shear strength τa is
determined by
and θj is defined as

3π τa = σt (σt − σn ) (22)
θj = − δ − φej (21)
2
where σt is the tensile strength of the rock bridge.
Chen et al. (2005) demonstrated that with these 2. Failure in shear
conditions, Vr and Vj , calculated by Eqs. (4) and (5), When σn reaches a certain higher level, the
will always be positive. rock bridge will fail in shear, at which τ is deter-
mined by
4 THE TREATMENTS WHEN TENSION 
OCCURS 1 (2cr + σn tan φr )2
τa = − σn2 (23)
2 1 + tan2 φr
4.1 The argument
Once the factor of safety is obtained, all internal forces where φr and cr are friction angle and cohesion of the
applied on a slice can be calculated based on the force intact rock respectively.

626
3. Failure at ultimate strength
If σn exceeds the compressive strength of the rock
bridge, the shear strength will be determined based
on that developed by crushed rocks with the following
criterion.

τa = σn tan φu (24)

where φu is the friction angle of crushed rocks.

In case of slope stability analysis, σn is on most
cases small, compared to σt . Therefore Case 1,
i.e., failure in tension, will always apply. Since on most
slope problems, σt is much bigger than σn , eq. [22] can
be approximated as

τa ∼
= σt

This means that the shear strength of a rock bridge Figure 6. An example that explains the treatments for
can be regarded as a purely cohesive material, whose alternative directions and tensions.
cohesion can be approximated to be its tensile strength.

4.3 The suggested method to deal with tensions

vertical when Sarma’s method was performed. The
Based on the understanding that the rock bridge could shear strength parameters were both assumed to be
behave like a purely cohesive material with a c value zero for the part of interface that passes through
approximately equal to its tensile strength, it is sug- the concrete dam body. This means that the concrete
gested that once tensile force appears on the base or parts of the slices 4, 5 and 6 are assumed to be a
interfaces of a slice when Sarma or the upper-bound stack of concrete blocks without any international fric-
method is performed, their strength parameters will be tion and cohesions. However the interfaces between
substituted with c = σt and φ = 0, and the problem block 6 and 7 is a longitudinal construction joint. The
will be recalculated with these new parameters. The shear strength parameters representing this joint were
solution is then ‘applicable’. assigned for this joint.
The shear strength parameters used in the stabil-
ity analysis are indicated in Table 2 from which one
5 AN ILLUSTRATIVE EXAMPLE may find that a rapid change in strength parameters at
point 7 that transit from the slice base 67 (represent-
Figure 6 shows a sketch of stability analysis for the ing the rock joint) to 78 (representing the concrete).
Three Gorges dam along a potential slip surface con- Both the issues of the alternative shear force direc-
sisting of a number of long persistent joints designated tions and tensions were raised in this case. In table 2
1234567. The failure mode assumes that a part of Alternative 1 employs the correct direction of inter-
failure surface cut the concrete dam at 78. slice shear forces according to the criteria described
Since the foundation consists of two sets of ver- in Section 3.2, while Alternative 2 invariably employs
tical joints the inter-slice faces were assumed to be Case 1 for all interfaces.

1. The first round of computation

The calculation employs the original strength
parameters as shown in Table 2 for this round.
Firstly, one may find that the Alternative 1
and 2 gave substantially different factors of safety,
which are 2.966 and 2.048 respectively. The lat-
ter underestimated the value of F because negative
velocities, obtained from Eq. (4), were involved
in the computation. Secondly, tensions associated
with negative values of the normal inter-slice forces
E on interfaces 6 and 7 for Alternative 1 and 2
Figure 5. A rock bridge under enforced shear (Lajtai, respectively. This means that the parameters on the
1969). interfaces require readjustment.

627
Table 2. The results for the three rounds of computations.

Number of the interfaces

Round Alternative Parameters and forces 2 3 4 5 6 7 8 F

1 1 c (kPa) 200 200 200 200 200 3000 0 2.966

φ (◦ ) 35.0 35.0 35.0 35.0 35.0 47.7 0
E (9.8 kN) 1331.8 981.5 584.0 70.8 −342.4 3310.4 2112.8
X (9.8 kN) 528.0 379.2 −277.5 326.2 492.1 −7618.0 0
2 c (kPa) 200 200 200 200 200 3000 0 2.048
φ (◦ ) 35.0 35.0 35.0 35.0 35.0 47.7 0
E (9.8 kN) 2103.9 1935.0 1384.3 1212.1 1029.9 −2827.2 2112.14
X (9.8 kN) 1027.9 875.1 675.4 808.1 888.2 10782.3 0
2 3 4 5 6 7 8
2 1 c (kPa) 200 200 200 200 200 3000 0 2.912
φ (◦ ) 0 0 0 0 0 47.73 0
E (9.8 kN) 1320.5 917.0 395.0 −95.0 −516.1 3231.6 2112.6
X (9.8 kN) 241.4 269.0 −203.8 313.1 398.0 −7730.0 0
2 c (kPa) 200 200 200 200 200 3000 0 1.966
φ (◦ ) 35.0 35.0 35.0 35.0 35.0 0 0
E (9.8 kN) 2219.5 2073.4 1544.1 1402.6 1250.8 −3267.4 2112.1
X (9.8 kN) 1112.0 961.0 760.6 909.6 1004.1 9645.9 0
2 3 4 5 6 7 8
3 1 c (kPa) 200 200 200 200 200 0 0 2.419
φ (◦ ) 0 0 0 0 0 0 0
E (9.8 kN) 1601.3 1292.0 895.9 525.7 219.7 −362.4 2113.6
X (9.8 kN) 290.5 323.8 −245.3 376.8 479.0 0 0
2 c (kPa) 200 200 0 200 200 0 0 2.455
φ (◦ ) 35.0 35.0 35.0 35.0 35.0 0 0
E (9.8 kN) 1675.9 1412.5 900.9 569.0 264.8 −302.1 2114.9
X (9.8 kN) 735.5 581.1 183.9 490.7 522.8 0 0

2. The second round of computation From these calculation results, we may conclude
In this round, we assigned φ = 0 for interface 2 that the correct answer is associated with Alterna-
to 6 for Alternative 1. For Alternative 2 the φ value tive 1 and factor of safety is between 2.912 and 2.419
for interface 7 was set to zero. The results contained depending on how much the cohesion at interface 7 can
in Table 2 still exhibit tension on interfaces 6 and 7 be developed. On the other hand, the Morgenstern-
for Alternative 1 and 2 respectively. However they are Price method gave F = 2.589, that is in general
considered to be acceptable since these interfaces do agreement to the upper-bound solutions.
not offer friction. The factors of safety are 2.912 and
1.966 for Alternative 1 and Alternative 2 respectively,
which still deviate from each other appreciably.
6 CONCLUSIONS
3. The third round of computation
In order to confirm that the results given by Alter- This paper provides an example demonstrating that the
native 2 was wrong; this round of calculation set the controlling equation (6) of the upper-bound method
value of c to be zero for interface 4 and 7. It can be proposed by Donald and Chen (1997) is reducible to
found in Table 2 that Alternative 2 gave a factor of that provided by the classical slip-line field method.
safety 2.455, greater than 1.966 of the second round of This useful numerical approach for slope stability
calculation in which c was not zero. This is obviously analysis thus has been offered a more sound theoretical
absurd. On the other hand, Alternative 1 decreased background.
factor of safety from 2.912 to 2.419 by reducing the This paper also discussed the two issues commonly
cohesion from 3000 kPa to zero at interface 7, which encountered when the upper-bound or Sarma’s method
can be regarded as a normal situation. is performed.

628
1. The requirement for identifying alternative Chen, Z.Y. and Ugai, K. 2008. Limit equilibrium and
directions of relative movement between two con- finite element analysis —a perspective of recent advances.
tiguous slices. Criteria for identifying correct Proceedings of the 10th International Symposium on
directions have been given. landslide and engineered slopes. Beijing.
2. Treatments when tension occurs in the calculations Donald, I. and Chen, Zuyu. 1997. Slope stability analysis by
an upper bound plasticity method. Canadian Geotechnical
based on the Lajtai’s criteria for rock bridge strength Journal, 34(11): 853–862.
criteria. Hoek, E. 1987. General two-dimensional slope stability
An illustrative example has been given indicating analysis-Analytical and Computational Methods in Engi-
that a proper handling of these two issues will ensure neering Rock Mechanics. 95, AlIen Unwin, London.
Hoek, E. 1983. Strength of jointed rock mass. Geotechnique
a correction solution for rock slope stability analysis, 33 No.3: 187–223.
especially when the slip surface exhibits rather irreg- Lajtai, E.Z. 1969. Strength of discontinuous rocks in direct
ularity in shape and abrupt changes in the strength shear. Geotechnique, (2): 218–233.
parameters. Morgenstern, N.R. and Price, V. 1965. The analysis of the
stability of general slip surface, Geotechnique, (15) l:
79–93.
REFERENCES Sokolovski, V.V. 1960. Statics of soil media. (Translated by
Jones DH and Scholfield AN) Butterworth, London.
Chen, W.F. 1975. Limit analysis and soil plasticity. Elsevier. Sarma, S.K. 1979. Stability analysis of embankments and
Chen, Z. and Morgenstern, N.R. 1983. Extensions to the gen- slopes. ASCE Journal of the Geotechnical Engineering
eralized method of slices for stability analysis, Canadian Division. 105(GT12): 1511–1524.
Geotechnical Journal. 20(1): 104–119. Chen, Z.Y. 1999. The limit analysis for slopes: Theory,
Chen, Z.Y., Wang, X.G., Yang, J., Jia, Z. and Wang, Y.J. methods and applications. Shikoko’99. Vol. 1. 15–30.
2005. Rock slope stability analysis: theory, method and Balkema.
computer programs, China Water Resources Press., 2005
(in Chinese).

629
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Slope stability analysis using graphic acquisitions and spreadsheets

L.H. Chen
Beijing Jiaotong University, Beijing, China

Z.Y. Chen & Ping Sun

China Institute of Water Resources and Hydropower Research, Beijing, China

ABSTRACT: This paper describes an approach that performs the conventional slope stability analysis by
catching the geometry information through a graphic information acquisition program STAB_E.LSP in Auto-
CAD and computing the factor of safety in an MS Excel spread sheet named LOSSAP.XLS. The coordinates
of various controlling points of the slope profile can be easily captured by STAB_E.LSP, which are imported to
STAB_E.LSP whose VBA facility has allowed a series of subroutines for calculating the weight, pore pressure,
strength parameters, etc., by which the factor of safety can be easily obtained.

1 INTRODUCTION • The calculation still involves a large amount of

manual work for determining the material and
Almost simultaneously with the advent of computers, geometry input data of each slice.
our profession started exploring its potential for facil- • The software has not been completely open through
itating the conventional limit equilibrium analysis of the web and is therefore not widely applicable to
slope stability problems by computer programs (Whit- potential users.
man and Bailey, 1967, Morgenstern and Price, 1965).
However, nowadays the conventional slope stability A handful approach, which is easy to understand
computation using software is still a luxury under- and self-explanatory, is proposed in this paper. The
taking for most geotechnical engineers. Some com- geometrical information of the slope profile and slip
mercial software is available but with the following surface shown in an AutoCAD image can be automat-
limitations: ically captured by a program coded in Lisp language.
• It is not accessible to most geotechnical engineers; These data are imported into the spreadsheet, from
• It disclaims any legal liability for the calculation which the in-built VBA subroutines will calculate the
results; basic data for slope stability analysis, such as weight,
• It requires training to the users, which is not always pore pressure, shear strength parameters of each slice.
practically possible. Lack of fully understanding to The various limit equilibrium methods for slope sta-
the program may lead to incorrect calculations. bility analysis have been arranged in separated sheets,
which calculate the factors of safety in a completely
Using a spread sheet to calculate the factor of safety transparent format.
can be an attractive although its functions may not be This paper describes the technical details of the
as powerful as commercial software. The calculation graphic acquisitions and spreadsheets facilities. The
details in a spread sheet are transparent, self-checkable concept of establishing a library of slope stability anal-
and ready for further extensions by users. Only a lim- ysis programs shared through web is also discussed.
ited amount of coding efforts are required. Therefore
theses spread sheets can be shared by geotechni-
cal engineers through the web on a non-commercial
basis. 2 DATA ACQUISITION FROM A GRAPHIC
Low and Lee (1997) reported their develop- IMAGE
ments of spreadsheets for calculating the factor of
safety by using the method proposed by Chen and In the vector based graphic software AutoCAD, the
Morgenstern (1983). However, their work has not property of each entity, such as point, line, surface,
been widely extended perhaps due to the following etc., has been stored in its own database. The in-built
reasons: language AutoLisp can access and operate the entity

631
property in the database directly, which provides a fea- The main flowchart of the acquisition program is
sible way to catch the geometry information of the shown in Figure 1, which includes identifying the Poly-
slope drawn as an AutoCAD image. line objects and acquiring the coordinates of all the
A graphic information acquisition program named vertex of the Polyline. Function ‘‘entsel’’ of AutoLisp
STAB_E.LSP was developed by AutoLisp language was employed to ask the user to select a Polyline and
to determine the input data for slope stability analysis return the name of the selected object, and Function
automatically using Polyline in the AutoCAD image ‘‘entget’’, for returning a list of the definition data of
(Figure 1). Before applying the program, the line the entity. An AutoLisp list is a group of related values
defining the border between two different soil layers, separated by spaces and contained in parentheses. The
as well as that defining the phreatic surface and the slip first value in the parentheses is called group code. The
surface, should be redrawn by Polylines. After loading group code for the entity type is 0. If the value follow-
the program STAB_E.LSP, the user-computer interro- ing 0 equals to LWPOLYLINE, a Polyline object is
gation will start at the command column (Figure 2). selected. The group code for the vertex coordinates of
The program requires the user to select Polyline objects the LWPOLYLINE entity is 10. By searching through
interactively by clicking them and inputting some fun- the list, all vertex coordinates of the selected Polyline
damental data such as the soil layer number underlying could be obtained. Figure 2 shows a picture during the
this line. process of data acquisition. The captured information
is saved in a data file with a default name ‘qqq.dat’.

Begin
3 FORMULATIONS FOR SLOPE STABILITY
ANALYSIS
Set UCS
3.1 Advantages of using the methods with integral
and explicit formats
Select an Object Most slope stability analysis methods involve itera-
tions for factor of safety by employing certain algo-
Yes rithms, which can fall into the following categories.
Is Empty 1. Trial-and-error
This technique employs some randomly selected
No No variables and expects the difference in factors of safety
between two consecutive trials is small enough to meet
Is Polyline the convergence criterion. For example, the Janbu’s
generalized method (1973) starts with a series of t(x)
(the gradient of shear force on the vertical interfaces)
Read Material type of soil underlying and an assumed value of factor of safety F. The algo-
the Polyline rithm expects the differences in t(x) for each slice and
F between two consecutive trials are all small enough.
There is no particular theoretical or practical ground
to ensure convergence. This technique is not suitable
Obtain Coordinates of all the Vertex of Polyline
for a spreadsheet.
2. Consecutive calculations
This technique is normally used for the method that
Change the Coordinates from WCS to UCS satisfies the force equilibrium method only, such as
the method proposed by Corps of Engineers, Lowe
and Karafiath. An initially guess of F is assumed and
Write the Coordinates and underlayer material the calculation starts at the top first slice, numbered 1,
type to the data file which determines the inter-slice force G1 applied on
the left side of the slice. Consecutive calculations will
enable the determination of G2 , G3 , . . . ,Gn for the sub-
End sequent slices. It is expected that Gn to be zero. Oth-
erwise F will be re-assumed until this criterion is met.
It is obvious that Gn is dependent on the slice
Figure 1. Flow chart of program STAB_E.LSP WCS: number n and is therefore unfriendly to the non-
World coordinate system UCS: User coordinate system. linear programming solver provided by Excel, which

632
Figure 2. Data acquisition in an AutoCAD interface.

requires the optimized object contained in a fixed cell them, Newton-Raphson method is the mostly com-
rather than in a dynamic position. Coding by Excel monly known and adopted by Morgnestern and Price
can be extremely difficult, if not possible. (1965) as:
3. Non-linear regressions
Normally, a generalized slope stability analysis Gn ∂M ∂Gn
∂λ − Mn ∂λ
n
Fi = Fi+1 − Fi = ∂Gn ∂Mn ∂Gn ∂Mn
(3)
method involves both force and a moment equilibrium ∂λ ∂F − ∂F ∂λ
equations designated respectively as
−Gn ∂M ∂Gn
∂F + Mn ∂F
n

f (F, λ) = 0 (1) λi = λi+1 − λi = ∂Gn ∂Mn ∂Gn ∂Mn

(4)
∂λ ∂F − ∂F ∂λ
M (F, λ) = 0 (2)
The iteration terminates when Fi and λi meet
the convergence criterion.
where F is the factor of safety and λ is a coefficient
As a matter of fact, Excel offers a powerful non-
defining the inter-slice force inclination. A number
linear programming solver, provided that the opti-
of researchers (Fredlund, 1967, Spencer, 1967) sug-
mized objects are all contained in fixed cells, free
gested creating the linear regression curves Ff and
of the slice number n as discussed previously. Our
Fm by solving Eqs. (1) and (2) respectively associated
approach to solving this problem is to set n a very
with different values of λ. F is determined by finding
large value covering all possible slice number the use
the point of intersection of the two curves (Fig. 3).
might take, say 100. Doing some calculations with nil
While this technique is feasible to normal computer
columns or rows would not be harmful for a spread-
programming, although not efficient, it is not easy to
sheet, provided that the formulation does not cause
be realized in a spreadsheet.
overflow problems due to a variable whose value is
4. Non-linear programming zero. This means that both f (F, λ) and M (F, λ) should
A variety of non-linear programming techniques be preferable in integral and explicit format. By ‘inte-
are available for solving Eqs. (1) and (2). Among gral’ we mean that both f (F, λ) and M (F, λ) should

633
 The two integrations involve all known valuables
except F and λ. They are suitable for spreadsheets
programming.
2. For the generalized method of slices (Chen &
Morgenstern, 1983), the formulations are:
 b
p(x)s(x)dx = 0 (7)
a
 b
p(x)s(x)t(x)dx − Me = 0 (8)
a

3. For Sarma or the upper-bound method, the

formulation is

 xn  
dW
(ce cos φe − u sin φe ) sec α − sin(α − φe )
Figure 3. Non-linear regressions for the factor of safety. x0 dx
(9)
 xn
E(x)dx − (cej cos φej − uj sin φej )L csc
eventually lead to manipulations of summing up the x0
valuables contained in a certain column in an Excel
dα
sheet, and by ‘explicit’ we mean that the variables × (α − φe − θj ) E(x)dx + Ki = 0 (10)
involved in the formulations for f and M should only dx
contain F and λ, free of any intermediate objects, such
as the inter-slice force Gi determined through consec-
utive calculations. It is probably due to this reason that 4 CALCULATING THE FACTOR OF SAFETY
Low and Lee (1997) selected Chen and Morgenstern’s THROUGH A SPREAD SHEET
formulations for their attempt in creating the first offi-
cially reported spreadsheet in the area of slope stability 4.1 Structure of the spread sheet
analysis. A spread sheet named LOSSAP.XLS is developed
for computing the factor of safety by inputting the
data contained in ‘qqq.dat’ and formulating Eqs.
(5) to (9). It includes three parts, six worksheets.
3.2 Formulations based on the integral
Worksheet ‘‘Finfo’’ is calculates the basic informa-
and explicit formats
tion of each slice, such as its average unit weight,
It is obvious that the formulation for Bishop’s simpli- pore pressure, strength parameters, etc. The second
fied method is explicit and involves the summations part ‘‘Load’’ works for inputting external loads. The
for the driving and resistant moments of all known third part includes several sheets that calculate the
variables except the factor of safety. It is suitable for safety factors using the methods proposed by Bishop,
spreadsheets. Spencer, Sarma (or Donald & Chen), and Chen and
In their keynote paper of this Proceedings, Chen Morgenstern (1983) respectively.
and Ugai (2008) summarized the analytical presenta-
tions as follows (the explanations for the symbols are
abbreviated). 4.2 Calculations for the basic information
Worksheet Finfo has four regions as shown in Figure 5.
1. For Spencer method, the formulations are: The left upper region is concerned with the fundamen-
tal information including the geotechnical properties
 b of soils, and the water level out of the slope. At the
p(x) sec(φe − α + β)dx = 0 (5) upper right there is a column that reads the input of
a
data ‘qqq.dat’, which has been saved after performing
 b STAB_E.LSP as described in Section 2. Immediately
p(x) sec(φe − α + β)(x sin β − y cos β)dx = Me after reading the geometry information contained in
a ‘qqq.dat’, the in built subroutines calculate geometri-
(6) cal and geotechnical information for each slice.

634
Figure 4. Spreadsheet Finfo.

The width of the slice, inclination at the base of each

slice, and the coordinates of the centre of the slice base
can be calculated directly by the simple formulas in a
spreadsheet. However, determination of the height and
average unit weight of the slice, as well as the shear
strength parameters and the pore water pressure ratio,
require several special sub-routines. These values will
be calculated by a user-defined Function named HWM
coded by VBA, which is a macro language to improve
the flexibility and automatic ability of Office software,
such as Word, Excel.
Function HWM first computes the intersection and
vertical distance between the center at the base of the
slice and each interface line segment. The intersection
points are sorted by their ordinates (the value of the y
axis), using a technique called ‘bubbling algorithm’.
This procedure allows the determination of weight of
the slice and the shear strength parameters of the slice
base, which are stored in Column T to X. Similarly,
the pore pressure, normally expressed in terms of pore
water pressure coefficient ru at the base of the slice,
can be calculated and stored in Column Y.

4.3 Computing the factor of safety

With the basic information for each slice calculated by
‘Finfo’ and ‘Load’, it is possible to compute the fac-
Figure 5. Spreadsheet for solve FS by Bishop’s method. tor of safety at the Worksheet ‘Morg-Chen’, ‘Bishop’,

635
‘Spencer’, and ‘Sarma’ with the help the non-linear in column AE and AF respectively. As mentioned
programming facilities provided by Excel. previously, our approach to summing up the sliding
For Bishop’s method, the task is relatively easy. and resistant moments is to set a very large slice
Calculations for sliding and resistant moments of number, n =100, which covers all possible value a
each slice are straight forward, which are contained user may employ. The formula of cell AF19 is: ‘‘=
sum (AI24:AI123)−sum (AJ24:AJ123)’’ (Figure 5).
In Solver tool’s dialog box, we set cell AE19 as vari-
able, in which an initial value of factor of safety, F, is
contained, and AF19, as target. By requiring AF19 to
be zero, the final solution for F will be caught immedi-
ately by the Solver and updated in cell AE19, as shown
in Figure 6.
Low and Lee (1998) described the technique of
implementing Chen-Morgenstern method in a spread
sheet. The Three-node Gauss-Legendre Quadrature
method was employed to solve the integration of the
force and moment equilibrium equations (7) and (8).

Figure 6. Pop-up dialog box of solver.

Table 1. The material property of the soil.

No. of γd γsat
30 soil φ(◦ ) c (kPa) (kN/m3 ) (kN/m3 ) ru
25

20
1 38 0 19.5 19.5 0
- 80 - 70 - 60 - 50 - 40 - 30 - 20 - 10
2 23 5.3 19.5 19.5 0
3 20 7.2 19.5 19.5 0
Figure 7. Example of ACADS.

Figure 8. Spreadsheet for solve F by Chen-Morgenstern method.

636
Similar technique can be used for Spencer method that Further work on updating the algorithms and
employs equations (5) and (6). extending its applicability is certainly an encourag-
ing deed, which should also be non-commercial, and
carried out by all geotechnical engineers interested in
5 AN ILLUSTRATIVE EXAMPLE this area. A concept of establishing a website entitled
‘Library Of Slope Stability Analysis Programs (LOS-
Figure 7 is an example that has been documented by SAP)’ has been developed. The authors wish it will
Chen and Ugai (2008) of this Proceedings, designated come to reality in the near future.
Example 5. The geotechnical property parameters are
shown in Table 1. The minimum factor of safety is
1.366. Using the same critical slip surface, the spread- REFERENCES
sheet solution is 1.368 for Chen-Morgenstern method
and 1.366 for Spencer method. Chen, Z.Y. & Morgenstern, N.R. 1983. Extensions to the gen-
Figure 8 shows the calculation procedures and eralized method of slices for stability analysis. Canadian
details related to Chen-Morgenstern (1983) method. Geotechnical J. 20 (1), 104–119.
Chen, Z.Y. & Shao, C.M. 1998. Evaluation of minimum
factor of safety in slope stability analysis. Canadian
Geotechnical J. 25 (4), 735–748.
6 CONCLUSIONS Fredlund, D.G. 1984. Analytical methods for slope stabil-
ity analysis. 4th International symposium on landslides.
The approach described in this Paper allows practi- 209–228.
tioners to perform routine slope stability analysis work Janbu, N. 1973. Slope stability computations. Embankment
without the need for commercial software and to check Dam Engineering, 47–86.
the computation details on their own effort. The data Low, B.K. & Lee, S.R. 1998. Slope stability analysis
acquisition technique in an AutoCAD interface per- using generalized method of slices. J. Geotechnical and
mits a rapid capturing of geometric information, which Geoenvironmental Engineering. 124: (4) 350–362.
Morgenstern, N.R. & Price, V. 1965. The analysis of the
is processed by the VBA programs in Excel for deter- stability of general slip surface. Geotechnique, 15 (l):
mining the geometry and geotechnical information of 79–93.
each slice. The integral and explicit forms for slope Spencer, E. 1967. A method of analysis of embankments
stability analysis make the procedures of calculating assuming parallel inter-slice forces. Geotechnique, 17:
factors of safety in a Spread Sheet very simple and 11–26.
straight forward. Whitman, R.V. & Bailey, W.A. 1967. Use of computers for
This work requires a limited amount of coding. The- slope stability analysis. J. Soil Mechs. Fnd. Div. ASCE.,
refore, it should be established on a non-commercial Vol. 93: 475–498.
basis. The data acquisition program and Spread sheet
described in this Paper has been announced at the
website: www.geoeng.iwhr.com/geoeng/download.htm.

637
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Efficient evaluation of slope stability reliability subject to soil parameter

uncertainties using importance sampling

Jianye Ching
National Taiwan University of Science and Technology, Taipei, China

Kok-Kwang Phoon
National University of Singapore, Singapore, China

Yu-Gang Hu
National Taiwan University of Science and Technology, Taipei, China

ABSTRACT: Evaluating the reliability of a slope is a challenging task because the possible slip surface is
not known beforehand. Approximate methods via the First-Order Reliability Method (FORM) provide efficient
ways of evaluating failure probability of the ‘‘most probable’’ failure surface. The tradeoff is that the failure
probability estimates may be biased towards the unconservative side. The Monte Carlo Simulation (MCS) is a
viable unbiased way of estimating the failure probability of a slope, but MCS is inefficient for problems with
small failure probabilities. This study proposes a novel way based on the importance sampling technique of
estimating slope reliability that is unbiased and yet is much more efficient than MCS. In particular, the issue
of the specification of the importance sampling Probability Density Function (PDF) will be addressed in detail.
An example of slope reliability will be used to demonstrate the implementation of the new method.

1 INTRODUCTION convert the problem into the standard Gaussian input

space. Monte Carlo simulation contains the following
Slope stability analysis has a long history in geotech- steps:
nical engineering. In its nature, the analysis of slope
stability is obscured by uncertainties, e.g.: material a. Draw Z samples {Zi : i = 1, . . . , N} from the
uncertainties and spatial variabilities. The conse- standard Gaussian PDF.
quence is that even for slopes with nominal safety b. For each sample Zi , conduct a deterministic slope
factor more than 1, they are not necessarily safe. Relia- stability analysis to find the most critical slip sur-
bility, namely one minus failure probability, rigorously face among all trial surfaces. If the safety factor of
quantifies such uncertainties. In recent years, analysis the most critical surface is less than 1, the entire
methods (e.g.: Chowdhury and Xu (1993), Christian slope is considered to fail for that Zi sample.
et al. (1994), Low and Tang (1997), Low et al. (1998) c. Repeat Step 2 for i = 1, . . . , N. The average of
and Griffiths and Fenton (2004)) have been proposed the failure indicators is simply an estimate of the
to evaluate reliability of slopes. However, evaluating failure probability of the slope.
reliability of a slope is not an easy task, primarily due
to the lack of knowledge of the failure surface. Mathematically, the MCS procedure can be sum-
A standard procedure of evaluating the reliability marized as the following equation:
of a slope is through Monte Carlo simulation (MCS)
 
1 
(Rubinstein 1981; Ang and Tang 1984). This approach N
was taken by Griffiths and Fenton (2004) where the P(F) ≈ I min FSω (Zi ) < 1 ≡ PFMCS (1)
N i=1 ω∈
MCS method is implemented with a random field
model for spatial distribution of shear strengths. The
procedure for MCS is straightforward. Let Z denote where F is the failure event of the entire slope; I [·]
all uncertain variables in the slope of interest. With- is the indicator function;  is the set of all trial slip
out loss of generality, let us assume Z is independent surfaces; ω is one of the trial surface, and FSω is the
standard Gaussian. In the case that Z is not stan- safety factor for that trial surface, which is clearly a
dard Gaussian, proper transformations can be taken to function of Z.

639
 Another way of interpreting the overall failure Note that the slip surface variable ω is augmented
probability in (1) is as follows: let Fω denote the event into the original FORM optimization problem. Let the
FSω (Z) < 1, i.e.: the failure event of the trial sur- solution of the optimization problem be z ∗ and ω∗ , then
face ω. The overall failure event F is simply the union (− | | z ∗ | |) is the estimated failure probability and
of all individual failure events ∪ω∈ Fω . This interpre- ω∗ is the ‘‘most probable’’ slip surface. This approach
tation is graphically depicted in Figure 1. The overall was taken by Low and Tang (1997) and Low et al.
failure probability is therefore the volume under f(z) (1998) for the evaluation of slope failure probability.
(the standard Gaussian PDF) within the F region. The FORM technique is efficient and convenient
The MCS method provides an unbiased estimate because the repetitive deterministic slope stability
of the actual failure probability. However, it can be analyses required by MCS are not needed. However,
very time consuming, especially for slopes with small the tradeoff is that the FORM methods may provide
failure probabilities. This is because the coefficient of biased and unconservative estimates of the actual over-
variation (c.o.v.; standard deviation divided by mean all failure probability (Oka and Wu 1990). This can
value) of the MCS estimator PFMCS is equal to be seen in Figure 1, where the thick curve indicates
the limit state function of the most probable slip sur-
 face ω∗ . It is clear that the volume under f(z) within
δ(PFMCS ) = [1 − P(F)]/[N · P(F)] the union region ∪ω∈ Fω (the actual overall probabil-
 ity) is always greater than or equal to that within the
≈ 1/[N · PFMCS ] (2) region Fω∗ (the FORM-estimated failure probability)
because the latter is a subset of the former.
where δ(.) denotes the c.o.v. Therefore, in order What is missing in the literature is a technique
to make the c.o.v. of the MCS estimator to be as small that can provide unbiased estimate of the actual over-
as 30%, it roughly requires 10/P(F) MCS samples, all failure probability and yet only requires a small
i.e.: 10/P(F) deterministic slope stability analyses. number of repetitive deterministic slope stability anal-
A more efficient method based on the first-order yses. The purpose of this study is to demonstrate that
reliability method (FORM) was proposed to estimate it is possible to implement the importance sampling
the failure probability. The idea is to solve for the (IS) technique (Rubinstein 1981; Shinozuka 1983;
following optimization problem: Melchers 1989) to achieve so. Moreover, it is shown
by examples that the c.o.v. of the failure probability
estimator made by the IS technique can be as small
min || z || such that FSω (z) = 1 (3) as 0.2 with only 100 deterministic slope stability anal-
z,ω∈
yses for practical range of failure probability. In the
paper, the discussion of the IS technique will be made
in the context of circular trial surfaces and method of
slices although its use is obviously not limited to this
scenario. The limitation of the IS technique will also
be addressed.

2 METHODS OF SLICES

Some popular methods of determining the safety

factor for a given trial slip surface and fixed soil
parameters are briefly reviewed. This presentation
does not aim to give a complete review of the slope
stability methods but just to give enough background
for the forthcoming presentation of the importance
sampling technique. Moreover, the presentation is
limited to methods of slices with circular trial sur-
faces although the importance sampling technique
may apply to general methods and non-circular trial
surfaces.

2.1 Ordinary method of slices

Figure 1. Illustration of various failure regions in the As shown in Figure 2, a circular trial surface is under
standard Gaussian space. consideration, and the soil parameters Z are fixed

640
 conservative estimates of safety factors. It is also the
only method of slices that does not require iterative cal-
culations to obtain the safety factor. In applications,
many trial surfaces are randomly generated, and the
trail surface with the smallest safety factor is the crit-
ical slip surface. If the safety factor of the critical slip
surface is less than 1, the entire slope is then considered
as unstable, or failure.

2.2 Simplified Bishop method of slices

The simplified Bishop method of slices (SBMS)
(Bishop 1955) takes a different assumption that all
inter-slice forces are horizontal. Equilibrium of ver-
tical forces in all slices and the overall moment gives
the following expression for the safety factor:
Figure 2. A demonstrative slope. Ns  ck lk cos(αk ) + (Wk −uk lk cos(αk )) tan(φk ) 
k=1 cos(αk ) + (sin(αk ) tan(φk ))/FSω
FSω = Ns
k=1 Wk sin(αk )
at some values (e.g.: for MCS, they are fixed their
sampled values for each deterministic slope stability (5)
analysis). The goal is to determining the safety factor
FSω of a slip surface ω. A very simplified method that The safety factor FSω can be found by iteratively
assumes no interacting forces between slices can be solving (5). Again, many trial surfaces are randomly
used to compute the safety factor: generated, and the trail surface with the smallest safety
factor is the critical slip surface. If the safety factor of

Ns
the critical slip surface is less than 1, the entire slope
FSω = [ck lk + (Wk cos(αk ) − uk lk ) is then considered failed.
k=1


Ns
2.3 Transformation to standard Gaussian space
× tan(φk )] Wk sin(αk ) (4)
k=1 Previously, it is assumed that all uncertain variables
are transformed to the standard Gaussian input space
where ck and φk are the cohesion and friction angle of for the ease of presentation. To demonstrate how the
the k-th slice; Wk is the total weight of the k-th slice; expressions of safety factors can be transformed into
uk is the pore water pressure at the middle point of the the standard Gaussian space, consider the example in
slice bottom; lk is the length of the slice bottom; αk is Figure 2, where the trial surface is ω, both soil layers
the inclination angle of the slice bottom. Equation (4) are homogeneous, and the c, φ and γ parameters for
is obtained based on the equilibrium equation of the both layers are uncertain and independent. Note that
overall moment for all slices about a chosen point. the first few slices share the same shear strengths and
This method of determining the safety factor is called unit weights, and similar for the last few slices. In the
the ordinary method of slices (OMS) (Fellenius 1936). case that the OMS is taken, the safety factor can be
expressed as:

⎡ Ns(1)   s(1) (1)  s(1) (1) (1)  ⎤

c(1) lk(1) + tan(φ (1) ) · γ (1) Nk=1 (hk cos(αk(1) )) − Nk=1 uk lk
⎣ k=1
 s(2) (2)   s(2) (2)  s(2) (2) (2)  ⎦
+ c(2) Nk=1 lk + tan(φ (2) ) · γ (2) Nk=1 (hk cos(αk(2) )) − Nk=1 uk lk
FSω =  (1)  (2)
(6)
γ (1) Nk=1
s
h(1) (1)
k sin(αk ) + γ
(2) Ns (2) (2)
k=1 hk sin(αk )

The OMS is the simplest method among all methods where c(n) , φ (n) and γ (n) are the cohesion, friction
of slices. It does not satisfy force and moment equi- angle and (average) unit weight of the n-th soil layer;
librium of individual slices, and it usually provides Ns(n) is the total number of slices whose bottom sides

641
are within the n-th layer; h(n)
k is the mid-height of the Gaussian PDF, is quite far from the failure region
k-th slice in the n-th soil layer. ∪ω∈ Fω , especially when the actual failure probability
Now let P be the mapping from a standard Gaus- is small. The consequence is that MCS requires many
sian random variable to the variable of interest, e.g.: if samples before a failure sample is obtained. The basic
c(1) is Gaussian with mean = μ(1) c and standard idea of the importance sampling method is to adopt
deviation = σc(1) , one can verify a shifted version of f (z) as the so-called importance
sampling PDF (IS PDF) q(z). That is,
c(1) = pc(1) (Zc(1) ) = μc(1) + σc(1) · Zc(1) (7)  
1 1
q(z) = √ n exp − (z − z  )T (z − z  ) (11)
2π 2
where Zc(1) is standard Gaussian; if c(1) is lognormal
with mean = μ(1) c and c.o.v. = δc(1) , where n is the dimension of Z; z  is the center of the IS
    PDF, which is chosen by the analyst. The IS technique
pc(1) (Zc(1) ) = exp ln μc(1) / 1 + δc2(1) is based on the following observation:

   
+ ln(1 + δc2(1) ) · Zc(1) (8) P(F) = E I min FSω (Z) < 1
ω∈
   
if c(1) is uniform with lower bound = lc(1) and upper = I min FSω (z) < 1 f (z)/q(z) q(z)dz
ω∈
bound = uc(1) ,  
1 
N
≈ I min FSω (Zi ) < 1 f (Zi )/q(Zi )
pc(1) (Zc(1) ) = lc(1) + (uc(1) − lc(1) ) · (Zc(1) ) (9) N i=1 ω∈

(12)
where  is the cumulative density function (CDF) of
standard Gaussian.
Therefore, (6) can be transformed into the following
standard Gaussian-space expression:

FSω (Z) =
⎡  N (1) (1)   N (1) (1)  N (1) (1) (1)  ⎤
(1)
pc(1) (Zc(1) ) k=1 lk + tan( pφ (1) (Zφ (1) )) · pγ (1) (Zγ (1) )
s
k=1 (hk cos(αk )) −
s
k=1 uk lk
s

⎢ $ %& ' $ %& ' ⎥

⎢ ⎥
⎢ a(1)  (1)
 ⎥
⎢  N (2) (2)  N (2) b(2)  (2) ⎥
⎢ + pc(2) (Zc(2) ) (2) Ns (2) (2) ⎥
k=1 lk + tan( pφ (2) (Zφ (2) )) · pγ (2) (Zγ (2) ) k=1 (hk cos(αk )) − k=1 uk lk
s s
⎣ ⎦
$ %& ' $ %& '
a(2) b(2)
 N (1) (1)
 N (2)
pγ (1) (Zγ (1) ) s
k=1 hk sin(αk(1) ) + pγ (2) (Zγ (2) ) s (2)
k=1 hk sin(αk(2) )
$ %& ' $ %& '
d (1) d (2)

[ pc(1) (Zc(1) )a(1) + pc(2) (Zc(2) )a(2) + tan( pφ (1) (Zφ (1) )) · b(1) + tan( pφ (2) (Zφ (2) )) · b(2) ]
= (10)
pγ (1) (Zγ (1) )d (1) + pγ (2) (Zγ (2) )d (2)

Note that a(1) , a(2) , d (1) and d (2) are deterministic,

while b(1) and b(2) are random. Correlation between where {Zi : i = 1, . . . , N } are independent samples
input standard Gaussian variables, e.g.: correlation from q(z). Therefore, the IS simulation contains the
(1)
between Z(1)
c and Zφ , can also be easily handled.
following steps:
a. Draw Z samples {Zi : i = 1, . . . , N } from q(z).
b. For each sample Zi , conduct a deterministic slope
3 IMPORTANCE SAMPLING TECHNIQUE stability analysis to find the critical slip surface
among all trial surfaces. If the safety factor of the
The inefficiency of MCS is primarily due to the fact most critical surface is less than 1, the slope is
that the main support region of f(z), i.e.: standard considered to fail for that Zi sample.

642
c. Repeat Step 2 for i = 1, . . . , N . The IS estimate of gωOMS
 < 0, the surface fails under the OMS criterion,
the overall failure probability is simply and vice versa. Let us consider the soil parameters
c(1) , c(2) , φ (1) , φ (2) , γ (1) , γ (2) to be uncertain. Note that
  if c(1) , c(2) , tan (φ (1) ), tan (φ (2) ) and γ (1) , γ (2) are
1 
N
P(F) ≈ I min FSω (Zi ) < 1 independent Gaussian, the limit state function in the
N i=1 ω∈
standard Gaussian space can be approximated as
 
1 1
× exp (Zi − z  )T (Zi − z  ) − ZiT Zi gωOMS
 (Z) = (μc(1) + σc(1) Zc(1) )a(1)
2 2
+ (μc(2) + σc(2) Zc(2) )a(2)
≡ PFIS (13)
+ (μtan(φ (1) ) + σtan(φ (1) ) Ztan(φ (1) ) )b̄(1)
Note that PFIS
is an unbiased estimator of the actual
overall failure probability regardless the choice of z  . + (μtan(φ (2) ) + σtan(φ (2) ) Ztan(φ (2) ) )b̄(2)
More importantly, if z  is carefully chosen so that the
− (μγ (1) + σγ (1) Zγ (1) )d (1)
main support region of q(z) is close to the failure
region ∪ω∈ Fω , the c.o.v. of PFIS can be quite small. − (μγ (2) + σγ (2) Zγ (2) )d (2) (16)
The c.o.v. of the estimator PFIS is simply

δ(PFIS ) = Var(PFIS )/E(PFIS )
)
*
*1  N      2 
+ 1  
1 T
≈ I min FSω (Zi ) < 1 · exp (Zi − z ) (Zi − z ) − Zi Zi − PF
T IS
PFIS (14)
N i=1 ω∈ 2 2

3.1 Choice of z 
Although PFIS is unbiased regardless the choice of z  , where μ and σ are the mean value and standard
the choice of z  may seriously affect the efficiency of deviation of the subscripted variable;
the IS estimator. An ideal choice of z  is simply the
(n) (n)
point on the failure limit state surface that is closest to 
Ns 
Ns
the origin, i.e.: the solution of (3), denoted by z ∗ (see b̄(n) = μγ (n) (h(n) (n)
k cos(αk )) − uk(n) lk(n) (17)
Figure 1). However, finding z ∗ is itself a challenging k=1 k=1
task. Nonetheless, if z  is taken to be a point that is
easy to obtain and yet is close to z ∗ , the resulting IS Note that (16) is an approximation of (15) because
estimator can still be satisfactory. b(1) and b(2) are in fact random, but we have delib-
In the following, an analytical way of finding such erately replaced them by b̄(1) and b̄(2) . By doing so,
a point that is based on the OMS is presented. Among (16) is now a linear function of the standard Gaussian
the slope stability analysis methods, the OMS seems input Z. It is soon clear that the OMS design point in
most linear. Moreover, it does not require iterative cal- the standard Gaussian space is roughly
culations. Therefore, it is proposed in this research to
 
implement the OMS to find the approximate design ∗ ∗ ∗ ∗ ∗ ∗
zcOMS zcOMS OMS
ztan(φ OMS OMS
(1) ) ztan(φ (2) ) zγ (1) zγOMS
point z  . Again, taking the slope in Figure 2 as an (1) (2) (2)

example, let the slip surface ω be a ‘‘representative’’ β OMS

slip surface, e.g.: the critical slip surface when all = · [−a(1) σc(1) − a(2) σc(2) − b(1) σtan(φ (1) )
uncertain soil parameters are fixed at their expected σgωOMS
values.
− b(2) σtan(φ (2) ) d (1) σγ (1) d (2) σγ (2) ]
One can re-write (4) as the following OMS limit
state equation: (18)

gωOMS
 = c(1) a(1) + tan(φ (1) ) · b(1) + c(2) a(2) where
(2) (2) (1) (1) (2) (2)
+ tan(φ )·b −γ d −γ d (15)
a(1) μc(1) + a(2) μc(2) + b(1) μtan(φ (1) )
+ b(2) μtan(φ (2) ) − d (1) μγ (1) − d (2) μγ (2)
where gωOMS
 is the OMS limit state function for the β OMS =
failure event of the representative slip surface ω : if σgωOMS

643
 )
* (1)2 2 b. Use (20) to obtain the design point z  for the
*a σ (1) + a(2)2 σ 2(2) + b(1)2 σ 2 (1) representative surface ω .
σgωOMS = + c c tan(φ )
+b(2) σtan(φ (1) 2
σγ (1) + d (2) σγ2(2) c. Draw Z samples {Zi : i = 1, . . . , N } from q(z), a
2 2 2
(2) ) + d
2
Gaussian PDF centered at z  with unit covariance
(19)
matrix.
d. For each sample Zi , conduct a deterministic slope
The above design point is in the standard Gaussian stability analysis (not necessarily the OMS) to find
space of c(1) , c(2) , tan(φ(1) ), tan(φ(2) ) and γ (1) , γ (2) . the critical slip surface among all trial surfaces. If
A simple algebraic operation shows that the the safety factor of the critical surface is less than
design point in the standard Gaussian space of 1, the slope is considered to fail for that Zi sample.
c(1) , c(2) , φ (1) , φ (2) and γ (1) , γ (2) is
⎡ ∗ ⎤ ⎡ ⎤
zcOMS
(1) −a(1) σc(1) β OMS /σgωOMS
⎢ ⎥ ⎢ ⎥
⎢ zcOMS
∗
⎥ ⎢ −a(2) σc(2) β OMS /σgωOMS ⎥
⎢ (2) ⎥ ⎢ ⎥
⎢ ⎥ ⎢ ⎥
⎢ ∗ ⎥ ⎢ p−1 (tan −1
(μ (1) 2
OMS )) ⎥
zφOMS (1) ) − b σ β /σ
OMS
⎢ (1) ⎥ ⎢ φ (1) tan(φ (1)
tan(φ ) g ⎥
⎢ ⎥= ⎢ ω
⎥ (20)
⎢ ∗ ⎥ ⎢ ⎥
⎢ zφOMS
(2) ⎥ ⎢ p−1 (tan −1
(μ (2)
tan(φ ) − b (2) 2
σ β OMS
/σ OMS )) ⎥
⎢ ⎥ ⎢ φ (2) tan(φ (2) ) gω
⎥
⎢ ∗ ⎥ ⎢ ⎥
⎢ zγOMS
(1) ⎥ ⎣ d (1) σγ (1) β OMS /σgωOMS ⎦
⎣ ⎦
∗
zγOMS
(2) d (2) σγ (2) β OMS /σgωOMS

where pφ (n) is the mapping from a standard Gaussian e. Repeat Step 2 for i = 1, . . . , N . The IS estimate of
random variable to φ (n) . In this study, the design point the overall failure probability is simply
defined in (20) will be taken as the center z  of the  
1 
N
IS PDF. P(F) ≈ I min FSω (Zi ) < 1
Note that the derivations of the design point in N i=1 ω∈
(20) are based on the following assumptions: (a)
c(1) , c(2) , tan(φ(1) ), tan(φ(2) ) and γ (1) , γ (2) are inde-  
1  T  1 T
pendent Gaussian and (b) the OMS is the adopted × exp (Zi − z ) (Zi − z ) − Zi Zi
method of slices. Those assumptions are taken because 2 2
the resulting design point z  has an analytical form.
Clearly, such a design point z  in general is not the ≡ PFIS (21)
actual design point z  because for the actual applica-
tion those assumptions may not be true, i.e.: the actual
adopted slope stability method may not be the OMS 5 NUMERICAL EXAMPLE
and the soil parameters may not be independent Gaus-
sian. Nonetheless, because the proposed IS approach Consider the following example extracted from the
works regardless the choice of z  , the resulting esti- STABL user manual (Siegel 1975): the slope in
mator PFIS is always unbiased. The only concern is Figure 3 underlain by a rock layer. The shear strength
whether the approximate design point z  is close to parameters c and φ of the soil are uncertain. It
the actual design point z ∗ . Empirically, it is found that is assumed that tension cracks are present near the
the approximate design point z  is usually close to the top surface of the slope up to a depth of 11 m
actual design point z ∗ even the two assumptions are and that the failure surface cannot propagate into
violated in the actual application. the rock layer. The uncertain cohesion c is lognor-
mally distributed with mean μc = 76.3 kN/m2 and
c.o.v. δc = 20%, while the friction angle is Gaus-
sian with mean μφ = 18◦ and standard deviation
4 PROCEDURE OF PROPOSED APPROACH
σφ = 1.8◦ . There is a negative correlation of −0.3
between Zc and Zφ . The simplified Bishop method of
The procedure of the proposed IS method is described
slices is taken as the slope stability method for this
as follows:
example.
a. Execute an OMS analysis where all uncertain soil The MCS method with sample size N = 100,000 is
parameters are fixed at their mean values to find taken to estimate the failure probability of the slope.
the representation slip surface ω . For each Z sample, one hundred trial surfaces are

644
 from q(z) to obtain PFIS via (13). Those samples are
plotted in Figure 4. It is clear that a large portion of
the samples are failure samples. The resulting PFIS
is 2.0e−4 , and its c.o.v. is estimated to be 26.1%
via (14). It is clear that the MCS method becomes
inefficient when the failure probability gets smaller,
while the IS method seems robust on that aspect. In
particular, with a total number of 100 samples, the IS
method performs satisfactorily even for small failure
Figure 3. The slope considered in the numerical example. probability.

6 CONCLUSIONS

A new method based on the importance sampling (IS)

technique is proposed to efficiently estimate failure
probability of slope stability. The center of the impor-
tance sampling probability density function (PDF) can
be found by a first-order reliability analysis based on
the ordinary method of slices. From the results of
the numerical example, it is found that the proposed
IS method is more efficient than the Monte Carlo
Figure 4. The MCS (left) and IS (right) samples in the stan- simulation (MCS). The proposed IS method provides
dard Gaussian space. The shaded region is failure region F. unbiased estimates for slope failure probabilities, and
yet it requires much less computation than MCS espe-
cially for the cases with small failure probabilities.
A limitation of the proposed IS method is that the
randomly generated to form the set of all trial slip dimension of uncertain soil parameters cannot be too
surfaces  (shown in Figure 3), and the factor of safety large due to the dimension limitation of the basic IS
is determined by solving (5), then (1) can be employed method. Therefore, the proposed method is not suit-
to find PFMCS . Figure 4 shows the locations of all MCS able for determining failure probability of slopes with
samples in the standard Gaussian space, where a dot spatial variability. It is left as a future research to
indicates a non-failure sample, while a ‘x’ indicates overcome this limitation.
a failure sample. One can see most of the samples
are non-failure ones. The analysis shows that PFMCS is
2.4e−4 and the c.o.v. of the estimator is 20.4%. REFERENCES
As a verification, the actual failure region F is found
by an exhaustive analysis described as follows. First Ang, A.H.S. and Tang, W.H. 1984. Probability Concepts in
divide the standard Gaussian space into dense grid Engineering Planning and Design, Volume II Decision,
points. At each grid point, a simplified Bishop stability Risk, and Reliability, John Wiley and Sons.
analysis is taken to see if the entire slope fails. Do so Bishop, A.W. 1955. The use of the slip circle in the stability
for all grid points to obtain the actual failure region F. analysis of slopes. Geotechnique, 5, 7–17.
Chowdhury, R.N. and Xu, D.W. 1993. Rational polynomial
The actual design point z ∗ can be readily identified in technique in slope-reliability analysis. ASCE Journal of
Figure 4 as the point on the failure boundary that is Geotechnical Engineering, 119(12), 1910–1928.
closet to the origin. Christian, J.T., Ladd, C.C. and Baecher, G.B. 1994. Relia-
For the IS technique, the representative surface ω . bility applied to slope stability analysis. ASCE Journal of
(shown in Figure 3) is first found by a single OMS sta- Geotechnical Engineering, 120(12), 2180–2207.
bility analysis where c and φ are fixed at their mean Fellenius, W. 1936. Calculation of the stability of earth
values. The approximate design point z  for the rep- dams. Transactions of 2nd Congress on Large Dams,
resentative surface ω . is determined from (20), and Washington, DC, 4, 445–462.
its location is shown in Figure 4. It is clear that z  is Griffiths, D.V. and Fenton, G.A. 2004. Probabilistic slope
stability analysis by finite elements. ASCE Journal
fairly close to the actual design point z ∗ even though of Geotechnical and Geoenvironmental Engineering,
the actual analysis method is the simplified Bishop 130(5), 507–518.
method, not the OMS, and even though in reality c Low, B.K. and Tang, W.H. 1997. Probabilistic slope analysis
and tan(φ) are not independent Gaussian. Following using Janbu’s generalized method of slices. Computers
the IS procedure, only 100 samples of Z are drawn and Geotechnics, 21(2), 121–142.

645
Low, B.K., Gilbert, R.B. and Wright, S.G. 1998. Slope Rubinstein, R.Y. 1981. Simulation and the Monte-Carlo
reliability analysis using generalized method of slices. Method, John Wiley & Sons Inc., New York.
ASCE Journal of Geotechnical and Geoenvironmental Shinozuka, M. 1983. Basic analysis of structural safety,
Engineering, 124(4), 350–362. ASCE Journal of Structural Engineering, 109, 721–740.
Melchers, R.E. 1989. Importance sampling in structural Siegel, R.A. 1975. STABL User Manual, Joint Highway
systems, Structural Safety, 6, 3–10. Research Project 75-9, School of Engineering, Purdue
Oka, Y. and Wu, T.H. 1990. System reliability of slope stabil- University, West Lafayette, Indiana.
ity. ASCE Journal of Geotechnical Engineering, 116(8),
1185–1189.

646
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Prediction of the flow-like movements of Tessina landslide by SPH model

S. Cola & N. Calabrò

Dept. IMAGE, University of Padova, Italy

M. Pastor
Centro de Estudios y Esperimentación de Obra Públicas (CEDEX) and ETS de Ingenieros de Caminos, Madrid, Spain

ABSTRACT: The SPH (Smoothed particle hydro-dynamics) method is a powerful tool for modelling
debris/mud flows, which can be described in terms of local interactions of their constituent parts. The char-
acteristics of the 1992 Tessina landslide make it appropriate for modelling by means of SPH method. In this
paper we will investigate the evolution and dynamics of the 1992 Tessina landslide and present future hazard
scenarios based on potential volumes of masses mobilised in the future.

1 INTRODUCTION 1994): that of Tessina one, being present for many

decades in North-East Italy. Model calibration is
Over the past decades, flow-like landslides (debris- obtained from the simulation of the 1992 event and,
flow and mud-flow) have been intensively studied, subsequently, used in forecasting flow expansion
because in their catastrophic form they have caused during a possible future reactivation.
many victims and serious economic damage around
the world. Their distinctive features and evolution are
strictly related to the mechanical and rheological pro- 2 MATHEMATICAL MODEL
perties of the materials involved which are responsible
for the long distances travelled and high velocities in The mathematical model is presented starting with the
some cases reached. Biot—Zienkiewicz equations for non-linear materials
Lately a number of new numerical models have and large deformation problems, from where the depth
been developed: the prediction of both run-out dis- integrated model can be derived.
tances and velocity can notably reduce losses inferred It is based on:
by these phenomena, providing a means for defining
and estimating the hazardous areas, and the working i. the balance of mass, which is:
out of appropriate design measures.
Most of the current numerical methods are based Ds ρ
+ ρdivvs = 0 (1)
on either structured (finite elements) or unstruc- Dt
tured (finite elements and volume) grids. An alter-
native approach is that of Cellular Automata (CA) where Ds refers to a material derivative following
which is appropriate in modelling debris/mud flow the soil particles, ρ is the density of the soil and vs
(Avolio et al., 2000). But recently another interesting is the velocity of soil skeleton;
and powerful tool for modelling natural systems has ii. the balance of linear momentum, which reads:
been provided by a new group of ‘‘meshless’’ numeri-
cal methods, among which the method of ‘‘Smoothed Ds vs
particle hydro-dynamics’’ (SPH) independently intro- ρ = ρb + divσ (2)
Dt
duced by Lucy (1977) and Gingold and Monaghan
(1977) and firstly applied to astrophysical modelling where b are the body forces and σ the Cauchy stress
(Benz W., 1990). This method relies on nodes (points) tensor.
to approximate functions or derivates, discluding any
element based information. The model is completed by suitable rheological
In this paper, an in-depth integrated SPH model and kinematical relations relating the stress tensor
(Pastor et al, 2001) is proposed to simulate the prop- to the rate of deformation tensor d, and the rate of
agation stage of a slow mud-flow landslide (Takeda, deformation tensor to the velocity field vs .

647
 Many landslides have average depths which are appreciable fraction of fine and plastic material whose
small in comparison with their length or width, so it behaviour may be controlled prevalently by viscosity.
is possible to simplify the 3D propagation model by The shear stress inside a Bingham fluid is depicted
integrating its equations along the vertical axis. The by the equation:
resulting 2D depth integrated model presents an excel-
lent combination of accuracy and simplicity, providing du
important information such as velocity of propagation, τ = τc + μb (5)
dy
time in reaching a particular place, depth of the flow
at a certain location, etc.
Depth integrated models have been frequently used where τ and τc are the shear stress and the yield
in the past to model flow-like landslides. It is worth strength respectively, μb the plastic viscosity and du/dy
mentioning the pioneering work of Hutter and his the shear rate.
co-workers (Savage and Hutter, 1991; Hutter and Koch
1991) or Laigle and Coussot (1997).
The equations for the depth averaged model are 3 TESSINA LANDSLIDE
obtained by integrating the balance of mass and
momentum equations along the vertical axis. The The Tessina landslide, which was first triggered in
balance of mass results: October 1960, is a complex movement with a source
area affected in the upper sector by rotational slides;
∂ ∂h downhill the slide turns into a mud flow through a
(hv̄j ) + =0 j = 1, 2 (3)
∂xj ∂t steep channel. The landslide develops in the Tessina
valley between the altitudes of 1220 m and 625 m
where h is landslide thickness and v̄j depth averaged a.m.s.l, with total longitudinal extension of nearly 3 km
velocity. and maximum width of about 500 m. The mud flow
Indeed, the balance of momentum, depth integrated passes very close to the village of Funes and stretches
equation is: downhill as far as the village of Lamosano (Figure 1).
The rock types involved in the landslide belong to the
  Flysch Formation, consisting of a rhythmic alternation
∂ ∂
ρ (hv̄j ) + (hv̄i v̄j ) of marly—argillaceous and calcarenite layers about
∂t ∂xj 1000–1200 m thick.
∂ During the 1960s, several reactivations, involving
= (hσ̄ij ) + tjA + tjB + ρbj h j = 1, 2 (4) about 5 million m3 of material, occurred causing the
∂xj
filling of the Tessina valley with displaced material
where the terms tjA and tjB are the normal stress acting
on the surface and bottom respectively.
It is important to note that the above results depend
on the rheological model chosen, from which basal
friction and depth integrated stress σ̄i,j are obtained.
In fact it is assumed that the flow at a given point
and time, with depth and in-depth averaged velocities
known, has the same vertical structure as a uniform
steady state flow. In the case of flow—like landslides
this model is often referred to as the infinite landslide
as it is assumed to have constant thickness and move
at constant velocity along a constant slope.
The main advantage of this numerical method is
a big reduction of the calculation time compared to
a standard finite element code formulated on the
Eulerian approach, because the computational grid
is separated from the structured terrain mesh used to
describe terrain topography.
In the first application of the model to the Tessina
phenomenon, the Bingham rheological model was
considered sufficiently appropriate to describe the
main features of landslide mass behaviour. This
is because the soil involved in the slide contains Figure 1. Distribution area of the Tessina landslide.

648
30–50 m thick. These movements seriously endan- The soil may be classified as a tout—venant with
gered the village of Funes, which is situated on a steep a percentage of clay and silt ranges from 5 to 20%
ridge—originally quite high above the river bed, but (Figure 2). The fine fraction is composed prevalently
now at nearly the same level as the mud-flow surface of medium-low plasticity inorganic clay (Figure 3).
(Dall’Olio et al, 1987). In order to evaluate the rheological behaviour of
During April 1992 a rotational slide with 20–30 m the soil composing the mud-flow, some tests with the
deep failure surface, affecting also the Flysch bedrock,
caused the collapse of an area 40.000 m2 wide, on the
left hand-side of the Tessina stream, with an approxi-
mate volume of 1 million m3 . The movements initially
caused the formation of 15 m high scarp, a 100 m dis-
placement of all the unstable mass and the destruction
CH
of drainage systems set up some years earlier, but they
continued with a certain intensity up to June 1992,
causing the mobilization of another 30.000 m2 more. CL
The highly fractured and dismembered material from
this area was channelled along the river bed where, OH
MH
due to continuous remoulding and increasing of water
content, it became more and more fluid, thus generat- OL
ML
ing some small earth flows converging into the main
flow body. After these events, the inhabitants of Funes
and Lamosano were evacuated.
After this important event, the landslide was not
stabilized but its evolution was kept under control by Figure 3. Classification of mud-flow material.
an accurate monitoring activity and an alarm system
set up. Some other collapses and mudflows, similar
in evolution to the 1992 event but involving smaller
portion of soils, were observed with an occurrence
of about 3 years. Anyway, the hazard level reached e
in 1992 was never overcome and the inhabitants were
evacuated no more.

4 SIMULATION OF THE 1992 TESSINA

LANDSLIDE

4.1 Properties of soils from experimental tests

c
Laboratory analyses were performed on some
remoulded samples collected inside the moved mass Figure 4. Viscosity curve.
along the closing channel near Funes village.

e
Percentage smaller (%)

Particle size (mm)

c
Figure 2. Typical grain size distribution of material
involved in mud-flow. Figure 5. Yield strength curve.

649
 a) t = 0 sec. b) t = 80 sec.

Figure 6. Simulation of the 1992 event obtained by SPH.

FANN V-G rheometer were performed on the fine In Figures 4 and 5 the best fitting equations are
fraction (passing to n.40 sieve). In accordance with the also reported. For the concentration analyzed the yield
facilities of this equipment, the strength was recorded strength and plastic viscosity range from 5 to 40 Pa and
at two speeds—i.e. at 300 and 600 r/min. From these from 0.005 to 0.08 Pa·s respectively.
two readings the Bingham viscosity and the yield
strength were determined.
The measurements were repeated with a solid con- 4.2 Numerical modelling
centration varying from 20 to 45%: Figures 4 and 5
sketch the experimental data obtained for sample n.3. In order to simulate the 1992 event and calibrate model
The rheological properties depend on solid con- parameters, the flow-like movements of slide were
centration cv according to the following exponential reproduced assuming a topographic mesh of 17464
laws (O’ Brien and Julien, 1988; Major and Pierson, nodes provided by a 15 × 15 m digital terrain model
1992; Coussot and Piau, 1994, Sosio et al, 2006): (DTM). Figure 6a shows the topographical base and
initial position of the failed mass, the latter being
modelled through 3304 particles with average spaces
of 3.5 m.
τc = αeβcv It is important to note that the analysis with SPH
(6) code doesn’t account for the triggering of the land-
μb = γ eδcv slide, when large portions of slope, including rock

650
Table 1. Assumed rheological properties. case of detachment. The detachment location may be
anywhere in the superior basin where the monitoring
Density (Kg/m3 ) 1700 has marked the highest movements: in this case, the
Solid concentration (%) 50 most critical area seems to be the zone known as ‘‘Pian
Bingham viscosity (Pa s) 600 de Cice’’ (Figure 7), where maximum displacements
Yield strength (Pa) 5200
of about 3 cm/ year were recorded in the last decade.
On the basis of these remarks, three different cases
were simulated, concerning the collapse of 0.5, 1 and
1.5 million m3 of material. To note that the value of
blocks coupled with debris and mud, move. This ini- 1 million m3 is more than that experienced in all the
tial phase exhausts itself quickly in comparison to the previous movements.
duration of the phenomenon, because thereafter such Also in this case the material is described by the
masses split continuously until reaching complete flu- Bingham rheological model, using the constitutive
idity. The numerical simulation considered that, the parameters obtained from the previous calibration.
whole mass involved in the landslide detaches instan- The simulation results, shown in Figure 8, under-
taneously. In reality the cohesion inside the unstable line how the mud path is similar to that of the 1992
mass is lost little by little: in future advances of this event. More, even if the involved volume changes,
model, this aspect would be taken into account by the maximum distance reached for the three cases are
assuming initial high viscosity of unstable mass and more and less equivalent because it is effected mainly
decreasing it gradually up to a complete fluidity state. by the topographic gradient, which is the same in all
The 1992 landslide was simulated assuming the three analyses.
Bingham constitutive model with viscosity coeffi- Velocities and thickness are different for the three
cients corresponding to a liquid state of the tested soils, cases (Table 2), the highest value resulting in the
as summarized in Table 1. simulation of 1.5 million m3 , that is, therefore, the
The results of the performed simulation are shown most hazardous case for Lamosano and Funes villages.
in Figure 6. In particular, Figure 6b shows the landslide Although the model can’t analyse recurrent collapses,
at the first simulation step (the elapsed time equal to and the real path length can’t be estimated exactly,
80 s), that means the masses have just moved away and it is possible to suppose that a detachment from this
the mudflow begins to channel correctly into the river area could involve downhill villages with a very high
valley. Simulations at the elapsed time of 200 s and risk, resulting in a necessary partial or even totally
1000 s (Figures 6c and 6d) show how the mudflow has evacuation.
canalized in the river valley and here runs correctly.
Comparison between the real event and this sim-
ulation shows substantial agreement in the landslide
development, even if the numerical analysis gives run-
out distances shorter than those observed. This result
depends on the fact that the SPH code is unable to
reproduce recurrent detachments of mass and, without
the contribution of other material moving later from
the upper zone, the flow stops earlier than in reality.
But, nevertheless, the mud path is properly identified
and the velocities (4 ÷ 5 mm/s) and thicknesses of
the landslide (6 ÷ 7 m) are in agreement with field
observations.

5 FUTURE HAZARD SCENARIOUS

The SPH code permitted to develop reliable simula-

tions of future events. To this aim, an accurate and Figure 7. Localization of Pian de Cice hazardous area.
update morphology of the landslide area has been
set up using the most updated DTM. As regards the
possible instable volume, on the basis of previously Table 2. Results for future hazardous scenarios simulation.
observed events, Avolio et al. (2000) indicated an
Volume (m3 ) 0.5 1 1.5
interval from 0.5 to 1.5 million on cubic meters:
Average velocity (mm/s) 2–4 4–6 9–10
these limit values represent the minimum mud quan- Average thickness (m) 11–10 8–11 7–10
tity needed to obtain a landslide event and the worst

651
 3 3 3

Figure 8. Simulation of the future scenarios due to mobilization of Pian de Cice area.

6 CONCLUSIONS Angelini’’ of Belluno and the CNR-IRPI of Padova

which covenant this activity research.
Results of first simulations using the depth averaged
model developed on the SPH approach are promis-
ing. Despite the limitations existing in representing
the trigger of the rotational superior slides and the REFERENCES
recurrent collapse of different area with this model,
Avolio, M.V., Di Gregorio, S., Mantovani, F., Pasuto, A.,
the reproduction of landslide path for the 1992 event
Rongo, R., Silvano, S. & Spataro, W., 2000. Simulation of
was appreciable, supporting the assumption that the the 1992 Tessina landslide by cellular automata model and
development of the initial part doesn’t influence the future hazard scenarios, Geomorphology, 2(1): 41–50.
simulation of the subsequently run-out phenomenon. Benz, W., 1990. Smooth particle hydrodynamics: a review,
On the basis of this positive result, possible future in The Numerical Modelling of Nonlinear Stellar Pulsa-
events, of a magnitude comparable or higher than tations, pp. 269–288, Buchler (Ed), Kluwer Academic
the previous slide, are simulated, hypothesizing that Publishers.
physical characteristics of the landslide don’t change Coussot, P., 1997. Mudflow Rheology and Dynamics. IAHR
significantly over time. The results obtained show a monograph, Balkema, Rotterdam, 260 pp.
Coussot, P. & Piau, J.M., 1994. On the behaviour of fine mud
situation of real risk affecting the hamlets involved in
suspensions. Rheological Acta 33, 175–184.
the landslide and underline the need to take adequate Dall’Olio, L., Ghirotti, M., Iliceto, V. & Semenza, E., 1987.
countermeasures. La frana del Tessina, Alpago (Bl). Atti VI Congresso
The experience acquired through the Tessina land- Ordine. Nazionale Geologi, Venezia 25–27 settembre
slide is useful for better identification and definition 1987, pp. 275–293.
of potential applications of this procedure in order to Gingold, R.A. & Monaghan J.J., 1977. Smoothed particle
forecast areas that may be involved in low—viscous hydrodynamics: theory and application to non-spherical
landslides, such mud flows. The SPH model may be stars. Monthly Notices of the Royal Astronomical Society.
successfully used in order to follow the progress of 81, pp. 375–389.
Gingold, R.A. & Monaghan J.J., 1982. Kernel estimates as
an event, predict its evolution and, finally verify the
a basis for general particle methods in hydrodynamics.
possible effects of flows on man’s intervention. J. Comput. Phys. 46.
Hutter, K. & Koch, T., 1991. Motion of a granular avalanche
in an exponentially curved chute: experiments and theo-
rical predictions. Phil. Trans. R. Soc. London, 334, pp.
ACKNOWLEDGEMENTS 93–138.
Laigle, D. & Coussot, P., 1977. Numerical modeling of mud-
This paper is a part of the research Project ‘‘MoVeMit: flows. Journal of Hydraulic Engineering, ASCE, 123(7),
pp. 617–623.
studio e modellazione di movimenti di versante finali- Lucy, L.B. 1977. A numerical approach to testing of fusion
zzati alla formulazione di interventi di mitigazione’’, process. Astronomical Journal, Vol. 82, pp. 1013–1024.
financed by the funding of the ‘‘Fondazione Cassa di Major, J.J. & Pierson, T.C., 1992. Debris flow rheology:
Risparmio di Verona, Vicenza, Belluno e Mantova. experimental analysis of fine-grained slurries. Water
At least the authors wish to thank the ‘‘Fondazione Resources Research 28, 841–857.

652
Monaghan J.J. & Gingold, R.A., 1981. Shock simulation by O’Brien, J.S. & Julien, P.Y., 1988. Laboratory analysis of
the particle method SPH. J. Comp. Phys. 52, pp. 374–389. mudflow properties. Journal of Hydraulic Engineering,
Pastor, M., Quecedo, M, Merodo, J.A., Herreros, M.I., 110, 877–887.
González, E. & Mira, P., 2002. Modelling of debris flows Savage S.B. & Hutter, K., 1991. The dynamics of avalanches
and flow slides. Revue française de genie civil: Numerical og granular materials from initiation to runout. Part I:
modelling in Geomechanics, vol 6, 1213–1232. Analysis. Acta Mechanica 86, pp 201–223, Madrid
Pastor, M., Sopeña, L., Fernández Merodo, J.A., 2001.
Quecedo, M., Mira, P., Herreros, M.I. & González, E., Sosio, R., Crosta, G.B., Frattini, P. & Valbuzzi, E., 2006.
2001. Modelización de deslizamientos de laderas y Caratterizzazione reologica e modellazione numerica di
olasprovocadas en embalses: aplicaciones prácticas. V un Debris flow in ambiente alpino, Giornale di geologia
Simposio Nacional sobre Taludes y Laderas Inestables, applicata, v. 3.
Madrid, v.2, pp. 431–460. Takeda, H.T., Miyama, S.M. & Sekiya, M., 1994. Numer-
O’Brien, J.S., Julien, P.Y. & Fullerton, W.T., 1993. Two ical simulation of viscous flow by smooth particle
dimensional water flood and mudflow simulation. Journal hydrodynamics, Progress of Theoretical Physics, V.92,
of Hydraulic engineering, 119, 244–259. 939–960.

653
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Applications of the strength reduction finite element method to a gravity

dam stability analysis

Q.W. Duan, Z.Y. Chen, Y.J. Wang, J. Yang, & Y. Shao

China Institute of Water Resources and Hydropower Research, Beijing, China

ABSTRACT: This paper examines an example that includes gravity dam stability analysis by using the strength
reduction finite element method (SRF), compared with the analytical results of the limit equilibrium approach
proposed by Sarma. Extensions of plastic zones along the specified seam, as well as the development of new
yield zone in the intact rock, can be clearly visualized as the reduction factor F increases, until a failure criterion,
either defined as divergence of the numerical iteration, or rapid increase of displacement at some characteristic
points, is reached. The associated values of F are in good agreement with those of Sarma’s method for two cases
investigated. It shows that strength reduction method is feasible in gravity dam stability analysis along specified
or potential slip surfaces.

1 INTRODUCTION displacement rate approaches infinity. This definition

for safety factor is the same as that proposed in the
During the past ten years, finite element analysis using limit equilibrium method. In the analysis, F, the
the strength reduction approaches (SRF) has been strength reduction factor, is gradually increased until
widely used in engineering (Griffths & Lane 1999, the structure reaches instability.
Ugai & Leshchinsky, 1995, Matsui & San, 1992).
It considers nonlinear constitutive relations of mate- c = c/F (1)
rials and complicated geological conditions of the 
rock mass. Various complicated loads, such as water- tan φ = tan φ/F (2)
storing, seepage, and earthquake can be included.
The analytical results include the stress region, spatial where c and φ are the strength parameters of the
distribution of deformation, the development process material.
of plastic zone etc., which are much more informa- Based on the shear strength reduction concept by
tive, compared to the conventional limit equilibrium Zienkiewicz et al. (1975) and the elastic-plastic finite
methods. element method, many domestic and foreign schol-
This paper studies the feasibility of strength reduc- ars developed a variety of research outcomes that
tion method in the deep anti-sliding stability of gravity have confirmed that SRF can always give calcu-
dams. The results obtained by this study are compared lated factor of safety almost identical to that given
with those obtained by the limit equilibrium method. by the conventional limit equilibrium methods. In
this paper, we present a study that concerns the sta-
bility of gravity dam along potential slip surfaces.
The analytical results again show good agreements
with the conventional approach that uses Sarma’s
2 THE STRENGTH REDUCTION METHOD method.
Strength reduction finite element method (SRF) is
firstly presented by Zienkiewicz in 1975 in the applica-
3 THE TEST EXAMPLES
tion of elastic-plastic by the Finite Element Analysis.
The reduction coefficient is defined as a ratio of the
3.1 The gravity dam example
available shear strength of the soil within slope to the
actual shear stress developed when the limiting equi- A gravity dam with height of 129 m is shown in
librium state is believed happened, either based on Figure 1(a), which contains two soft layers AD and
the criterion that the numerical iteration fails or the BC intersecting at point B. The inclination of AD is

655
 (a)

(b)

Figure 1. An example of gravity dam stability analysis. (a) Double soft layers, (b) Single soft layer.

Table 1. Mechanical parameters of the dam and abutment.

Deformation Tensile
Modulus Poisson’s Friction Cohsion strength Density
Material (GPa) ratio angle (MPa) (MPa) kN/m3

Dam body 20 0.167 35 2.0 1.85 24

Bedrock 10 0.26 35 1.3 0 27
Layers 2.5 0.35 19.8 0.115 0 25.6

15◦ downstream, inclination of BC is 10◦ upstream.

Figure 1(b) shows the dam containing only a weak
layer AD as described in Figure 1(a). Water thrust of
128 m height and the water pressure at the ground
surface of upstream base are considered. The param-
eters of dam body and dam base are shown in Table 1.
Mohr-Coulomb yield criterion is adopted for the
Figure 2. Analysis by Sarma’s method, the double-layer
elastic-plastic analysis.
model.

3.2 Calculation results by Sarma method b

Figure 2 shows the calculations for double-layer model
associated with the safety factor of 2.19, with vertical
interfaces.
For the calculation of single layer model illustrated a
in Figure 3, the method of slices with inclined inter-
faces is adopted and the initial and critical sliding Figure 3. Analysis by Sarma’s method, the single-layer
surfaces of calculation results are shown respectively. model.
The factors of safety with regard to initial and critical NOTE: The arrows ‘a’ and ‘b’ refer to the initial and critical
sliding surfaces are 4.732 and 3.190. slip surfaces respectively.

656
4 THE DOUBLE - LAYER MODEL BY SDF Calculation started with F = 1.0, with a results
shown in Fig. 5(a), followed by F = 1.2, F =
4.1 The iteration process 1.4, . . . , until F = 2.2 (Figure 6), when the iteration
broke.
The safety factor for calculation model shown in
The calculation results have shown that the criti-
Figure 4 is 2.12 by the commercial software FLAC.
cal sliding surface is along the soft layer ABC. In
Its corresponding surface is shown in Figure 4. The
order to study the progressive development of destruc-
result by the authors has good agreement with that
tion of dam base with the reduction of strength, the
obtained by Sarma method.
plastic zones are listed in Figure 5 with regard to differ-
ent reduction coefficients (from 1.0 to 2.0 with equal
intervals of 0.2, 2.12 and 2.20).
As shown in Figure 5, greater portions of the
soft layer become plastic as the reduction coefficient
increases. When the reduction coefficient reaches 2.0,
the whole weak layer is plastic; however, the dam
base has not reached the ultimate Status. When the
reduction coefficient increases to 2.12, the whole weak
layer is plastic, moreover, a large plastic area appears
between the dam site and the turning point, the dam
base has been at the ultimate status.
Figure 6 shows the plastic zone at the moment the
numerical iteration broke, associated with a factor of
Figure 4. Sliding Plane in ultimate State, F = 2.12. safety F = 2.20.

(a) F=1.0 (b) F=1.2

(c) F=1.4 (d) F=1.6

(e) F=1.8 (f) F=2.12

Figure 5. The plastic zones for different reduction coefficients.

657
4.2 The displacement and iteration features
From the reduction factor of distribution of plastic
zone, the relationship can be seen between the increase
in the reduction factor and development process of the
plastic area. So we can come into conclusion that the
appearance of the plastic area is an essential stage from
this process from all the plastic zone transfixion to the
ultimate Status of dam base. Therefore, it is the ulti- Figure 7. Reduction factor and the curve of the horizontal
mate Status when secondary sliding surface appears, displacement of the entrancing.
state as illustrated in Figure 5.
From Figure 5, one may see the gradual develop-
ment of plastic zone on the slip surface as F increases.
At F = 2.12, which corresponds to the solution that
Sarma’s method offered, plastic yielding developed
along the whole seam, and in the meanwhile, a ver-
tical plastic zone developed near the toe. The failure
modes of SRF and Sarma coincide with each other.
Table 2 shows information of the iterative con-
vergence times and horizontal displacement of some Figure 8. Reduction factor and the curve of iterative step.
characteristic points located at the entrance, exit points
of the seam and dam crest (designated A, C, and E
respectively in Figure 1a) associated with the reduction feature points, but the extent that horizontal displace-
factor. Figure 7 is for the curves of the reduction fac- ment increases as reduction factor increases remains
tor and horizontal displacement of the point of entry, the same level, and there is a reduction factor of 2.12,
Figure 8 is for the curves of iterative steps of conver- there exists a mutation and the largest displacement.
gence. From Table 2 we can see, there are different As shown in Figure 7, when rapid development of
displacement takes place, factor amounts to about 2.1,
As shown in Figure 8, with the increase of reduc-
tion factor, the iterative steps in tandem increase,
corresponding mutations range is from 1.8 to 2.12.

5 THE SINGLE - LAYER MODEL BY SDF

5.1 The iteration process

It can be seen from Figure 9 that as the reduction factor
Figure 6. The final stage at which the numerical algorithm increases, yield length of soft layer increases. When
broke (F = 2.20). the reduction factor reaches 1.8, the weak layer is in
plastic state. There appears a small range of plastic area
in the dam, as shows in Figure 9(e). As the reduction
Table 2. Horizontal displacement of feature points in reduc- factor continues to increase, the plastic area expands.
tion factors.
When the reduction factor arrives at the range from
Displacement, cm 2.8 to 3.0, as shown in Figure 9(j) and (k), a plastic
Reduction Iteration zone between the soft layer and the dam developed in
factor, F A C E number a downstream direction.
When the reduction factor increases to 3.15, there
1.0 0.6644 0.1571 0.2521 40920 will appear a long plastic zone which tends to the
1.2 0.6495 0.1699 0.308 52415 upstream of the dam between the earth surface and
1.4 0.6445 0.1882 0.4007 56524 plastic area, which implies the dam base has arrived at
1.6 0.6840 0.2148 0.5592 62013
the ultimate status, as shown in Figure 9(l). The grad-
1.8 0.9165 0.2544 0.8283 67215
2.0 1.643 0.528 1.542 84727 ual development of the plastic zone and related sliding
2.12 15.04 13.41 26.64 114876 surface is noticeable as can be seen in Figure 11.
2.2 203.8 200.1 379.7 140920 Figure 10 shows the plastic zone at the moment the
numerical iteration broke, associated with a factor of
NOTE: The locations of A, C, E are indicated in Figure 1(a). safety F = 3.15.

658
 (a) F=1.0 (b) F=1.2

(c) F=1.4 (d) F=1.6

(e) F=1.8 (f) F=2.0

(g) 2.2 (h) 2.4

(i) 2.6 (j) 2.8

(k) 3.0 (l) 3.15

Figure 9. The plastic zones for different reduction coefficients.

659
 Figure 13. The reduction factor and horizontal displace-
ment curves of dam top.

Figure 10. The final stage at which the numerical algorithm

broke.

Figure 14. The reduction factor and iterative frequency

required to achieve convergence curves.

Figure 11. The single sliding plane’s horizontal displace-

ment in ultimate status.

Figure 15. The shear strain rate vector under ultimate status.
Figure 12. The single sliding plane, at Fr = 3.15.

6 CONCLUSIONS

The dam stability analyses of two typical failure modes

5.2 The displacement and iteration features have been performed by finite difference strength
Figure 11 shows the contour of horizontal displace- reduction method that confirms the results in good
ment under ultimate State. Figure 12 gives the sliding agreement with those by Sarma method.
surface under ultimate status, which has good agree- The strength reduction analysis for the double soft
ment with that shown in Figure 3 with Sarma method. layers calculation model provides the development of
Figure 13 shows the curves between reduction fac- plastic zone along the seam with the increase of reduc-
tor and the horizontal displacement at the dam top, tion coefficient. Both SRF and Sarma method gave a
Figure 14 shows the curve between reduction fac- factor of safety F = 2.12. Different failure criterion
tor and iterative steps required by the convergence might give slightly different results. For this partic-
of the calculation. As can be seen from Figure 13 ular problem, judging based on the distribution of
and Figure 14, when the reduction coefficient is 3.15, plastic zone of dam base may result in a little con-
divergence occurs in both the two curves. Figure 15 servative estimation for the factor of safety, while
is shear strain rate vector under the ultimate status, the use of criterion of displacement divergence can
which shows the overall deformation trend of the give closer results to that of the limit equilibrium
dam base. method.

660
 The deep sliding pattern of single soft layer models Matsui, T. & San, K.C. 1992. Finite element slope stability
is obtained by strength reduction method. The main analysis by shear strength reduction technique. Soils and
sliding surface can be found in the dam base, together Foundations 32(1):59–70.
with sub-sliding surfaces passing through dam toe. Sarma, S.K. 1979. Stability analysis of embankments and
Not only the factor of safety but also the failure zone slopes. Journal of the Geotechnical Engineering Division,
ASCE 105(GT12):1511–1524.
are in good agreement with those obtained by Sarma’s Zienkiewicz, O.C, Humpheson, C. & Lewis RW. 1975. Asso-
method. ciated and nonassociated visco-plasticity and plasticity in
soil mechanics. Geotechnique. 25(4):671–89.
Ugai, K. & Leshchinsky, D. 1995.Three-dimensional limit
REFERENCES equilibrium and finite element analysis: a comparison of
results. Soils and Foundations 35(4):1–7.
Griffths D.V. & Lane P.A. 1999. Slope stability analysis by
finite elements. Geotechnique 49(3):387–403.

661
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Study on deformation parameter reduction technique for the strength

reduction finite element method

Q.W. Duan & Y.J. Wang

China Institute of Water Resources and Hydropower Research, Beijing, China

P.W. Zhang
Central Research Institute of Building & Construction, Beijing, China

ABSTRACT: Some problems related to stability analyses of slopes and dam foundations against sliding by
employing the strength reduction method are discussed in this paper. In view of the fact that the plastic zone
appeared during the reduction of strength parameters does not conform to the real case when using the strength
reduction method, a relationship between the friction angle, and Poisson’s ratio, is proposed, that is, among which
is a parameter related to the depth, unit weight and cohesion of geomaterials. According to this inequality, the
minimum friction angle can be determined for different depth, cohesion, unit weight and Poisson’s ratio. This
inequality can also take into account the more general cases, such as zero cohesion, infinite depth of geomaterial.

1 INTRODUCTION safety to the the Bishop’s Method. However, the

strength reduction technique implies that only the
Over the past decades, Finite-element (FE) method shear strength parameters are reduced, simultaneous
(including Discrete Element Method and Finite Dif- reduction of deformation parameters, such as Young’s
ference Method), has been widely used to analyze modulus, E, and Possion’s ratio, μ, are not considered.
problems of slope stability and stability of dam foun- Zheng (2002) discussed the necessity of reduc-
dation against sliding. Compared to limit equilibrium ing the shear strength and deformation parameters
methods, FE methods have advantages of being able simultaneously when using the strength reduction FE
to simulate the non-linear stress-strain relationship method, and found that when these two kinds of
of geomaterial and complex geometric and bound- parameters are reduced, the Poisson’s ratio, μ, and
ary conditions, and to take into account construction internal friction angle, ϕ, should satisfy the following
process, seepage, earthquake and reinforcement. In inequality, that is,
addition, FE methoda can also predict distribution
of stress or displacement, progress of plastic zone. sin ϕ ≥ 1–2μ (1)
However, FE Methods has a difficulty in estimating
the factor of safety similar to the limit equilibrium However, Equation 1 was derived based on: (1) a
method, which leads to no essential progress for null value of cohesion of geomaterial; and (2) an infi-
solving stability problems by using FE methods. nite depth of analysis area below the ground suface.
In recent years, FE methods combined with a Actually, these two assumptions are not often satisfied,
technique called strength reduction (or shear strength which leads to Equation 1 not be stood. Neglecting
reduction) have been developed to evaluate the struc- the deficiency mentioned above, utlization of Equa-
ture safety. The factor of safety is just one by which tion 1 may produce some abnormal plastic zones
the shear strength parameters are reduced, and the when the strength and deformation parameters are
reduced strength parameters make the caollapsed of reduced simultaneously. In view of the deficiency
strucure occur. The FE method combined the shear embedded in Equation 1, an update of Equation 1
strength reduction technique called strength reduc- is made in this paper. The validity of the update is
tion FE method in this paper. Griffith (1999) firstly also discussed. In addition, the necessity of reducing
applied the strength reduction FE method to slope Young’s modulus and its criterion are also discussed
stability problems and gave a compartive factor of for strength reduction FE technique.

663
2 STRENGTH REDUCTION FE METHOD

The strength reduction technique was first forward by

Zien-kiewicz et al. (1975), when performing elasto-
plastic analysis for soil stability by using FE method.
The concept of strength reduction is defined as a ratio
of the shear stress produced by the external loads act-
ing on the structure to the maximum mobilized shear
strength of soil. This strength reduction factor deter-
mined by this definition is conceptually the same as
the facor of safety used by the Bishop’s method. The
general procedure for determining the factor of safety
by using the strength reduction FE method is as follow- Figure 1. The distribution of stress in semi-infinite rock-
ing: (1) the material shear strengths are progressively mass with depth of h.
reduced by a reduction factor, F; (2) preforming FE
analysis based on the reduced strength parameters until 
collapse occurs. then he factor, F, is just the factor of σ1 = σ2 = −Kγ h
(3)
safety for the strength reduction FE method. σ3 = −γ h
Obviously, the value F determined has the same
meaning as material strength reservation factor (Pan where γ is the unit weight of geomaterial, K the lat-
1980). That is, If material strength parameters are eral pressure coefficient, which is defined as K = μ/
divided by the value of F, the structure system would (1 − μ).
collapse. If it is assumed that the stress field shown in Figure 1
⎫ is satisfied the Mohr-Coulomb criterion, thus
f ⎪
fi = ⎪
⎪
Kf ⎬ (1 + sin ϕ)σ1 − (1 − sin ϕ)σ3 ≤ 2c cos ϕ (4)
(i = 1, 2, . . . , n) (2)
C ⎪⎪
⎪
Ci = ⎭ Substituting Equation 4 into Equation 3 gives
Kf
−Kγ h(1 + sin ϕ) + γ h(1 − sin ϕ) ≤ 2c cos ϕ (5)
where fi , ci are the reduction factor at time i.
By combining the concept of strength reduction
Let h in Equation 5 tend to infinite or the cohesion is
technique with the elasto-plastic FE method, a number
equal to zero, the inequality sin ϕ ≥ 1–2μ, proposed
of strength reduction FE methods have been devel-
by Zheng (2002), can be obtained.
oped, and be applied to preform slope stability analysis
Rearrangement of Equation 5 leads to
and stability analysis of dam foundations against
sliding (Li 2000, Ma et al. 1984) 2c
(1 + K) sin ϕ + cos ϕ ≥ 1 − K (6)
γh
3 RELATIONSHIP BETWEEN FRICTION
ANGLE AND POISSON’S RATIO and then,
Generally speaking , the larger values of material 1−K
strength parameters c and ϕ are, the larger its elastic sin(ϕ + α) ≥   2 (7)
modulus is, and the smaller its Poisson’s ratio is. For 2c
example, the value of Young’s modulus of hard rock is (1 + K)2 +
γh
on the level of GPa, and the value of Poisson’s ratio is
not greater than 0.25. By contrast, the Young’s modu- Where,
lus of soft rock is on the level of MPa, and the value of
Poisson’s ratio is above 0.3. So far, the geneal way is 1+K
to reduce the strength parameters, c and ϕ, only, and cos α =   2 (8)
simultaneous reduction of deformation parameters, E 2c
(1 + K)2 +
and μ, is not reported when using the strength redution γh
FE technique.
For a semi-infinite geomaterial space as shown in 2c/γ h
sin α =   2 (9)
Figure 1, is only subjected to gravity, stresses acting 2c
on an element at depth, h, below the ground surface, (1 + K)2 +
can be defined as follows: γh

664
 It can be from Equations from 6 to 9 obtained that, 1 3
when the depth h tends to infinite or the cohesion c is
zero, then α = 0.
μ
By substituting K = 1−μ into Equation 7, one
obtains [ 1 3]o

sin(ϕ + α)/ cos α ≥ (1–2μ) (10)

[ 1 3 ]r
when α = 0, Equation 10 can be returned to
Equation 1. Eo
Equation 10 has demonstrated that there is a rela-
tionship between friction anger, ϕ, and Poisson’s
ratio, μ. When preforming the strength reduction FE
analysis, the abnormal plastic that is inconsistent with
Er
the real case will occur if sin(ϕ+α)/ cos α < (1−2μ),
so adjusting Poisson’s ratio, μ, according to Equation
10 is necessary to gurantee the unreaonable plastic
does not occur. Figure 2. The concept of reducing Young’s modulus E based
on the hyperbolic stress strain relationship, the subscript
‘o’ and ‘e’ stand for the original and reduced variables and
parameters.
4 ELASTIC MODULUS REDUCTION THEORY
AND METHOD

For the soil whose mechanical characteristics follows τ = c + σ tan ϕ (15)

E-ν model, The elastic constants accord with hyper- 2c cos ϕ + 2σ3 sin ϕ
bola relation, as shown in Figure 2, where a and b are (σ1 − σ3 )f = (16)
1 − sin ϕ
constants, σ1 and σ3 are major and minor principal
stresses respectively, ε1 is the axial strain, Ei is the
initial elastic modulus. (σ1 − σ3 )ult is the asymptotic It can be obtained from Qian (1994) that
line of (σ1 − σ3 ) when ε1 → ∞.
Et = (1 − Rf s)2 Ei (17)
ε1
σ1 − σ3 = (11)
a + bε1 (σ −σ )
Where, s = (σ(σ11−σ
−σ3 )
3 )f
, Rf = (σ11−σ33)ultf , which are known
According to differential relation, the tangent mod- as failure ratio,
ulus can be obtained as  n
σ3
Ei = Kpa (18)
∂(σ1 − σ3 ) a pa
Et = = (12)
∂ε1 (a + bε1 )2
Where, pa is atmospheric pressure, K and n are
Elastic modulus is derived from two conditions, constants.
When ε1 → ∞ It can be known from Equation 17 that, the initial
elastic modulus is invariable when Rf of soil is given.
And the elastic modulus of soil is decreased gradually
(σ1 − σ3 ) → 1/b = (σ1 − σ3 )ult (13) when s is increased, which means that when using the
strength reduction FE method to determine the factor
When ε1 → 0 of safety of a slope, the value (σ1 − σ3 ) at Mohr circle
shown in Figure 3 is not changed because the stress
within slope due to external loading is not changed.
Et = 1/a = Ei (14)
When shear strength parameters c and ϕ are reduced,
the corresponding strength determined by the Mohr-
According to Mohr-Coulomb criterion, the shear Coulomb are also changed as shown in Figure 3, and
stress at failure can be obtained by using the following the stress state of a point within the will approache to
equation, the yield state progressively.

665
 Table 1. Mechanic parameters of soil in Example 1.

Deformation Friction Tension Unit

modulus Poisson’s angle◦ Cohesion strength weight
(GPa) ratio (MPa) (MPa) kN/m3

10 0.26 35 1.3 0 27

Figure 3. Strength reduction FEM sketch map.

Figure 4. Analysis result without adjusting the Poisson’s

It is also known from Equation 17 that, the elastic ratio (F = 2.0).
modulus of soil is related to the stress state at fail-
ure, which means that for a set of material parameter
c1 and ϕ1 reduced already by a reduction coeffi-
cient F1 , the maximum stress at failure, (σ1 − σ3 )ult ,
can be determined by using the Mohr-Coulomb cri-
terion. Therefore, the elastic modulus of materials
can be determind by using Equation 17, the detailed
procedures are as follows:
1. Determine the elastic modulus corresponding to
parameter c and ϕ by Equation 17
  Figure 5. Analysis result with adjusting the Poisson’s ratio
(σ1 − σ3 ) 2
E1 = 1 − Ei (19) (F = 2.0).
(σ1 − σ3 )ult

2. Compute c2 and ϕ2 by a given strength reduction

coefficient F2 , and determine E2 as

/  02
E1
E2 = 1 − F + F Ei (20)
Ei
seam

3. Input E2 , c2 and ϕ2 into FE model to perform FE

analysis; Figure 6. Gravity dam and foundation model.
4. If the limit state is reached, then F2 is the factor of
safety obtained by strength reduction FE method,
and the calculation terminates, otherwise, increas- This special example is designed to illustrate how
ing the strength reduction coefficient and repeat the adjustment of the Poisson’s ratio affect the appear-
Step 1 to Step 4. ance of plastic zone when strength reduction FE
method is applied. For this simple semi-infinite soil
mass, all the surrounding boundaries are only moved
5 EXAMPLES in the direction of normal to their side and the geo-
material is only subjected to gravity. When using the
Example 1 For a semi-infinite soil, the physical and strength reduction FE method to analyze this prob-
mechanical parameters are summarized in Table 1. lem, the Poisson’s ratio is firstly not changed, and the

666
Table 2. Adapting parameter of dam and foundation.

Deformation Tension Unit

Material modulus Poisson’s Friction Cohesion strength gravity
name (GPa) ratio angle◦ (MPa) (MPa) kN/m3

Dam 20 0.167 35 2.0 1.85 24

Foundation 10 0.26 35 1.3 0 27
Interlayer 2.5 0.35 19.8 0.115 0 25.6

Figure 7. Horizontal displacement-reduction coefficient

curve without adjusting Poisson’s ratio. Figure 9. Distribution of plastic zone without adjusting
Poisson’s ratio (F = 2.12).

Figure 8. Distribution of plastic zone without adjusting

Poisson’s ratio (F = 2.12).
Figure 10. Horizontal displacement-reduction coefficient
curve with adjusting Poisson’s ratio.
soil below a depth of 200 m is in a plastic state when
the reduction factor F = 2.0, as shown in Figure 4.
However, when both strength parameters and Pois-
son’s ratio are adjusted, espe-cially the Poisson’s ratio
is changed from 0.26 to 0.29, the plastic zone appeared
in Figure 4 disappeared, as shown in Figure 5, which
is more identical with the real case.
Example 2 stability analysis of dam foundation
against sliding.
This example is focused on a gravity dam with
height of 129 m, under which two weak seams BC
and CD are present as shown in Figure 6, among them, Figure 11. Geometry model of slope (unit: m).
weak seam BC dips at 15◦ downstream, and weak seam
CD dips upstream at 10◦ .
For this case, except the gravity is considered in Table 3. Material parameter of slope.
analysis, only water pressure acting on the upper
stream side of dam and upstream foundation is Deformation Unit
considered. The physical and mechanical parame- modulus Poission’s Friction Cohesion gramity
(GPa) ratio angle◦ MPa) kN/m3
ters of geomaterials are summarized in Table 2 the
Mohr-Coulomb failure criterion is applied also in anal- 0.1 0.4 20 0.01 20
ysis, following conditions and parameters mentioned

667
Figure 12. Distribution of plastic zone.

above, the strength reduction FE method gives a factor factor of safety for the case of reducing Poisson’s ratio
of safety of 2.12, which is close to Fos = 2.19 is 2.1 or so, just a little different rom that for the case
determined by traditional Sarma’s method. of no adjusting the Poisson’s ratio. However, charac-
For the case of no adjustment of Poisson’s ratio, teristics ofdisplacement for the two cases of adjusting
curves of horizontal displacement at characteristic and no adjusting the Poisson’s ratio are quite different,
points, A, B, C, D, E, and F, with reduction coefficient as shown in Figure from 7 to10, which indicates that
are shown in Figure 7. If utilization of the phenomenon Poisson’s ratio adjustment during the strength param-
that the displacement curve for the characteristic point eters reduction, provides not only a correct estimation
changes sharply at a vertain value of the reduction of the factor of safety, but also a pertinent evolution of
coefficient as judging criterion for the occurrence of plastic zone.
limit failure state, the factor of safety of Example 2
is 2.12 for the case of no adjusting the Poisson’s ratio. Example 3
The corresponding plastic zone is shown in Figure 8. It Example 3 is a homogeneous slope with a slope
can be seen from Figure 8 that, the shear plastic zone of 26.57◦ , its geometry is shown in Figure 11, and
occurs not only in weak seams, but also in the deep mechanical parameters for the slope material are listed
foundation. actually, the plastic zone appeared in the in Table 3. For the FE model, the bottom boundary
deep foundation are due to the disaggrement of Pois- is fixed, left and right boundaries can only move in
son’s ratio with the reduction strength parameters and vertical direction.
Eqigure 9 is violated. For this example, three different cases are con-
To avoid occurrence of the unreasonable plastic sidered when performing the strength reduction FE
zone, and make Equation 9 stand, the value of Poisson’s analysis, namely, (1) Case 1: adjusting both deforma-
ratio of foundation material is adjusted to 0.33. the tion modulus and Pois-son’s ratio; (2) Case 2: adjusting
plastic zone in deep foundation as shown in Figure 10 Poisson’s ratio only; (3) Case 3: not adjusting defor-
disappear except the weak seams. The curves of dis- mation modulus and Poisson’s ratio. The plastic zones
placement at characteristic points with the strength for these three cases are illustrated in Figure 12. It
reduction coefficient as shown in Figure 10 shows the can be found from Figure 12 that, (1) an adjustment

668
of Poisson’s ratio plays an imortant role on the dis- • The reduction of Young’s modulus of soils obeying
tribution of plastic zone; (2) an adjustment of elastic E-νhyperbolic model has been studied and its effect
modulus has no minimal influence on distribution of on the plastic zones and factors of safety has been
plastic zone and the factor of safety; however, tak- discussed.
ing much small value of elastic modulus would cause
unreasonal factor of safety; (3) no adjustment of elas-
tic moduls and Poisson’s ratio will produce abnormal REFERENCES
and impractical plastic zone.
PAN J.Z. 1985. Engineering geology calculation and foun-
dation treatment. Beijing: Water Resources and Electric
6 CONCLUSIONS Power Press. (in Chinese)
Zheng H. & Li C.G. 2002. Finite eIement method for soIv-
ing the factor of safety, Chinese Journal of Geotechnical
Based on the analysis results, the following conclu- Engineering, V24(5), 626∼628.
sions can be drawn: Griffith D.V. & Lane P.A. 1999. Slope stability analysis by
• The previous strength reduction FE method only finite element. Geotechnique, 49(3), 387∼403.
consider reduction of shear strength parameters, Zienkiewicz O.C., Humpheson C & Lewis R.W. 1975. Asso-
ciated and Non-Associated Visco-Plasticity and Plasticity
adjustment of Poisson’s ratio and Young’s modu- in Soil Mechanics. Geotechnique, 25(4), 671∼689.
lus are not available. In this paper, an adjustment LI P.J. 2002. Analysis on deep anti-sliding of gravity dam,
criterion for Poisson’s ratio and Young’s modu- Design of water resources and hydroelectric projects,
lus, sin(ϕ + α)/ cos α ≥ (1–2μ), are proposed 2000, V19(1), 49–52.
for propore utlization of strength reduction FE MA L., ZHANG Y.M. & ZHANG L.Q. 1984. Several prob-
technique. lems concerning the analysis of stability against deep
• Although the strength reduction FE method not sliding in gravity dam foundation. Journal of Hydraulic
considering the Poisson’s ratio be adjusted may Engineering, (1), 27–35. (in Chinese)
give the factor of safety in a good agreement with LUAN M.T., WU Y.J. & NIAN T.K. 2003. A criterion for
evaluating slope stability based on development of plas-
that determined by limit equilibrium method, how- tic zone by shear strength reduction FEM. Journal of
ever, some abnormal and impractical plastic zones Seismology, 23(3), 1–8. (in Chinese)
appears. The strength reduction FE method with ZHAO S.Y., ZHENG Y.R. & DENG W.D. 2003. Stabil-
adjusting the Poisson’s ratio according to the cri- ity analysis of jointed rock slope by strength reduction
terion presented in this paper is proved to provide FEM[J]. Chinese Journal of Rock Mechanics and Engi-
only an appropriate factor of safety, but also a sound neering, 22(2), 254–260. (in Chinese)
plastic zone.

669
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Stability and movement analyses of slopes using Generalized Limit

Equilibrium Method

M. Enoki
Tottori University, Tottori, Japan

B.X. Luong
University of Transport and Communication, Hanoi, Vietnam

ABSTRACT: The authors proposed Generalized Limit Equilibrium Method, which can analyze every type
of plasticity problems. This method was developed from the ordinary slice method, but the assumptions
on the forces on inter-block planes are more reasonable. By introducing inertia forces caused by the earth-
quake accelerations, GLEM can treat the dynamic problems. In this paper GLEM is introduced and applied
to some plasticity problems including static stability and dynamic behavior. Finally, the applicability of clas-
sical theories of plasticity including GLEM is discussed in regard to the assumption of rigid-perfect plastic
material.

1 INTRODUCTION 2 STATIC GLEM

Geotechnical engineers often encounter the dynamic 2.1 Formulation of static GLEM
and plastic problems, such as deformations of foun-
The Generalized Limit Equilibrium Method was
dations or slopes subjected by an earthquake or
developed from the ordinary slice method (Enoki et al.
dynamic loading. Newmark’s method and the deriva-
1990), but the failing soil mass is divided into n pieces
tive methods are famous to analyze these problems
of triangular or quadrangular blocks, as shown in
(Newmark1965). However, they have the common
Fig. 1.
weak point that they cannot obtain the actual solution
The following equilibrium conditions of every
for the static problems, for example, the Prandtle’s
block give 2n equations under a given geometry of
solution for the bearing capacity problem.
block system:
The authors already proposed the Generalized Limit
Equilibrium Method (GLEM), which can analyze Equilibrium toward the direction normal to the
every type of plasticity problems such as bearing block base plane
capacity problems, earth pressure problems and slope
stability problems (Enoki et al. 1991a, b). −Hj cos(βj − αi ) + Vj sin(βj − αi )
GLEM can give the approximate solution obtained
+ Hj−1 cos(βj−1 − αi ) − Vj−1 sin(βj−1 − αi )
by Slip Line Method (like Kötter’s equation) and the
upper bound solution in the strict sense of Limit + Mi g cos αi = Pi (1)
Analysis Method.
By introducing inertia forces caused by the earth- Equilibrium toward the direction tangential to the
quake accelerations, GLEM can treat the dynamic block base plane
problems, such as the movement of slopes (Enoki et al.
2005). Hj sin(βj − αi ) + Vj cos(βj − αi )
In this paper GLEM is introduced and applied to
− Hj−1 sin(βj−1 − αi ) − Vj−1 cos(βj−1 − αi )
some plasticity problems including static stability and
dynamic behavior. + Wi sin αi = Ti (2)

671
 Hn+1 geometry of block system. In the case of slope problem
Vn+1 the safety factor Fs becomes a variable instead of Hn+1 .
n th block Thus, the static plasticity problem is transformed
i th block into an optimization problem in which bearing capac-
1st blockst ity or safety factor is minimized for the geometry of
1st interblock Vj (n+ 1) th block system with restrained conditions expressed by
plane Mig Hj Interblock equations.
plane
This kind of optimization problem can be easily
Hj-1 Block base plane solved by ‘‘Solver’’, additional function of Excel.
Vj-1 Interblock The readers can download the Excel sheets and FOR-
plane
Ti Pi TRAN source programs free for this analysis from the
following web:
http://www.denkishoin.co.jp/index.html
Figure 1. Block and force system used in Generalized Limit
Equilibrium Method.
(Only Japanese language for the present)

2.2 The meaning of optimization

Table 1. Number of variables and equations in static GLEM.
In the limit equilibrium method including GLEM, the
Equations Equilibrium condition process of optimization is required. For example, in a
Normal to base plane n bearing capacity problem the ultimate bearing capac-
Tangential to base plane n
ity is minimized for the geometry of block system.
Failure condition Total 4n
On base plane n The physical meaning of this process in GLEM is
On interblock plane n demonstrated in Figs. 2 and 3.
Variables Forces on block base plane As shown in Fig. 2, in the initial geometry the stress
Normal force P1 ,---, Pn n state of every block (expressed by a Mohr’s stress cir-
Tangential force T1 ,---, Tn n cle) has intersections with two failure lines, failure
Forces on interblock plane Total 4n conditions on block base plane and interblock plane.
Normal force H2 ,---, Hn n However, the stress circles exceed the failure lines.
Tangential force V2 ,---, Vn n The centers of stress circles are not always on σ -axis,

BL i : Block nimber
(a)Initial Geometry
r B i : Block base plane number
I i : Interblock plane number
The following failure conditions are applied not
only to the block base planes but also to the interblock Surcharge
r q=1
I1 I6
planes, and 2n equations are obtained: B1
BL1
I2
I5 BL5
B5
I4
c=0, =30 , BL2
I3
Failure condition on the block base plane =0, q=1, B=1 B2
B3
B4 BL4
60 BL3
e
lan
(b)Stress state of every block ep
tan φ c kb
as
Ti = Pi + li (3) B4 B
lo c
Fs Fs 40
B5
Failure condition on the interblock plane BL4
BL5
20 B3
tan φ c
Vj = Hj + lj (4)
Fs Fs B1 B2 BL3
I6
0
20 40 60 80 100
Then, total 4n equations are obtained. In a bearing I1 I2
I3
capacity problem the variables are normal and tangen- I4
BL1
tial forces, P1 , ---, Pn and T1 , ---, Tn on the block -20
BL2 Int
base plane, and H2 , ---, Hn and V2 , ---, Vn on the erb
loc
interblock plane. Hn+1 is the objective force (ultimate k pI5
lan
bearing capacity Qu ), Fs = 1, and H1 and V1 are e
known surcharge.
The number of variables is also 4n, as shown in Figure 2. Stress state of every block before optimization in
Table 1, and the problem is determinate under a given GLEM.

672
and this means equilibrium of moment of rotation is of bearing capacity is a circle of which center is the
not satisfied. left end of foundation and radius is B.
As shown in Fig. 3, after the minimization of bear- Stability of a slope is analyzed by static GLEM
ing capacity for the geometry of block system, the under the assumption of toe failure and the result is
stress state of every block hardly exceed the failure shown in Figure 5.
lines, and the centers of stress circles are almost always
on σ -axis. In GLEM the equilibrium of moment is not
used, but it is satisfied as a result.
Thus, the authors believe that GLEM can give the 3 DYNAMIC GLEM
approximate solution obtained by Slip Line Method
(like Kötter’s equation) and the upper bound solution in 3.1 Introduction of inertia forces
the strict sense of Limit Analysis Method. The detailed To formulate the dynamic GLEM, the horizontal
reason should be referred to Enoki 2007. In the limit and vertical accelerations of the i-th block, αhi and
equilibrium methods other than GLEM, such as all αvi , are introduced into the static GLEM as inertia
the slice methods, the failure condition is not adopted forces, −Mi αhi and −Mi αvi using d’Alembert’s prin-
on inter-slice planes, then the physical meaning of ciple (Enoki et al. 2005). The kinematic condition of
optimization is not clear and not same with GLEM. the foundation is required to solve the problem. Now,
we give the mass of the foundation, Mf , and assume
the rough base that means the foundation and the n-th
2.3 Illustrative examples of static GLEM block move as one, as shown in Figure 6.
Then, the following motion equations are obtained
So-called Nq problem, ultimate bearing capacity of
for every block instead of Equations 1 and 2 for a
unit width foundation on weightless ground (c = 0,
dynamic problem under a given geometry of block
φ = 30◦ ) under unit surcharge, is analyzed by static
system:
GLEM. The bearing capacity and slip surface obtained
are added to Hansen’s figure and shown in Figure 4.
In this analysis initial slip surface before minimization
31.1 Bishop
c=0, =30 , =0, q=1, B=1 27.9 Janbu1
22.6 Rendulic
Bishop, Janbu1 22.5 Hansen
Nq 19.5 GLEM
(a)Geometry after optimization Prandtl, Janbu2
18.4 Prandtl
13.0 Janbu2
BL i : Block number Hansen 9.0 Frict. Circ.
Surcharge B i : Block base plane number Unit surcharge 5.6 Conventional
q=1 I i : Interblock plane number

I1 I6 GLEM
BL1 BL5
B1 I5 B5
I2 Rendulic
I3 I4
c=0, =30 , BL2 B4 BL4 Fric. Circ., Conv.
=0, q=1, B=1 B2 B3 BL3
ne
10 pla
b ase Figure 4. Bearing capacity and slip surface for Nq problem
(b)Stress state of every block ck analyzed by static GLEM and other methods (after Hansen
Blo
1970).
B5
5 BL2 B4
BL1 B3 BL5
B2 Safety factor Fs = 2.095 (10,5.774) (12,5.774)
B1 BL4
BL3 H=5m, = 30 r
I6 ppe (7.5,4.33)
0 of u ne (12,5.774)
c = 1, = 30 , =2 tes pla
5 10 15 19 20 s
r d inalock
I1 coo terb
I2 tial f in (10.5,4.2)
Ini int o (5,2.89)
I3 po
I4
Toe failure
--5 I5 (2.5,1.44) (8,2.3)
Int owe
r
erb (0,0) (6,1.3) of l ne
loc 30
i n atecs k pla Slip surface (Initial)
kp r d o
(3,0.3) coo rbl
lan (0,0) ial inte
e Initint of Slip surface
po (after minimization)

Figure 3. Stress state of every block after optimization in Figure 5. Safety factor and slip surface of slope problem
GLEM. analyzed by static GLEM (Enoki 2007).

673
 Mass of Table 2. Number of variables and equations in dynamic
foundation Mf GLEM at the instant of failure.
Rough base
Acceleration Equations Equilibrium condition
of i-th block Normal to base plane n
n-th block Tangential to base plane n Total 4n–1
i-th block Failure condition
1st block On base plane n
1st interblock Vj Foundation and On interblock plane n–1
Hj n-th block move
plane as one Variables Forces on block base plane
Mi hi Normal force P1 ,---, Pn n
Hj-1 Block base plane
Mi(g+
Tangential forceT1 ,---, Tn Total 4n–1
vi)
Vj-1 Interblock plane Forces on interblock plane
Ti Earthquake Normal force H2 ,---, Hn−1 n–1
Pi acceleration
Tangential force V2 ,---, Vn−1 n–1
Critical earthquake
acceleration
Figure 6. Block and force system used in dynamic GLEM. Horizontal α h,cr 1

*α v is given as the earthquake acceleration.

Equilibrium toward the direction normal to the
base-block plane, n’

v’
−Hj · cos(βj − αi ) + Vj · sin(βj − αi )
h’
+ Hj−1 · cos(βj−1 − αi ) − Vj−1 · sin(βj−1 − αi ) n’ = n

+ Mi (g + αvi ) · cos αi − Mi · αhi · sin αi = Pi n

(5)
v h
Equilibrium toward the direction tangential to
the base-block plane,
Figure 7. The continuity condition of accelerations.
Hj · sin(βj − αi ) + Vj · cos(βj − αi )
3.3 After the failure
− Hj−1 · sin(βj−1 − αi ) − Vj−1 · cos(βj−1 − αi )
When the actual earthquake acceleration αh exceeds
+ Mi (g + αvi ) · sin αi + Mi · αhi · cos αi = Ti αh,cr at t = t1 , the ground fails, and this process will
(6) begin. In this process every block moves relatively
to each other, and all the accelerations αh1 , αh2 , . . .,
The failure conditions, Equation 3 and 4 are still αhn and αv1 , αv2 , . . . , αvn are variables. In order to
applicable, and Fs = 1. determine these accelerations, the continuity condition
Equations 1 and 2 are a special case of Equations 5 of accelerations and the conservation of slip surface
and 6, where every acceleration of block, αh1 , αh2 , . . ., are useful.
αhn , αv1 , αv2 , . . ., αvn , are equal to zero. As for the The continuity condition of accelerations was found
horizontal and vertical accelerations of every block, by one of the authors, when the motion of the surface
there are two processes in dynamic GLEM. layer during an earthquake was analyzed. It says that
the components of the accelerations normal to a slip
plane, αn and αn , are equal to each other across the slip
3.2 The instant of failure plane, and Equation 7 is written, where αh , αv and αh ,
Before and at the instant of failure, every block moves αv are accelerations of one side and another side across
similarly to the base of the ground, or αh1 = αh2 = the slip plane, respectively, and θ is the inclination of
· · · = αh and αv1 = αv2 = · · · = αv . the slip plane, as shown in Figure 7.
In this case the objective is to determine the critical
acceleration αh,cr that causes the ground to fail under αn = αv cos θ − αh sin θ = αv cos θ − αh sin θ = αn
the given αv , and the problem is still determinate as (7)
shown in Table 2.
The critical acceleration αh,cr must be minimized This means the conservation of mass during the
for the geometry of the slip surface. motion, and also written as(αv −αv )/(αh −αh ) = tan θ.

674
Table 3. Number of variables and equations in dynamic Geometry of st ructure B = 3m
GLEM after the failure. at critical state
Geometry of st ructure
Equations Equilibrium condition at moment t = 0.52 sec Foundati
Normal to base plane n Inter -block on
Tangential to base plane n plane
Failure condition Total 5
On base plane n 6n–2 1
Sub -base 2 3 4
On interblock plane n–1
Continuity condition of Sliding
surface
acceleration
On base plane n
On interblock plane n–1 Figure 8. Displacement of block system in foundation
Variables Forces on block base plane Total problem at moment t = 0.52 sec.
Normal force P1 ,---, Pn n 6n–2
Tangential forceT1 ,---, Tn n
Forces on interblock plane
Normal force H2 ,---, Hn−1 n–1
Tangential force V2 ,---, Vn−1 n–1
Acceleration of every block
Horizontal α h1 ,---, αhn n
Vertical αv1 ,---, αvn n

The conservation of slip surface says that the slip

surface remains unchanged from the initial one dur-
ing a single motion. This is not always necessary to
Figure 9. Horizontal accelerations of sub-base, foundation
formulate the dynamic GLEM, but convenient in the
and soil blocks.
calculation and realistic for the actual problems.
After applying these conditions, the problem
becomes determinate as shown in Table 3, and the
accelerations of every block can be determined. The
optimization of the acceleration for the geometry of
the slip surface is not required in this process.
When the relative velocity of arbitrary block
becomes zero at t = t2 , the slip stops, as written in
Equation 8.

 t2
(αhi − αh )dt = 0 (8) Figure 10. Vertical displacements of foundation and soil
t1 blocks.

3.4 Illustrative examples of dynamic GLEM

The model test of finite slope was conducted on a
The behavior of a shallow foundation with a rough shaking table, as shown in Figure 11. Toyoura sand
base, Mf = 70 t and B = 3 m, during an earthquake with density of 1.8 t/m3 , void ratio e = 0.8, moisture
is analyzed by dynamic GLEM. The soil has density content 1%, and strength parameters of cpeak =
of 1.8 t/m3 and strength parameters of c = 10 kPa, 1.25 kPa, cresidual = 0.85 kPa, φpeak = φresidual =
φ = 32◦ . Only a horizontal wave αh = 5.75 sin 36.4◦ was used. The slope had height h = 30 cm and
2πt (m/s2 ) is given as an earthquake acceleration. inclination 50◦ . Only a horizontal wave αh = 10 sin
A system of five triangular blocks is used as shown 12πt (m/s2 ) was given as an earthquake acceleration.
in Figure 8. The slip surfaces observed in the experiment is
After the given acceleration exceeds the critical compared with the analyzed one in Figure 12. The hori-
acceleration αh,cr = 3.94 m/s2 , all the blocks move zontal accelerations and displacements observed in the
differently from each other as shown in Figure 9. experiment are shown in Figure 13. It can be compared
Within one wave the failure occurred twice, and the with the analytical ones shown in Figure 14. Accord-
settlement of foundation is about 6 mm as shown in ing to the experimental results, the sliding begins to
Figure 10. occur from the second wave of input acceleration at a

675
Figure 11. Model test of finite slope on shaking table.

Figure 14. Analytical results of slope motion by dynamic

GLEM.

theoretical analysis is a little larger than the experi-

mental one (2.76 cm compared to 2.02 cm).
Although there are still some differences, it can
be said that the experimental results have rather good
agreement with the analytical ones.
The comparison of experimental result of founda-
tion model on a shaking table with the analytical one
showed not so good agreement as in the case of slope.

Figure 12. Slip surfaces observed and analyzed. 4 DISCUSSION ON THE APPLICABILITY
OF GLEM

The stability analysis based on the classical theories of

plasticity such as Slip Line Method and Limit Anal-
ysis Method, requires the assumption of the perfect
plasticity (with no hardening nor softening) of the
material. This assumption can be easily satisfied in the
unsaturated medium-dense sand. However, the motion
analysis based on the classical theories of plasticity
requires the assumption of the rigid-perfect plastic-
ity. This assumption cannot be easily satisfied in the
foundation problems where the change in stress in the
soil is large, and volumetric compression and elastic
Figure 13. Experimental results of slope motion. deformation cannot be neglected. On the other hand,
the change in stress from the initial state till the fail-
ure state is small in the slope problems; therefore the
‘‘rigid’’ assumption can be approximately satisfied.
critical acceleration of about 800 gal. Then, the criti- As the static or dynamic GLEM is a kind of the
cal acceleration decreases and has the value of about classical theories of plasticity, the applicability may
600 gal corresponding to the third wave, and 365 gal have the same limitation. The authors believe that this
for the following waves. is the reason why the analytical results agrees well with
This fact indicates that for the early stage of the the observed one in dynamic slope problems but not
failure, the peak shear strength of the soil are mobilized agrees in dynamic foundation problems.
to resist the sliding, thereafter the strength of the soil
will decrease and be at residual value. In the theoretical
analysis, with the use of the residual strength of the 5 CONCLUSIONS
soil, the sliding begins to occur just from the first wave
with the critical acceleration of about 438 gal. The outline of the static and dynamic Generalized
Consequently the residual displacement in hori- Limit Equilibrium Method is introduced in this paper.
zontal direction of the sliding soil mass obtained by The static GLEM can give the approximate solution

676
obtained by Slip Line Method and the upper bound REFERENCES
solution obtained by Limit Analysis Method.
The procedure of optimizing the external force Enoki, M., Yagi, N., Yatabe, R. & Ichimoto, E. 1990. Gen-
(earth pressure or bearing capacity) or safety factor in eralized Slice Method for Slope Stability Analysis. Soils
GLEM plays a role to prevent stress state from exceed- and Foundations 30(2): 1–13.
ing the failure criterion. It also plays a role to satisfy Enoki, M., Yagi, N., Yatabe, R. & Ichimoto, E. 1991a. Gener-
alized Limit Equilibrium Method and Its Relation to Slip
the equilibrium of moment of rotation in a block. Line Method. Soils and Foundations 31(2): 1–13.
Since the static GLEM is formulated as a special Enoki, M., Yagi, N., Yatabe, R. & Ichimoto, E. 1991b.
case of the dynamic GLEM, where acceleration of Relation of Limit Equilibrium Method to Limit Analysis
every block is equal to zero, there is no essential Method. Soils and Foundations 31(4): 37–47.
difference between them. Thus, the authors suppose Enoki, M., Luong, X.B., Okabe, N. & Itou, K. 2005.
that the dynamic GLEM gives the approximate solu- Dynamic Theory of Rigid-Plasticity. In Soil Dynamics and
tion obtained by the dynamic Slip Line Method and Earthquake Engineering: 635–647, London, Elsevier.
the upper bound solution obtained by the dynamic Enoki, M. 2007. Analyses of Stability and Deformation in
Limit Analysis Method, although they have not been Soil-like Frictional Material: Tokyo. Denki-Shoin (in
Japanese).
established yet. Hansen, J.B. 1970. A Revised and Extended Formula for
GLEM seems to have the limitation on application, Bearing Capacity, Geotekniks Institute, Bull. 28: 5–11.
that is, unloading and active problems can be properly Newmark, N.M. 1965. Effect of earthquakes on dams and
analyzed, but loading and passive problems cannot embankments. Fifth Rankine Lecture Géotechnique 2:
properly be analyzed, because of the assumption of 139–160.
rigid-perfect plasticity of the material.

677
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Long-term deformation prediction of Tianhuangpin ‘‘3.29’’ landslide

based on neural network with annealing simulation method

Faming Zhang, Chenxin Xian, Jian Song & Binyue Guo

Earth Science and Engineering Department of Hohai University, China

Zhiyao Kuai
Geological Department of Chang’an University, China

ABSTRACT: Landslide has already become one of the most dangerous geo-hazards in China, there are a
lot of economy loss and personnel casualty. Recently, the safety research on stability analysis or predict to the
hazard taken place for landslides is more and more important. But, it is difficult to predict when the sliding will
happen accurately since the factors affected on landslide are complex. Thus, in these years, monitor scheme is
adopted for many landslides, how to use the monitoring data to predict the landslide will occur is an important
problem in geological field. In this paper, based on adopting neural network with simulated annealing method,
the model of landslide deformation prediction in long-term was set up, which using simulated annealing method
to overcome the disadvantage of BP neural network, furthermore, by using dynamic forecasting technique to
reduce the influence of the prophase displacement, it can get better forecast precision. Finally, the Tianhuangpin
‘‘3.29 landslide’’ deformation in long-term was predicted by using the model discussed in this paper, the result
is coincident to the real condition of the slope.

1 GENERAL INSTRUCTIONS and predicted value are companied with the factors
affected on landslide, so, it is difficult to forecast the
Tianhuangpin pump-storage power station is located at deformation of landslide in long term. The main fac-
Tianhuangpin town in Zhejiang Province, it is 57 km tor is the monitored data are changed with seepage
apart from Hangzhou and 180 km from Nanjing. and weather, In this paper, based on adopting neural
The 329 Landslide is distributed in the left bank of the network with simulated annealing method, the model
lower reservoir of the project, it is 430 m from the of landslide deformation prediction in long-term was
dam in upstream. The first sliding of the landslide was set up, which using simulated annealing method to
occurred in the morning of 29, March of 1996, which overcome the disadvantage of BP neural network, the
the 2 × 106 m3 slide mass was rushed into the reser- prediction result of Tianhuangpin landslide is more
voir. The further exploration work found the instability accurate with monitored data.
slope body was more than 19 × 106 m3 , which slip
surface of the landslide is in middle weathered break
rock mass. From 1997, the monitoring network was
2 GEOLOGICAL CONDITIONS
set up. In the past 10 years, there are many mon-
itored data was collected. But, how to evaluate the
2.1 Topography and physiognomy
landslide stability and predict the long-term defor-
mation performance by using the monitored data is ‘‘3.29’’ landslide of Tianhuangpin power station is
an important subject in safety running of the power located in Tianmu Mountain area, the famous moun-
station. tain in China, the highest altitude of the landslide is
It is well known that the research work of predic- 970 m, the vertical difference of the landslide is more
tion for landslide are only recent 30 years, the method than 400 m. The strike direction of the landslide is
used in forecasting normally is statistics and experi- S∼N, and the slope degree is 40∼50◦ . The width of
ence [Liu 1996, 1998]. In recent years, grey system the landslide is from 15 m (top area) to 600 m (lower
theory, fuzzy mathematic etc. are used in landslide part) (Figure 1). All the accumulation of the landslide
deformation prediction [Deng 1985, 1987]. When body is consisted of weathered volcanic rock mass
talk about the methods, the errors between monitored (Figure 2).

679
 2.4 Hydro-geological property
Atmosphere precipitation is the supply source for
underground water, the buried depth of the table is
more than 50 m since the joint closed spaced.

2.5 Weathering phenomenon

The depth of complete weathered rock mass subface
is 2.0∼5.0 m (EL464.90∼461.90 m), the maximum
depth is 59.50∼60.10 m (in EL506.97∼506.37 m).
The strong weathered is 60.10∼70.90 m (EL506.37∼
495.57 m), and the weak weathered is 40∼85 m which
under the stress-release interface 10∼30 m.

3 MONITORING SCHEME

Figure 1. ‘‘3.29’’ landslide scope. 3.1 Monitoring scheme

After the ‘‘3.29’’ landslide happened, the accumulation
of the landslide was clean up by the Tianjian Corpora-
tion, the slope was reinforced by taken downwearing
and revetment, the stability of the slope is strength-
ened. From 1999, the monitoring scheme was taken
in the surface deformation, slope indicator, surface
settlement, rainfall and piezometer etc. The scheme
map of landslide monitor is illustrated in Figure 3,
which HP is indicated as slope indicator, TP is indi-
cated as horizontal surface displacement and LD is
vertical surface displacement.
For surface deformation monitoring, there are
18 monitor sites lain on the surface of the slope, which
take 15 days as one cycle. For the interior deformation

Figure 2. Accumulation of ‘‘3.29’’ landslide.

2.2 Stratum and rock mass property

In the landslide distributed area, the rock mass is Juras-
sic system up series Hangjian group (Jh1 3 ) and Lacun
group (JL1(d)
3 ) volcanic rock with intrusive granite-
porphyry (γ π52–3 ). The Quaternary loose deposits
were distributed on volcanic rock. Huanjian group
(Jh1
3 ) is grey tufflava and ryholitical agglomerate.
Lacun group (JL1(d)
3 ) is gravel agglomerate.

2.3 Structure
The structure in landslide area is well developed, the
main faults are F105 and F106 . These two faults are
0.4∼2.0 m width, which filled with fault mud and
mylonite. The joints are closely spaced. Figure 3. Monitoring scheme of ‘‘3.29’’ landslide.

680
monitoring of the slide body, there are 11 clinometers in Figure 4. The curves of deformation along the depth
installed from 1999. were figured in Figure 5 and Figure 6.
The other monitoring sites of interior deformation
were lain out in Table 1.
3.2 Deformation information statistics
It is obvious that the interior deformation especially on
the sliding surface may representative the stability of
the landslide or slope, so, in this paper, the deformation
prediction is only discussed for interior deformation
performance. On considering the probability sliding
surface of the landslide, the deformation with depth
of 10 monitoring holes was considered from 2000.
Take an example, monitoring hole 2 (called HP2),
installed at 450 m, the depth is 38.5 m. There are 2
abnormal changes at 18 m and 34.5 m in direction A.
In direction B, there are 2 obvious variations along the
depth at 17.5 m and 37 m. The 8 m depth’ deforma-
tion changes with time in direction A were illustrated

Figure 6. Deformations with depth in direction B of HP2.

Table 1. The maximum exceptional deformation changed

in monitoring hole with depth.

The depth The

Moni- of obvious maximum
toring Dire- deformation deformation
location ction Time variation (m) variation (mm)

hp1 A 2004-11-15 8 8.80

2005-6-18 18 7.63
2005-6-18 33 3.21
B 2001-9-18 17.5 6.94
2003-7-22 25.5 4.22
Figure 4. Deformation with time curve of 18 m for HP2.
hp3 A 2003-8-21 21.5 7.74
B 2002-3-15 21.5 8.08
2002-3-15 24 3.52
hp4 B 2005-7-18 21 13.13
2003-9-16 35 15.36
2005-7-18 42.5 9.75
hp5 A 2002-7-25 19.5 6.19
2004-1-10 26 4.61
B 2002-7-25 27 6.96
hp6 A 2003-12-5 18.5 5.86
2005-4-15 31.5 2.43
B 2001-6-16 27 4.72
hp7 A 2003-7-22 7 8.76
2002-2-28 33 1.16
B 2002-7-25 27.5 4.51
hp9 A 2001-7-27 12.5 5.02
B 2002-6-20 12.5 14.10
hp10 A 2003-1-20 12.5 5.78
B 2001-4-13 7 7.06
Figure 5. Deformations with depth in direction A of HP2.

681
4 MODEL OF LONG-TERM DEFORMATION the plentiful specimen data is required. It is the best
PREDICTION way to use interpolation method for unequal inter-
val data.
4.1 Prediction model of simulated annealing
with neural work
4.2.3 Data alternate process
Since the BP neural network technology can’t guar- Characteristic function of BP neural network normally
antee the convergence at the minimum value, it is used is S (Sigmoid) function the interval of the func-
difficult to get the best result in optimization pro- tion value is (0, 1), so, it is necessary to change the
cess [Sandro 1997], but not the same with BP neural monitored data value into (0, 1) interval. The method
network technology, simulated annealing method can can be used as following methods.
research the optimum result [Kang etc., 1994]. As we The first method is normalizing, it changes the
known, the study of neural work process is to look for import data to normalize to (0, 1) interval. The nor-
a suitable mapping or optimalizing weight and critical malizing formula is:
point threshold, so, the simulated annealing method
can be used in dealing with the study of neural work
Y − Ymin
process. In this reason, the model of simulated anneal- X = Xmin + (Xmax − Xmin ) (2)
ing with neural work was set up to forecasting the Ymax − Ymin
deformation of landslide.
In the model of simulated annealing with neural where Ymax and Ymin are the maximum and the mini-
work, the network structure is the same as BP neural mum values of monitored data, Xmax and Xmin are the
network, it adopt three layers, but use the simulated normalized values corresponding to Ymax and Ymin .
annealing method to search the weight and critical When the monitored data were forecasted by BP
point threshold instead of the opposite propagation neural network method, the output values must restore
errors in BP neural work method. So, the weight value to the original state as the actual data by using the
and the critical point threshold value is obtained from following formula:
interval of [−1, +1] by using random search method
of simulated annealing, the objective function is:
X − Xmin
Y = Ymin + (Ymax − Ymin ) (3)

m 
m 
q Xmax − Xmin
E= Ek = (ytk − ctk )2 /2 (1)
k=1 k=1 t=1 The symbols in the formula (3) represent the same
meanings as in formula (2).
where ytk is output victors, Y k = (y1k , y2k , . . . , yqk ), The second method is standardizing, the stan-
k = 1, 2, . . . , m, q is the number  of output units; dardizing process is expressed in the formula (4)
ctk = f (Qtk )t = 1, 2, . . . , qQtk = nj=1 vjt bkj − rt t = to (6).
1, 2, . . . , q, vjt is the weight from middle layer to out-
put layer, rt is critical point threshold of output layer,
bkj is activation value of j unit of middle layer, m is the 
N
xi
number of specimen (n) is imported unit. More details i=1
x̄ = (4)
about neural network method can reference to Zhang’s N
paper [Zhang 1983]. )
* N
*
* (xi − x̄)2
4.2 Process of neural network with simulated + i=1
σ = (5)
annealing method for deformation prediction N −1
4.2.1 Analysis of monitored data
xi − x̄
Since affected of monitoring equipments and artificial Si = (6)
factors, the monitored data will be unreal some- Std
times, for these reasons, it is necessary to analyze the
data reality before prediction by using neural network where xi is the import data (the monitored data), i =
method. 1, 2, . . . , N , N is the number of specimen, x̄ is the
mean value of import data, σ is the standard deviation,
4.2.2 Select the specimen of data Si is the standardized value of xi .
The specimen data is used in training process of neural The modeling flow chart of BP neural network with
network method for building the rules of deforma- simulated annealing method for predicting landslide
tion changing, so, in obtaining the reasonable rule, deformation was showed in Figure 7.

682
 Begin

Import the original monitored data

Determining the units of middle layer

Evaluate random value in [-1, 1] interval for weight and threshold value

Select a specimen data randomly to import to neural network

To make better for weight and threshold value by using simulated annealing method

N Finished specimen study

Y
Calculating the overall situation errors E

Y
Output the searched network structure

End

Figure 7. The modeling flow chart of BP neural network with simulated annealing method for predicting landslide
deformation.

4.3 Model of BP neural network with simulated the original monitored data is x(0) , then, the generated
annealing method data under one time totting-up is:
According to the process discussed in the paper above,
the model of BP neural network with simulated anneal- 
t
x(1) (t) = x(0) (m) = x(1) (t − 1) + x(0) (t) (8)
ing method for predicting landslide deformation was
m=1
set up in this paper. In the model, the import para-
meters are (x(t − 1), x(t − 2), x(t − 3), t, L), where
x(t − 1), x(t − 2) and x(t − 3) are real monitored where x(1) (t) is the accumulated deformation value in t
deformations, t is the period of deformation from ini- moment. When cutting the accumulated data in certain
tial monitoring to forecasting time, L is the water time, training is carried out by using the BP neural
level of underground water in t moment, suppose network with simulated annealing method, and, after
the output value is x(t) then, it can be described as the training is finished, the prediction for deformation
function: can be executed.

x(t) = f (x(t−1) , x(t−2) , x(t−3) , t, L) (7)

5 DEFORMATION PREDICTING
For the model described above, there are 3 layers, FOR ‘‘3.29’’ LANDSLIDE
such as import layer, middle layer and output layer,
there are 5 nodes in import layer and 1 node in output Take the monitored data before Sept. 15 of 2003 as
layer, but, for the middle layer, the middle nodes and the training specimen, let the data from Sept. 15 of
the weight and critical point threshold value can be 2003 to Sept. 15 of 2004 as the checking specimen,
obtained by using simulated annealing method. then import the training specimen in BP network with
Additionally, for reducing the affects of random simulated annealing structure, the prediction of defor-
interfere, the import data are not used the original mon- mation for interior deformation was showed in Table 2
itored data directly but adopting totting-up method for and Figure 8 to Figure 9. On comparing the predic-
original monitored data and set in equal intervals. If tion results with the monitored data in site, it is easy

683
Table 2. Comparing with the prediction and monitored data in site of HP1 and HP2.

hp1 hp2

Direction A Direction B Direction A Direction B

Prediction Monitoring Prediction Monitoring Prediction Monitoring Prediction Monitoring Prediction

time value (mm) value (mm) value (mm) value (mm) value (mm) value (mm) value (mm) value (mm)

9-15-04 3.19 3.228 4.415 3.804 4.36 5.104 −2.425 −2.211

10-15-04 1.26 1.15 5.255 6.183 2.89 3.251 −2.975 −3.217
11-15-04 6.92 8.043 6.875 6.722 1.12 1.088 −2.705 −3.017
12-15-04 1.07 0.967 4.165 3.586 1.89 1.573 −2.615 −2.961
2-15-05 1.54 1.449 5.925 5.334 1.89 1.55 −2.265 −2.346
3-15-05 3.63 4.051 4.105 4.59 6.13 7.126 −3.325 −3.251
4-15-05 1.36 1.597 2.895 2.382 3.13 2.669 −3.035 −3.581
5-15-05 1.94 1.966 2.915 2.496 1.3 1.114 −2.545 −2.354
6-15-05 −6.13 −5.129 3.115 3.456 10.32 9.406 −3.175 −3.706
7-15-05 1.84 1.689 4.325 4.744 8.3 7.163 −2.995 −2.959
8-15-05 0.926 5.313 3.769 −6.057
9-15-05 1.594 11.12 8.726 −4.715
Average
error 9.66% 12.21% 13.50% 9.09%

Figure 8. The diagram of deformation predicted with mon- Figure 9. The diagram of deformation predicted with mon-
itored of HP1 in direction B. itored of HP4 in direction B.

6 CONCLUSIONS

to find the maximum error is 16.26%, the maximum The neural network methodology possess the function
absolute error is 0.8 mm, most of the errors are less of dealing with high nonlinearity and parallel calcula-
than 10%, that means the model discussed in this tion problems, the simulated annealing method can be
paper is suitable to forecast the future deformation for used in searching the minimum in overall situation for
landslide. time-sequence problems, so, it is possible to set up the
The results for predicting the interior deformation in model for forecasting the value based on deformation
3 years, 5yeares and 10 years are indicated in table 3. It monitored data in combining with BP neural network
is not difficult to get the long term deformation value and simulated annealing method. As an example, the
from the tables. It can be known that the maximum Tianhuangpin ‘‘3.29 landslide’’ deformation in long-
deformation is 26 mm in July, 15 in 2015, the location term was predicted by using the model discussed in
is in the middle of the landslide, so, the 3.29 landslide this paper, the result is coincident to the real condition
is stable in future 10 years. of the slope.

684
Table 3. Deformation predicted in long term of monitoring site.

Hp1 Prediction value (mm) Hp2 Prediction value (mm) Hp3 Prediction value (mm)

Prediction time A B A B A B

2008-7-15 1.783 8.683 9.046 −3.957 2.275 −12.744

2010-7-15 3.339 10.921 11.506 −5.45 3.514 −14.888
2015-7-15 6.197 16.872 13.476 −8.953 6.728 −18.903
Hp4 Prediction value (mm) Hp5 Prediction value (mm) Hp6 Prediction value (mm)

Prediction time A B A B A B

2008-7-15 7.181 −17.75 13.875 −29.43 1.093 1.285

2010-7-15 10.532 −21.551 17.654 −40.203 1.445 2.909
2015-7-15 17.584 −27.607 25.981 −57.45 2.824 4.14
Hp7 Prediction value (mm) Hp9 Prediction value (mm) Hp10 Prediction value (mm)

Prediction time A B A B A B

2008-7-15 7.253 2.452 −2.765 6.011 −3.764 1.525

2010-7-15 9.338 0.35 −3.236 8.864 −4.205 0.764
2015-7-15 13.358 −1.544 −5.816 13.753 −7.272 −3.396

ACKNOWLEDGEMENTS Liu Hangdong, 1996. Prediction Theory and Method of

Landslide failure. Zhenzhou: Yellow Water Publishing
The research work was financed by Jiangsu Natural House.
Science foundation (No.Bk2006171). Liu Hangdong, 1998. Develop stage of slope failure forecast-
ing, The 5th Rock Mechanics and Engineering Congress
in China, Beijing: Chinese Sciences and Technology
Publishing House.
REFERENCES Sandro Ridolla, Stefano Rovetta & Rodolfo Zunino, 1997.
Circular back-propagation Networks for classification,
Brown, E.T. 1987. Analytical and computational methods in IEEE Transaction on Neural Networks, 8 (1): 84–97.
engineering rock mechanics. John Wiley (eds), New York. Zhang liming. 1983. Artificial neural network model and
Deng Julong, 1985. Grey control system, Wuhan: Huazhong it’s application. Shanghai: Fudan University Publishing
Technology University Publishing House. House.
Deng Julong, 1987. The basic method of grey system.
Wuhan: Huazhong Technology University Publishing
House.
Kang lishan, Xie yun, You siyong etc., 1994. None numer-
ical parallel arithmetic—simulated annealing method.
Beijing: Science Publishing House.

685
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

New models linking piezometric levels and displacements in a landslide

R.M. Faure
Tunnel Study Centre, Bron, France

S. Burlon
Public Works Research Laboratory of Lille, France

J.C. Gress
Company Hydrogéo, Fontaines, France

F. Rojat
Public Works Research Laboratory of Toulouse, France

ABSTRACT: The aim of this article is to present two models linking the piezometric level to the displacements
of a landslide. These two relations are validated with the study of the Petit Caporal landslide (Boulogne-sur-
Mer, France). The obtained results are very correct since both models manage to follow the evolution of the
displacements. These relations stem from two distinct approaches: the first one based on experimental concepts
and the second one based on the study of creep. These models are relevant alert systems, able to detect changes
in the move of the slide. Their easy use encourages to validate them on other sites and to generalise this kind of
studies to other unstable slopes.

1 INTRODUCTION these models and their possible improvements are also

discussed.
The constant lack of free building areas may lead to
set up constructions and infrastructures on unstable
zones. Slow landslides can be dangerous when their 2 THE ‘‘PETIT CAPORAL’’ LANDSLIDE
speed increases, as this phenomenon often announces
an imminent failure. Therefore, models allowing to This landslide does not concern a very important vol-
understand better the behaviour of such slides con- ume of material. It is about 220 m long and 200 m wide
stitute a major issue for both experts and administra- with a regular slope around 5.5◦ . It is delimited uphill
tors. The site of ‘‘Petit Caporal’’ (Boulogne-sur-Mer, by the RN1 road and downhill by the Liane stream.
France) that has been studied for nearly thirty years The global geology appears quite complex. Figure 1
by the technical services of French State is precisely shows the typical longitudinal section that was used to
one of these slow landslides. It is crossed by a national carry out the study.
road (RN1) and by gas and water pipes, which implies Two kinds of geological formations can be identi-
security problems and important financial interests for fied: first the Kimmeridgian stratigraphic formations
the neighbouring population. This landslide is fully (upper Jurassic, secondary era) and on the other hand
tooled up with monitoring systems such as pluviome- the superficial formations of the Liane Valley (quater-
ter, inverse pendulum, piezometers and inclinometers. nary era). To simplify matters, the reader should keep
The data gathered during several years allow build- in mind three main guidelines about the geology of
ing and validating various models linking piezometric the site:
levels to displacement speed. Two kinds of models are
presented in this paper.The first one bases on empirical – the material is globally clayey despite the presence
and statistical considerations as it aims at determin- of two strata with more sandy characteristics in
ing at best the value of two behaviour parameters. which most water flows happen;
The second one rests on more physical considera- – the ‘‘Moulin Wibert’’ clays constitute an imperme-
tions as it insists on creep modelling. The validity of able substratum;

687
 R
N1 (1) “Chatillon” clays
1
Instrumentation
(2) “Chatillon” sands and
siltstones
2
(3) “Moulin Wibert”
limestones and marls
3
(4) “Connincthun”
6

5 4

7 T

Figure 3. Rainfall and piezometric levels between August,

2nd 1999 and May, 7th 2000.
Figure 1. General geology of the ‘‘Petit Caporal’’ site
(drawing not to scale).

c’ (kPa) (˚) (kN/m3)

(1) “Chatillon”clays 20 25 20.4

(2) “Chatillon” sands 10 30 20.2

and siltstones

(3) “Moulin Wibert” 12 19 20.5

limestones and marls

(4) “Connincthun” 5 26 20.5

sands

(5) “Moulin Wibert” 0 8 21

clays (residual)

(6) Uphill soils 0 20 20.2 Figure 4. Rainfall and piezometric levels between August,
2nd 2000 and May, 8th 2001.
(7) Downhill soils 0 25 20.2

Figure 2. Geotechnical characteristics of the landslide. Coming back to Le Petit Caporal, the models pre-
sented in this paper are tested on two observation
periods of 280 days, the first one beginning on August,
– the main surface of failure is located at 6 m depth 2nd 1999 and the second one beginning on August,
at the interface between this clay layer and the 2nd 2000.
Connincthun sands. The rainfall and piezometric readings on both
Movement prediction from rainfall measures is an periods are presented hereafter on figures 3 and 4.
old challenge; a synthesis was done in (Favre et al.,
1992) and very good data were obtained on the full-
scale experimental site at Sallèdes in France (Pouget 3 A FIRST MODEL FOR DISPLACEMENT
et al, 1994). ASSESSMENT
(Vuillet et al., 1996) tested new approach using neu-
ral networks, with a conclusion saying that a great The first model that was used is the one proposed by
amount of data is necessary for building an accurate P. Alphonsi (Alphonsi, 1997) who had carried out sim-
model. ilar research programs about the ‘‘Clapière’’ landslide

688
(Alpes-Maritimes, France). The relationship he had
established was as follows: 80

70
Hj+1

Displacements ( mm/10 )
Vj+1 = Vj exp(−β) (1) 60
Hj
50

with: 40
30
Displacements measured
Vj the displacement speed at day j in mm/day 20 Displacements estimated
Hj the piezometric level at day j in kPa 10
β a coefficient expressing ground sewage.
0

02/08/99

02/09/99

02/10/99

02/11/99

02/12/99

02/01/00

02/02/00

02/03/00

02/04/00

02/05/00
In fact, it appeared that this model was giving
no relevant results in the case of the ‘‘Petit Capo-
ral’’ landslide. Therefore, a second formulation has Observation day
been elaborated:

Hj Figure 5. Comparison between observed and calculated
Vj+1 = α1 exp (2) displacements over the first observation period.
α2

with:
160
Vj+1 the displacement speed at day j + 1 in mm/day 140
Hj the piezometric level at day j in kPa
Displacements ( mm/10 )

α1 a scale parameter in mm/day 120

α2 a piezometric level of reference, for which the 100

displacement speed is minimal. 80

60 Displacements measured
This relationship simply expresses that the sliding
Displacements estimated
speed increases when the piezometric levels are higher. 40
It can be used to evaluate the ground displacements 20
over a few days but not the precise daily speed. Indeed, 0
it appeared too ambitious to build a two-parameters
02/08/00

02/09/00

02/10/00

02/11/00

02/12/00

02/01/01

02/02/01

02/03/01

02/04/01

02/05/01
model only based on piezometric data to express
explicitly the daily speed changes, so equation (2)
Observation day
expresses an average behaviour. The results obtained
are compared with the readings from the inverse
pendulum that provides cumulated displacements.
Figure 6. Comparison between observed and calculated
This formula was used in two steps. First, the obser- displacements over the second observation period.
vation period beginning on August, 2nd 1999 allowed
determining both α1 and α2 parameters. The indicator
that was chosen to assess the quality of the cor-
relation is the average of the absolute values of the correlation shows to be very precise during the first
difference between measured and calculated displace- 100 days, and then a constant drift is observed as if the
ments. Thus, the following values of parameters α1 and slide was reacting differently to piezometric changes.
α2 have been determined: α1 = 0.01 mm/day and α2 = The sudden inflexion in the curve of calculated dis-
20.5 kPa, with = 2.86 mm. As shown in figure placements can be easily explained by a higher—and
5 below, considering that the displacement range is constant—piezometric level, due to important rain-
around 50 mm, the obtained correlation is satisfactory. falls. So the question is to know why the landslide did
In a second step, the above-determined values of not accelerate as well with this increase of the water
parameters α1 and α2 have been applied to the other level. The first hypothesis is that the calculation model
observation period. In this case reaches the value does not use enough parameters to describe correctly
of 12.82 mm: the correlation is not very good, as it the kinematics of the landslide as a whole. The sec-
appears on figure 6. ond hypothesis takes into account that reinforcement
In fact, this second observation period has to be works were realized during the observation period.
analysed more carefully, as it seems to indicate a In fact, two test trenches filled with 7500 m3 of gran-
change in the kinematics of the landslide. Indeed, the ular materials were installed at the bottom of the slide.

689
Due to their position with respect to the slope, these 2002). It describes the three kinds of creep mentioned
trenches have no noticeable influence on piezomet- above and can be written as follows:
ric levels. However, substituting clays and silts with
a sandy material must have increased the global resis- v0 (s) sinh((s1 − s)(t + T ))
tance to shear at the level of the failure surface, slowing δ(s, t) = λ(s) + · (3)
s1 − s (cosh (t + T ))s1 −s
down the observed displacements. This last hypothe-
sis has to be confirmed through the study of a second
with:
calculation method.
Thus, this first simple model, despite its uncertain-
s the tangential stress ratio τ /τmax at the level of the
ties, seems to translate very correctly the interactions
failure surface, with τ = W · sin(α) and τmax =
between piezometric levels and displacements in a
c + (W · cos(α) − u) · tan(ϕ  ) (u being the pore
slow landslide. Its main advantages are its very simple
pressure).
use and its ability to detect changes in the kinemat-
s1 a reference stress ratio expressing the stress state
ics of the sliding area. Validating this model on other
in the ground;
landslides of the same kind would be of great interest.
λ the instant displacement when the material is sub-
mitted to the shear stress τ . In this study, λ was
supposed equal to 0;
4 A SECOND WAY TO ASSESS SLOPE ν0 the initial slope of the creep function;
DISPLACEMENTS, BASED ON CREEP t a time increment;
MODELLING T a time variable expressing ground damage.
The study of creep phenomena usually allows to bring The parameters T and t above have the dimension
out three kinds of creep, each of them being asso- of a time. However, the exact time scale they refer to
ciated with a characteristic strain speed: no speed for has not been determined yet as it would require com-
primary creep, constant speed for secondary creep and plete laboratory creep tests, which was not possible in
regularly increasing speed—until failure—for tertiary this study. For the moment, they can be simply con-
creep (see Figure 7). However some authors (Ter- sidered as time increments necessary to calculation
Stepanian, 1996 e.g.) contest this characterisation and progress. Further research may allow to identify bet-
define four kinds of creep corresponding to four levels ter their real physical meaning and to determine how
of material state and stress conditions. the scale changes between laboratory tests and a real
The creep function that was used in this research landslide should be taken into account. Applying the
had been presented in a previous article (Faure et al., creep function above also requires defining an average
cohesion and friction angle at the level of the failure
surface in order to assess the evolution of the shear
stresses. From the laboratory tests realised on the var-
ious kinds of soils encountered in the ‘‘Petit Caporal’’
slope, the following values were inferred: c = 5 kPa
and ϕ = 25◦ . The soil density is: γ = 21 kN/m3 .
The displacements are estimated with the creep for-
mula (3) by giving t a fixed value and by determining
T and s1 in order to get the best possible correla-
tion. Fixing t comes down to the hypothesis that the
observation (and modelling) period is very short in
comparison with the complete lifetime of the landslide.
It allowed getting round the above-mentioned difficul-
ties on parameters T and t, by using only the transitions
from one kind of creep to another (Y-axis) and not the
time-dependence (X-axis) of the creep function.
To determine the various parameters, the fitting
procedure was similar to the one described in the
previous sections. For the first observation period,
= 2.41 mm is obtained with s1 = 0.27, T = 6 and
ν0 = 0.1. For the second observation period, the same
parametering yields = 12.90 mm, which is quite
deceptive. Figures 8 and 9 give an overview of the
results: it clearly appears that the observed behaviour
Figure 7. The three kinds of creep. is very similar to what had been inferred from the

690
 90
80
Displacements ( mm/10 )

70
60
50
40
30
20 Displacements measured
Displacements estimated
10
0
02/08/99

02/09/99

02/10/99

02/11/99

02/12/99

02/01/00

02/02/00

02/03/00

02/04/00

02/05/00
Observation day

Figure 8. Comparison between observed and calculated Figure 10. Comparison between observed and calculated
displacements over the first observation period. displacements after soil strengthening.

160 logical as equation (3) uses s1 − s. However, it must be

140 emphasized that increasing s1 or increasing τmax refer
Displacements ( mm/10 )

120
to two different physical approaches.
In the first case the reference stress ratio is
100
changed, which means different creep thresholds
80 (primary/secondary/tertiary) and a change in the
60
Displacements measured
behaviour of the soil that deforms less for the same
40 Displacements estimated stress level.
In the second case the creep curves remain the
20
same, but the maximal shear strength is increased,
0 which reduces the stress level τ /τmax . It is likely that
02/08/00

02/09/00

02/10/00

02/11/00

02/12/00

02/01/01

02/02/01

02/03/01

02/04/01

02/05/01

both approaches play a role in the real behaviour of

the slope. This third observation period shows that
Observation day
interesting data can be obtained from such a model
when trying to quantify ground improvement. For
instance, testing the model on a more recent period
Figure 9. Comparison between observed and calculated could have allowed knowing whether the effects of the
displacements over the second observation period. reinforcement works were still perceptible.
Finally, it seems interesting to give additional com-
ments about the definition of the parameters and
first calculation method (equation (2)). The change variables of the creep function. In the case of the ‘‘Petit
in the kinematics of the landslide during the second Caporal’’ landslide, the variable s directly expresses
observation period seems to be confirmed. the influence of the piezometric changes on displace-
In order to know if a change in the parameters ments. It allows determining how the material might be
of the model—in particular s1 and T—could allow a subjected to creep. In the studied cases, the calculated
better fitting, the second part of the second observa- variations of s show that the material does not behave
tion period has been studied as a third observation the same according to the season. When the water
period (i.e. from November 1st, 2000 to May 8th, table is high, s > s1 and tertiary creep is observed.
2001). With the same calculation method as previ- On the contrary, when the water table is low, primary
ously, a very good correlation ( = 1.91 mm) can be creep happens. Besides, the parameter T expresses the
obtained by simply increasing s1 from 0.27 to 0.33. material fatigue. During the 39th Rankine Lecture,
The results are shown in figure 10. This increase S. Leroueil clearly explained this fatigue phenomenon
in parameter s1 clearly indicates an improvement of with loading and unloading cycles when the soil is
ground strength. Quite similar results can be obtained submitted to ground water level variations. The vari-
by keeping s1 constant and by increasing τmax (for ous calculations performed for this study showed that
instance with ϕ = 29◦ instead of 25◦ ), which is the parameter T does not vary much on the observation

691
period, as small changes in T may lead to important the present time, when security and risk management
displacements. This observation is in good agreement become more and more sensitive topics, such research
with the hypothesis that the observation period is rather works should have good prospects.
small in comparison with the total lifetime of the slide.
In any case, finding how T varies in function of the
stress-strain history of the material, even if of great REFERENCES
interest for the improvement of the model, remains
rather complex. Alfonsi P. 1997. Relation entre les paramètres hydrologiques
et la vitesse dans les glissements de terrain. Exemples de
la Clapière et de Séchilienne (France). Revue française de
5 CONCLUSION géotechnique, no 79, p.3–12.
Faure R.M., Gress J.C. & Rojat F., 2002. An easy to use
The validity of two formulas expressing the displace- model for taking in account rainfall in soil displacement.
European Slope Stability Symposium, Rybar, Prague.
ments of a landslide from piezometric changes has Favre J.L., Gevreau E. & Durville J.L. 1992. Prévoir
been tested on the ‘‘Petit Caporal’’ case history. The l’évolution des mouvements de terrain. Revue française
first relationship shows that a simple formula based de géotechnique, no 59, p.65–73.
on intuitive concepts can allow over a few months a Leroueil S. 2001. Natural slopes and cuts: movement and
good assessment of the evolution of the slope. The failure mechanisms. Geotechnique 51, No. 3, p.195–243.
second relationship resting on a creep model yields Pouget P. & Livet M. 1994. Relations entre la pluviométrie et
quite similar results that can be analysed in function la piézométrie et les déplacements d’un versant instable
of various parameters, in particular s and T. (site expérimental de sallèdes, Puy-de-Dôme), Etudes et
Both relations constitute interesting tools for con- Recherches des Laboratoires des Ponts et Chaussées, série
géotechnique, GT 57.
trol or even forecast (when no fundamental change Ter-Stepanian G. 1996. Concentration du fluage avec le
occurs in the landslide behaviour). They might be temps. Revue française de géotechnique, no 74, p.31–43.
used as well to assess the efficiency of reinforce- Vuillet L., Cornu T. & Mayoraz F. 1996. Using neural net-
ment works on slope motions. Curve fitting before works to predict slope movements. Balkema-Rotterdam:
and after reinforcement may yield very useful quanti- Senneset, p.295–300.
tative data concerning the improvement of the global
shear strength. However additional validations on
other unstable sites and through laboratory experi-
ments are still necessary. It must be emphasised that at

692
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

3D slope stability analysis of Rockfill dam in U-shape valley

X.Y. Feng & M.T. Luan

Dalian University of Technology, School of Civil and Hydraulic Engineering, Dalian, China

Z.P. Xu
China Institute of Water Resources and Hydropower Research, Beijing, China

ABSTRACT: By using the method of non-linear elastic-plastic finite element analysis with the software of
ABAQUS, the three dimensional slope stability of an earth core rockfill dam constructed in U-shape valley was
analyzed. The constitutive model of the materials in the analysis is Mohr-Coulomb model and the factor of safety
of the slope is determined by strength reduction method. In the analysis, the development of equivalent plastic
strain of dam slope under different strength reduction factors was presented. The final critical state of slope
failure is defined as the plastic strain zone run through the slope and the sudden changes of equivalent plastic
shear strain and displacement occurred. Compare with the conventional limit equilibrium method, the proposed
strength reduction method can not only provide the similar factor of safety, but also the strain and deformation
changes of the slope. In addition, in 3D analysis, it can also present a clear spatial sliding surface of the slope.

1 GENERAL INSTRUCTIONS factor of strength reduction at this time will be the

factor of safety of the slope (Song, 1997. Griffiths,
The slope stability is one of the most important issues 1999. Dawson, 1999. Manzari, 2000. Zhao, 2002).
in the safety of high rockfill dam. At present time, In recent years, with the development of computation
the conventional methods for conducting slope stabil- technology and the progress in numerical analysis of
ity analysis include: limit equilibrium method, limit geotechnical engineering, many large scale FEM soft-
analysis method, slip line field method, etc. The most ware which are applicable for soil and rock materials
popular method in engineering application is the limit are developed. Most of those software are powerful in
equilibrium method. In general, those methods are pre and post processing, which are helpful in applying
all based on the theory of limit equilibrium of rigid FEM analysis to study the stability of various slopes.
body. It can neither consider the internal stress-strain
relationship of the soil, nor the development of the
slope failure and the interaction of different material 2 COMPUTATION METHOD OF STRENGTH
zones. Besides, it is also necessary to assume the shape REDUCTION
of the sliding surface, such as circle, broken line or
logarithmic spiral, etc. 2.1 Yield criteria
The application of FEM in slope stability analy-
sis has no need to assume soil as a rigid body. It can The yield criteria employed in the paper is Mohr-
not only meet the requirement of equilibrium, but also Coulomb model. It is widely accepted yield criteria in
take into account of the stress-strain relationship of elastic-plastic analysis. The main defect of the model
soil material. It can simulate the procedure of slope is the incontinuous edge points in the three dimen-
failure and the real shape of sliding surface and it is sional stress space. The Mohr-Coulomb yield criteria
applicable in any complex boundary conditions. The is expressed as:
first application of FEM in slope stability analysis was 
started in 1970s. But at that time, due to the limitation 1 1
I1 sin φ − cos θ σ + √ sin θ σ sin φ
of computation conditions, the error of the compu- 3 3
tation is relatively large. It was not widely accepted 
(Zienkiewicz, 1975). By gradually reduce the strength × J2 + c cos φ = 0 (1)
of material in FEM analysis to make the structure
in an unstable status, which may finally lead to the where I1 is the first stress invariant of stress tensor, J2
unconvergence of the iteration of the computation, the is the invariant of deviatoric stress tensor, θσ is Lode’s

693
angle, c is the cohesion, φ is the internal fraction angle Table 1. Material parameters.
(Li, 2006).
In the analysis, the modified Mohr-Coulomb crite- Strength index
ria was developed (Abaqus, 2003), where the devia- Density
toric stress space has no edge point, the flow potential Material E/Mpa υ kg/m3 c/KPa φ/o ψ/o
is totally smooth and only one flow direction can be Rockfill 100 0.3 2200 100 40 20
developed. For general states of stress the model is Core 100 0.3 2160 30 25 12.5
more conveniently written in terms of three stress
invariants as
F = Rmc q − p tan φ − c = 0 (2)
where
1  π
Rmc (, φ) = √ sin  +
3 cos φ 3
1  π 
+ cos  + tan φ (3a)
3 3
 is deviatoric polar angle (Chen and Han, 1988).
The flow potential G is chosen as a hyperbolic func- Figure 1. Mohr-Coulomb yield criteria (p − Rmc q plane).
tion in the meridional stress plane and the smooth
elliptic function proposed by Menétrey and Willam
(1995) in the deviatoric stress plane: reduction can be expressed as (Luan, 2003): under
 the condition of unchanged external load, the factor
G = (εc|0 tan ψ)2 + (Rmw q)2 − p tan ψ (4) of shear strength reduction is the ratio of maximum
shear strength of the soil to the real shear stress of the
where soil. The method of strength reduction is conducted
by dividing the original strength indexes of c, φ with
π 
Rmw (, e) = Rmc ,φ a factor Fs to get a new strength indexes c , φ  . With
3 the new indexes, FEM analysis will be carried out.
4(1 − e2 ) cos2  + (2e − 1)2 By gradually increasing the factor Fs to let the slope
×  finally reach to the critical state, then the last Fs will
2(1 − e ) cos  + (2e − 1) (1 − e2 ) cos2  + 5e2 − 4e
2
be the minimum factor of safety of the slope (Zhang,
(5) 2003). The strength indexes c , φ  can be calculated
by (7) and (8). Elastic modulus E and Possion’s ratio
and ν are kept unchanged during the computation.
π  3 − sin φ
c
Rmc ,φ = (6) c = (7)
3 6 cos φ Fs

tan φ
In the analysis presented in the paper, a certain dila- φ  = arctan (8)
tancy of soil material is considered. If the property of Fs
dilatancy of soil material is not considered (ψ = 0),
the computation results will be too conservative. If the
2.3 Criteria for determining slope failure
dilatancy of soil material is overestimated (ψ = φ),
the deformation will be too large. Therefore, the dila- In the slope stability analysis by using strength reduc-
tion angle is accepted as half of the friction angle tion FEM method, the criteria for determining the
(ψ = φ/2), which may lead to a nonassociated plastic critical state of slope failure is one of the key points. In
flow and unsymmetric stiffness matrix. practice, some researchers use the un-convergence of
calculation as the criteria (Lian, 2001), which means
the un-convergence of the calculation represents the
2.2 The principle of strength reduction method
stress distribution cannot guarantee the general equi-
The idea of strength reduction was first introduced by librium of the slope (Liu, 2005). Ugai (Ugai, 1989)
Zienkiewicz et al. in 1975. It has the same concept as specifies the iteration number of 500 as the limit of un-
Bishop (Bishop, 1955) presented in limit equilibrium convergence. Dawson assumes the ratio of imbalance
method. The definition of the factor of shear strength nodal force to the external load greater than 10−3 is

694
the criteria of slope failure. Lian Zhenyin (Lian, 2001)
considered the character of slope failure should be
thorough development of the generalized shear strain
from the top of the slope to the bottom of the slope.
The appearance and the development of plastic strain
in slope indicate that the occurrence of unrecoverable
residual deformation. Therefore, the development of Figure 4. Distribution of plastic strain (Fs = 1.5).
plastic strain could essentially represent the procedure
of the yield and failure of the slope.
In the analysis presented in this paper, the stability
of slope is judged by the conditions of the distribution
of equivalent plastic strain. If the plastic strain zone
is not yet run through the slope, it means the slope is
still stable. Further reduction of the strength indexes
of soil material will be conducted. If the plastic zone is
run through from the top to the bottom of the slope, it
means the slope will be in critical status. The strength
reduction factor will be defined as the factor of safety Figure 5. Distribution of plastic strain (Fs = 1.9).
of the slope.

3 THE APPLICATION OF STRENGTH

REDUCTION FEM METHOD

According to the computational method presented

above, the 3D slope stability of an earth core rock-
fill dam constructed in U-shape valley was analyzed
by ABAQUS software. The finite element mesh of the Figure 6. Distribution of plastic strain (homogeneous dam).
computation model is shown in Figure 2. The upstream
and downstream slopes of the dam are 1:2 and 1:1.5
respectively. The upstream and downstream slope of
central core of is 1:0.2. The constitutive model of
rockfill material and earth material is Mohr-Coulomb

Figure 7. Distribution of plastic strain (3D, original condi-

tion).

Figure 2. 3D FEM mesh.

Figure 3. Distribution of plastic strain (original condition). Figure 8. Distribution of plastic strain (3D, Fs = 1.5).

695
 lity analysis. Compare with the conventional method,
the development procedures of plastic strain zone in
the slope can be clearly identified. Especially, with the
application of three dimensional analysis, the spatial
sliding surface of the slope can be presented in the
analysis.
In the application of strength reduction method, the
developing of plastic strain zone is suggested to be
the criteria for determining the slope failure. Besides,
the observation of deformation changes of the slope
could also be provided as a secondary judgment.
Figure 9. Distribution of plastic strain (3D, Fs = 1.9).

REFERENCES

ABAQUS. 2003. ABAQUS analysis User’s manual[M]. inc.

Bishop A.W. 1955. The use of the slip circle in the stability
analysis of slopes[J]. Geotechnique, (5):7∼17.
Dawson E.M, Roth W.H, Drescher A. 1999. Slope sta-
bility analysis by strength reduction[J]. Geotechnique,
49(6):835∼840.
Griffiths D.V, Lane P.A. 1999. Slope stability analysis by
finite elements[J]. Geotechnique, 49(3):387∼403.
Li C.Z, Chen G.X, Fan Y.W. 2006. The analysis of slope
stability of strength reduction FEM based on ABAQUS
software[J]. Journal of Disaster Prevention and Mitigation
Figure 10. Distribution of plastic strain (homogeneous Engineering, 26(2):207∼212.
dam). Lian Z.Y, Han G.C, Kong X.J. 2001. Stability analysis of
excavation by strength reduction FEM[J]. Chinese Journal
of Geotechnical Engineering, 23(4):407∼411.
model. The parameters of materials are shown in Liu Z.Q, Zhou C.Y, Dong L.G, etal. 2005. Slope stability
Table 1. The bottom and abutment boundary is treated and strengthening analysis by strength reduction FEM[J].
as fixed boundary. In the analysis, several different Rock and Soil Mechanics, 26(4):558∼561.
Luan M.T, Wu Y.J, Nian T.K. 2003. A criterion for evaluat-
strength reduction factors (1.5, 1.7, 1.8, 1.9) were ing slope stability based on development of plastic zone
selected for conducting the computation. Besides, the by shear strength reduction FEM[J]. Journal of Disaster
results of the cases in homogeneous dam and central Prevention and Mitigation Engineering, 23(3):1∼8.
core rockfill dam are presented. Manzari M.T, Nour M.A. 2000. Significance of soil dilatancy
From the computation results, it can be noticed in slope stability analysis[J]. Journal of Geotechnique and
that: under the original condition, the plastic strain Geoenvironmental Engineering, America Society of Civil
zone mainly distributed in the region of central core. Engineers, 126(1):75∼80.
No continuous plastic strain zone occurred on down- Song E.X. 1997. Finite element analysis of safety factor
stream slope. When the strength reduction factors for soil structure[J]. Chinese Journal of Geotechnical
Engineering, 19(2):1∼7.
reach to 1.5, 1.7 and 1.8, plastic strain zone occurred Ugai K.A. 1989. Method of calculation of total factor
on downstream slope. But the plastic strain zone not of safety of slopes by elaso-plastic FEM[J]. Soil and
run through the slope. When Fs = 1.9, an obvious Foundation , 29(2):190∼195.
circular plastic strain zone developed through the top Zhang L.Y, Zheng Y.R, Zhao S.Y, etal. 2003. The feasibility
to bottom of the slope. Large deformation occurred at study of strength reduction method with FEM for calcu-
dam crest and downstream toe (as shown in Figure 6 lating safety factors of soil slope stability[J]. Journal of
and 10). It can be concluded that the final factor of Hydraulic Engineering, (1):21∼27.
safety of the slope is 1.9. Zhao S.Y, Zheng Y.R, Shi W.M, et al. 2002. Analysis on safety
factor by strength reduction FEM[J]. Chinese Journal of
Geotechnical Engineering, 24(3):343∼346.
Zienkiewicz O.C, Humpeson C, Lewis R.W. 1975. Asso-
4 CONCLUSIONS ciated and nonassociated visco-plasicity in soil mechan-
ics[J]. Geotechnique, 25(4):671∼689.
From the computation results of the elastic-plastic
finite element analysis of the earth core rockfill
dam, it can be found that the strength reduction
method is applicable in three dimensional slope stabi-

696
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

3-D finite element analysis of landslide prevention piles

K. Fujisawa, M.Tohei & Y. Ishii

Public Works Research Institute, Tsukuba, Japan

Y. Nakashima & S. Kuraoka

Nippon Koei co., Ltd., Tsukuba, Japan

ABSTRACT: The landslides prevention piles is one of effective countermeasures against landslides. The
design of landslides prevention piles is usually conducted by 2-dimentional analysis. Recently, 3-dimentional
limit equilibrium analysis such as Hovland method is applied in order to achieve rational design of piles. In
this method, stabilizing force (The force to achieve planed factor of safety) is calculated. Then, this force is
divided by the number of piles in order compute the load that need to be carried by each pile. It is assumed that
stability force is uniformly distributed to the piles. However, landslides have 3-dimentional (3-D) geometries.
For instance, thickness of moving body, slip surface angle and material constants for soils are not uniform.
Therefore, the load that acts on each pile is different and load may exceed the capacity of piles. For this case, it is
possible to apply 3-D finite element analysis that models 3-D geometries of landslide mass and each prevention
pile. This paper presents study of 3-D finite element analysis for simulating restraining mechanism of each piles
constructed in landslide that is triggered by groundwater. Especially, this paper presents modeling techniques for
slip surface using joint element and group of piles with 3-D beam elements. Simulation of actual pile behavior
has been performed to validate the model. Possible application for rational design is also discussed.

1 INTRODUCTION by a beam element. Validation of proposed model was

performed by comparison between analysis results and
Many landslides that require prevention piles in Japan measurement results.
are those with pre-existing sliding surface. In such
cases, three dimensional limit equilibrium methods
such as 3-D Janbu’s Method and Modified Hovland’s
Method are applied to reduce amount of piles by taking 2 MODEL FOR LANDSLIDES PREVENTION
into account 3-D geometry of landslide body. PILES
It is important that how the total prevention forces
obtained from 3-D stability analysis is distributed to 2.1 Outline of analysis model
each pile. One idea is to assume that the load of piles is The model for landslides prevention piles consist of
uniformly distributed. However, the thickness of land- landslide mass, sliding surface, base of landslide, and
slide body is thick at center, and thin in the edge. In piles. The landslide moving mass and base are mod-
addition, the inclination of sliding surface and mate- eled by the solid elements as elastic material. The
rial properties of landslide mass are not uniform. As sliding surface is modeled by the interface (joint) ele-
a result, load of piles is different depending on the ments as elasto-viscoplastic material is described 2.2.
position of piles, in some piles, load may exceed the The piles are modeled by the beam elements as elastic
capacity. material.
For this case, it is possible to apply three dimen-
sional finite element method (3-D FEM), but it is nec-
essary to validate whether the 3-D FEM can simulate
2.2 Model for sliding plane
the real landslide.
The objective of the research described in this The landslide modeled in this research is the type
paper is to develop a practical three dimensional finite where soil and rock mass move as a single mass over
element model for active landslide. A simple elasto- the pre-existing sliding surface. It is therefore assumed
viscoplastic interface model for sliding surface was that the landslide mass itself does not fail and modeled
developed, and each prevention piles were modeled by elastic solid elements.

697
 It is thought that the finite element of the sliding
surface is solid elements and interface (joint) element.
In this paper, interface (joint) element was chosen
for the reasons described as follows.
There are two types of stress concepts, one is the
surface traction on the surface of a body, and another
is the internal stress in a body.
The former is the normal direction stress (σn )
on the surface and shear stress (τs ), and they are
called surface traction. On the other hand, the latter
is defined as tensor with six components expressed
(σx , σy , σz , τxy , τyz , τzx ), and they are called internal
stress.
Stresses that control failure of sliding plane are the
surface traction and therefore it is necessary to use
joint elements. If solid elements are used, the direction
of failure will be different from that of the prescribed
sliding surface as described below. The direction of Figure 2. 16-node hexahedron joint element.
failure base on internal stresses is determined by the
Mohr’s circle and the failure envelope (Figure 1).
Let us suppose that the direction of principal stress
is equal to the direction of gravity. Then we can esti- the direction of failure is equal to the direction of slid-
mate the direction of failure. The direction of this ing surface because the direction of plane of the joint
surface is not necessary in the direction of sliding sur- element corresponds to sliding surface.
face. Especially, if the inclination of sliding surface is It is thought that a solid element is suitable if the
gentle, the direction of failure of a solid element may purpose is to simulate the process formation of sliding
be significantly different from the direction of sliding surface. However, when the sliding surface has already
surface. On the other hand, in case the joint element, been formed, assuming that the large deformation take
place, it is more suitable that the sliding surface is
modeled by joint element.
We have been studying the joint element with liner
shape functions. If the sliding surface geometry is
simple, it is not especially a problem. However large
number of joint elements may be required to gain better
solution for soil-pile interaction.
So in this paper, modified version of the joint ele-
ment is introduced. The element has midside nodes
and has quadratic shape functions.
As an example, natural coordinates (ξ η) is intro-
duced for numerical integration of 16-node hexa-
hedron joint elements. Shape functions of eight-node
plane serendipity elements are expressed as:

N1 = 1/4 (1 − ξ ) (1 − η) (−1 − ξ − η)
N2 = 1/4 (1 + ξ ) (1 − η) (−1 + ξ − η)
N3 = 1/4 (1 + ξ ) (1 + η) (−1 + ξ + η)
N4 = 1/4 (1 − ξ ) (1 + η) (−1 − ξ + η)

N5 = 1/2 1 − ξ 2 (1 − η)

N6 = 1/2 (1 + ξ ) 1 − η2

Figure 1. The schematic diagram relationship between the N7 = 1/2 1 − ξ 2 (1 + η)

shear failure plane based on internal stress and the angle of

sliding surface. N8 = 1/2 (1 − ξ ) 1 − η2 (1)

698
 
This shape function is introduced to translate phys- F= τrs2 + τsr2 + σn tan φ − c (6)
ical coordinates to natural coordinates.
The relative displacement between the upper sur-
face and lower surface of the joint element is given by F0 = c · tan φ (7)

where τrs and τsr is shear stress for joint element, σ n is

{W } = [B] {U } (2) normal stress act on joint element plane, φ is friction
angle.
The plastic potential Q is expressed as,
where {W } = {δu δv δw}T is the relative displacement
of 8-node, and [B] is the matrix of 3 by 48 that trans- 
lates displacements vector {U } of 8-node to relative Q= τrs2 + τsr2 (8)
displacements.
The stiffness matrix of the joint element is given by Compression is negative such that F > 0 in case
of shear failure. For tensile failure mode, the yield
 and plastic potentials, and reference value for normal
[K] = [B]T [D] [B] dA (3) direction are formulated in terms of tensile strengths.
A The yield and plastic potentials for the shear direction
are the same as Eq. (6) and (7) except that angle of
friction and cohesion are set to zero.
where [D] is constitutive matrix for joint element.
Also, we introduced 12-node penta type element to
model more general shape of landsides. 2.3 Model for prevention piles
The elasto-viscoplastic model is based on the The beam element is introduced for the model of pre-
fundamental theory developed by Perzyna in 1966, vention pile. We can choose other type pile model
and it is extension of the two dimensional model composed of solid elements, if it necessary to assess
introduce by Sekiguchi (1990) with modification of detailed stress distribution around the piles.
yield and plastic potential functions. The model per-
mits both shear and opening failure modes, while
compression failure is not allowed. When the inter- 2.4 Model for sliding mechanism
face opens due to tensile failure, displacements The judgment of shear failure of joint element is per-
in shear are permitted since there is no shear formed by subtract the pore water pressure from the
resistance. normal stress.
When there is no failure, elastic analysis is per- The landslide sliding mechanism is expressed by
formed. If failure takes place, viscoplastic flow increasing the groundwater level at slopes.
rule applies. The viscoplastic flow rule is defined
for relative displacement, uvp in local coordinates,
since stresses are functions of relative displace- 3 APPLICATION
ments,
3.1 Outline of the area and measurement
3  4
1 2 F ∂Q
u̇vp =γ  (4) Validation of proposed model for landslide and preven-
F0 ∂σ tion pile is performed by comparison between analysis
and measurement results.
Where γ is the fluidity parameter, Q is the plastic The Arahira Landslide is located in Miyazaki Pre-
potential, F is a yield function, F0 is a reference value, fecture in western Japan. It is 120 m long (south-north
and  is a scalar flow function for x > 0 and the direction) and 120 m wide (east-west direction). The
notation <> is defined as, maximum depth of moving soil mass is about 19 m
(Figure 3 and 4).
The geology is composed of phyllite from the
(x) = (x) (for x > 0) Makimine Formation in the Morozuka Group.
(x) = 0 (for x ≤ 0) (5) Figure 4 is the section of the landslide due to field
investigation and exploratory borings.
The sliding surface shape is plane towards the toe
The rate of relative displacement mainly depends of landslide.
on the choice of functions Q and . When the interface The counter measure such as drainage well and
is in shear failure mode under compression, the yield prevention piles were designed as the target factor
function, F is taken as, of safety 1.15 (the increase with piles is 0.106).

699
These were constructed from March through August 3.2 Analysis model and procedure
of 2000.
Finite element mesh of Arahira landslide is shown
Strain gauges were attached to the pile by 1 m pitch
Figure 6, in which the sliding plane is modeled by
in the direction of depth to measure bending moment.
the joint elements described 2.2, whereas rest of the
The measurement is conducted at the frequency about
domain is modeled by 10-node tetrahedron solid ele-
once a Month. Also, the extensometer (S-1, S-2) and
ments as elastic material, the prevention piles was
the groundwater level meter (W-1, B11-2) were set up.
modeled by beam element. All of the mesh was made
Figure 5 shows the results.
based on borehole investigation and topographical
The bending moment of pile increased rapidly
measurement.
from May 19, 2000 (measurement was started)
The validation analysis was performed in two steps.
through October 13, 2000. In the same period,
In the first step, initial stress condition was generated
the extensometer (S-2) recorded the displacement of
under gravity loads where the ground water level was
4.5 mm, and the groundwater level meter (W-1, B11-2)
set to low water level (L.W.L), and piles were not gen-
recorded increase of water level. The rainfall of total
erated at this step. Subsequently, in the second step, the
671 mm was recorded from September 8, 2000 to
piles were generated and the ground water level was
September 16. Thus, it is thought that the landslide
set to high water level (H.W.L) to simulate landslide
was caused under the influence of increased ground
by inducing plastic failure in joint elements.
water level, and load acted on the prevention piles.
The validation analysis was performed by simu-
The pressuremeter test was performed near the pile
lating behavior of piles from May 19, 2000 through
by 1 m pitch in the direction of depth to investigate
October 13, 2000 where ground level increased by
deformation property of geomaterial (Table 1).
3.4 m (from May 27, 2000 to June 20, 2000 at W-1).

3.3 Analysis case and parameter

The parameter for analysis, such as cohesion and the
angle of internal friction, modulus of deformation, unit
weight, Poisson’s ratio is shown Table 2.
Unit weight and Poisson’s ratio was set to typical
value, the angle of internal friction of joint elements
was back-calculated and by assuming that F.S. of slid-
ing plane is 1.0 with 212.5 m of ground water level on
September 5,2000 (W-1).
In the analysis, F.S. was calculated as the ratio of
sliding force which is summation of each joint element
and resistant force.
The fluidity parameter γ in Eq. (4) was set to 0.025.
The modulus of deformation was set to the mean
value of the pressuremeter test (Table 1) from sliding
Figure 3. Plan view of Arahira Landslide and location of plane to distance of 1/β in the direction of depth.
instruments, piles.

Figure 4. Sectional view of Arahira Landslide.

700
 Bird’s-eye view
landslide mass

node :38630
element :26760 sliding surface

base

116.5 m

229 m

z
y 120.2 m

250 m
x landslid mass base

Piles and sliding surface

sliding surface

10-node solid
piles
sliding surface

12-node joint

Figure 5. Measurement results. pile

base

Table 1 Results of the pressuremeter test.

Depth Geology Classification Modulus of deformation Figure 6. Finite element mesh of Arahira landslide.
(m) E0 (kN/m2)
1 Colluvium- 2682
2 5090 Table 2. Parameter of the validation analysis.
3 4305
4 3503 Angle of
5 5022 Unit internal Poisson’s
6 D 9085 weight Cohesion friction ratio
7 8470
8 9297 Landslide
9 8429
mass 18 kN/m3 - - 0.3
10 7717
11 106890 Sliding -
12 27700 surface 19.0 kN/m2 12.17◦ -
13 Phyllite 18640 Base 18 kN/m3 - - 0.3
14 CL 14750
15 73850
16 20070
17 555090
18 Sliding surface D 69160 Where kh is coefficient of horizontal subgrade reac-
19 22580
20
CL
17230
tion, d is diameter of piles, E is Young’s modulus
21 D 24710 of steel, I is geometrical moment of inertia for piles
22 155390 section.
23 CM 79840
24 129420
3.4 Results of analysis
The displacement and moment distribution in depth
The value of β which decides the property of piles of measured pile is shown in figure 7. The bending
is given by moment from FEM analysis shows a good agree-
ment with the measurement results. From this figure,
5 the bending moment distribution shows similar peak
kh · d moments at upper and lower depth from sliding plane.
β=
4
(9)
4·E·I This tendency agrees with measurement results.

701
Figure 7. Displacement and moment of measured pile.

Figure 9. Displacement contour of sliding plane from FEM

analysis when there is no pile.

Figure 8. The maximum moment distribution of each pile.

C

The maximum moment distribution of each pile

from FEM analysis and measurement results (pile
No.17) are shown in figure 8. The maximum moment
is different in each pile. In the left side area of land-
slide block, the maximum moments of piles are large.
On the other hand, in right side area of landslide block, D
the maximum moments of piles are relatively small.
It is thought that this trend is due to the differences in
the thickness of the landslide. However, this reason can
not explain the fact that the bending moments of the
piles No.1 to 5 are relatively large, where the thickness
of left side area is relatively thin.
The reasons for the high moment of piles on the
left region may be related to the inclination of sliding E
plane. Figure 9 shows displacement contour of sliding
plane from FEM analysis when there is no pile, and
figure 10 shows displacement contour of each section.
From these figures, the inclination of sliding plane
of the left side is steeper than that of the right side,
causing the left side of the displacement to be larger
than that of the right side. Figure 10. Displacement contour of each section.

702
4 CONCLUSIONS Kuraoka, S., Ota, K. & Nakashima, Y. 2003. International
Conference On Slope Engineering, Vol. II, pp. 834–839.
This paper presents modeling techniques for slip sur- Owen, J & Hinton, E. 1980. Finite Elements Plasticity;
face using joint element and group of piles with 3-D Theory & Practice, Pineridge Press, pp. 272–318.
beam elements. Simulation of actual pile behavior has Sekiguchi, K., Rowe, R.K. & Lo, K.Y. 1990. Time step
selection for 6 noded non-linear joint element in elasto-
been performed to validate the model. viscoplasticity analyses. Computers and Geotechnics,
It is thought that the 3-D finite element method Vol. 10, pp. 33–58.
enables more reasonable design of piles because it can Desai, C., Samtani, N.C. & Vulliet, L. 1995. Constitutive
consider 3-D effects. modeling and analysis of creeping slopes. Journal of
Geotechnical Engineering, Vol. 121, No. 1, pp. 43–56.
Miura, F., Okashige, Y. & Okinaka, H. 1985. Rupture prop-
REFERENCES agation analysis by the three dimensional finite element
method with joint element. Research report University
Ghaboussi, J., Wilson, E.L. & Isenberg, J. 1973. Finite ele- of Yamaguchi, Dept. of Engineering, Vol. 36, No. 1,
ments for rock joints and interfaces, Jl. Soil Mechs. Dn., pp. 81–87.
ASCE, 94, SM3.
Goodman, R.E., Taylor, R.L. & Brekke, T. L. 1968. A model
for the Mechanics of Jointed Rock. Proc. ASCE, 94 SM3,
pp. 637–659.

703
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Integrated intelligent method for displacement predication of landslide

W. Gao
Institute of Rock and Soil Mechanics, The Chinese Academy of Sciences, Wuhan, China

ABSTRACT: Displacement predication of landslide is very important in the control of landslide disaster.
Considering the monotonously increasing character of time series of the landslide displacement, a new intelligent
prediction method ton combine the Grey System and the Evolutionary Neural Network (ENN) is proposed here.
On the basis of the principles of displacement decomposition, the trend of time series is extracted by the Grey
System, while the deviation of the Grey System is approximated by the new ENN proposed. The architecture
and algorithm parameters in the new ENN can evolve simultaneously through the modified BP algorithm and
Immunized Evolutionary Programming proposed by the author. This new method is applied in the study of Xintan
landslide, and the results show that the new method is good in generalization and can predict the displacement
of landslide very well.

1 INTRODUCTION fluctuant section, et al. For different sections, different

methods should be taken. But in previous studies
Landslide is a very serious geological disaster. For using neural network, this problem have not been men-
there are lot of mountains in the west of china, as the tioned. To solve this problem very well, in this paper,
progress of West Development Project in China, more an intelligent method combining Grey System and
and more landslide disasters will be encountered. So, Evolutionary Neural Network is proposed.
how to control the landslide has become a very impor-
tant work. To control landslide disaster, the forecasting
of landslide is a very powerful method. But the devel- 2 NEW INTELLIGENT METHOD
opment of landslide is a very complicated dynamic FOR LANDSLIDE PREDICATION
procedure. To describe this system very accurately is
very hard. But the measured displacement series can 2.1 Division of time series of the landslide
describe the general laws of landslide development. displacement
So, some methods (Feng 2000, Gao & Zheng 2000,
Huang 1999, Liu & Fan 1992, Shi & Xu 1995) for Supposing ui is measured time series of landslide, it
displacement predication of landslide are proposed. can be divided as follows,
From analysis of those methods (Feng 2000, Gao
2002), we can find that, the neural network method ui = usi − vi (1)
and evolutionary neural network method can describe
displacement series more accurately and more easily, where usi is the trend section of displacement time
and are two better methods. For predication problem, series; and vi is the deviation section.
the generalization of neural network is very important. The previous studies show that (Liu & Fan 1992),
From the theory studies on generalization of neural the trend of landslide displacement time series can be
network, the precision of inner interpolation can be described by grey system very well. But the devia-
ensured, but the error of outer interpolation is very tion section is still a very complicated time series.
large. For predication of one monotonously increas- For describing this time series, the evolutionary neural
ing time series, the outer interpolation is need. So, network is a very suitable method.
for predication of one monotonously increasing time
series, the neural network is not very good. But the
2.2 Grey system model for trend section
landslide displacement is one monotonously increas-
ing time series, so to predict this time series, the new Generally, the trend section of landslide displacement
method must be proposed. Generally, the time series of can be concluded into two types. One type is a kind
the landslide displacement can be divided into some of monotonously increasing curve that is a concave
sections, such as, even section, periodic section and line. To describe this curve, the GM (1, 1) model in

705
grey system is a very suitable one. Another type is 2. One network construction is generated by two ran-
an S shape curve. To describe this curve, the Verhulst dom numbers in search range of input neuron
model or DGM (2, 1) model in gery system is an very and hidden layer neuron. And also, one kind of
suitable one. MBP algorithm is created by two random num-
Here, the grey system is only to extract the trend, bers in search range of parameters η and α. And
so its precision is not exigent. And the only objective then, one individual can be generated by the four
is that the deviation is not a monotonously increasing parameters.
curve. The details of the grey system can be found in It must be pointed out that, for two numbers of
reference (Liu et al. 1999). neuron are integer numbers and two MBP param-
eters are real numbers, so the expressions of one
individual must be structural data.
2.3 Evolutionary neural network model 3. To one individual, its fitness value can be gotten by
for deviation section follow steps.
As in above description, the precision of grey system a. The whole time series of landslide displace-
is not high, so the deviation section is still a very com- ment is divided to construct the training samples
plicated series. Even if the precision of grey system based on number of input neuron and number of
is high, it is very hard to guarantee that the deviation hidden layer neuron. And also, the total number
section is a simple random series. So, to describe the of samples is noted.
deviation section very well, the neural network is very b. The whole learning samples are to be divided
suitable. into two parts. One part is the training samples,
So, to construct a neural network model for mod- which is to get the non-linear mapping network.
eling displacement time series, the construction of The other part is the testing samples, which is
neural network is the main problem to be solved. to test the generalization of network.
Because in this problem, the hidden layer construction c. The initial linking weights of network individual
and input layer construction all must to be confirmed. are generated randomly.
This problem can be solved by evolutionary algorithm d. The iterating step of MBP algorithm is taken as
very well. Here, as a primary study, the evolutionary j = 1.
neural network which construction is confirmed by e. This network individual is trained by testing
evolutionary algorithm and which weight is confirmed samples, and the square error E( j) is computed,
by MBP algorithm is proposed. To make problem sim- and this error is taken as minimum error of
pler and generalization bigger, the three layers neural the whole training, min E = E( j). If minE is
network is studied. So, here, only the number of input smaller than the error criteria of evolutionary
neuron and number of hidden layer neuron are to be algorithm, then the fitness value is minE. And
confirmed. In MBP algorithm, there are two para- the computing process is transferred to step (3).
meters, iterating step η and inertia parameter α, to f. This network individual is trained by training
be confirmed. These two parameters affected MPB samples. If its training error is smaller than
algorithm very seriously. So, these two parameters are iterating error criteria of MBP algorithm, then
all confirmed by evolutionary algorithm. And then, in the fitness value is also the minE. And the
evolutionary neural network, there are four parameters computing process is transferred to step (3).
to be evolved. In order to get the better effect, the new g. The whole linking weights are adjusted by MBP
evolutionary algorithm- immunized evolutionary pro- algorithm.
gramming (Gao & Zheng 2003) proposed by author is h. j = j + 1, and the computing process is
used in evolutionary neural network. transferred to step e.
The details of this new evolutionary neural network i. If j is lager than iterating stop criteria of MBP
are given as follows. algorithm, then the fitness value is also the
minE. And the computing process is transferred
1. The search range of input neuron and hidden layer to step (3).
neuron are given firstly. And also the search ranges
of two parameters in MBP algorithm are given. And 4. If the evolutionary generation reaches its stop cri-
some evolutionary parameters, such as evolution- teria or computing error reaches error criteria of
ary generation stop criteria, individual number in evolutionary algorithm, then the algorithm stop. At
one population, the error criteria of evolutionary this time, the best individual in last generation is the
algorithm, number of output neuron in neural net- searching result.
work, iterating stop criteria and iterating error 5. Every individuals in population are mutated. For
criteria in MBP algorithm are all given. there are different data types in one individual, the
It must be pointed out that, to construct the suit- different mutation types are used for each parame-
able samples, the number of input neuron must be ter. For numbers of input neuron and hidden layer
smaller than total number of time series. neuron are integer number, the uniform mutation

706
 Generate training samples

Generate network weight

Iterating step in BP, j=1

Calculated error of testing samples, E

Yes
E< evolutionary criteria E is fitness

No
Calculated error of training samples, E 1

Yes
E < error criteria
1
E is fitness

No
Adjust network weights by BP

j=j+1

Yes
j > iterating criteria E is fitness

No

Figure 1. Flow chart of evolutionary neural network.

Parameters confirmed

Initial population of network construction generated

Calculated fitness value

Generated Initial population

Individual mutation

Calculated individual fitness

Immunized selection

Evolutionary generation increase

No
Evolutionary generation bigger than criteria
Yes
End

Figure 2. Flow chart of calculated error E.

707
Table 1. Measured displacement and calculated displacement of Xintan landslide.

Training
Time Measured Trend of Deviation samples of Generalized Calculated
step ones GM (1, 1) section ENN results results

1 0.077
2 0.092 0.414 0.322 0.677
3 0.615 0.468 −0.146 1.146
4 0.65 0.529 −0.120 1.120
5 0.69 0.599 −0.091 1.091
6 0.738 0.677 −0.091 1.060
7 0.846 0.766 −0.080 1.080
8 0.962 0.866 −0.096 1.096
9 1.0 0.979 −0.020 1.020
10 1.03 1.108 0.078 0.922
11 1.061 1.253 0.192 0.808
12 1.077 1.417 0.34 0.660
13 1.1 1.602 0.502 0.498
14 1.23 1.811 0.581 0.4184 0.460 1.267
15 2.46 2.048 −0.411 1.411 1.382 2.431
16 2.754 2.316 −0.437 1.437 1.359 2.675
17 2.83 2.619 −0.210 1.210 1.268 2.888
18 2.92 2.962 0.042 0.957 1.087 3.049
19 3.46 3.350 −0.109 1.110 1.116 3.467
20 4.00 3.788 −0.211 1.211 1.135 3.923
21 4.25 4.284 0.034 0.966 0.826 4.110
22 4.38 4.844 0.464 0.535 0.559 4.403
23 4.615 5.478 0.585 5.064
24 5.77 6.195 0.804 5.999

is used. For parameters η and α are real numbers, 7

the adaptive Cauchi mutation is used. And then the Measured diaplacement
6
offspring population is generated.
displacement/mm

6. The fitness value of each individual in offspring Computing displacement

5
population is calculated by the method in step (3).
4
7. The set of offspring population and parent popu-
lation is selected by selection operation based on 3
thickness, then the new offspring population is
2
generated.
8. The number of evolutionary generation increases 1, 1
then the computing process is transferred to
0
step (4).
1 6 11 16 21
time step
From the above algorithm, the four parameters,
number of input neuron, number of hidden layer neu-
ron, two parameters η and α in MBP algorithm can Figure 3. Measured displacement and computing displace-
ment of Xintan landslide.
be confirmed. So, the optimization neural network for
deviation section of landslide displacement series can
be gotten.
The flow charts of evolutionary neural network are
as Figures 1 and 2. The displacement time series on key measured point
of A3 is showed as in follow Table 1.
The measured displacements can also be showed
3 APPLICATION OF NEW INTELLIGENT in Figure 3. From the Figure, we can see that, the
METHOD IN REAL ENGINEERING trend of displacement is a kind of concave exponen-
EXAMPLE tial curve. To model it, the GM (1, 1) model is a
suitable one. Also, to test the forecasting capability
To verify the above algorithm, the displacement time of our algorithm, the data at step 23 and 24 cannot
series of Xintan landslide (Huang 1999) is used. be used.

708
 According to the model process of GM (1, 1), the forecasting very well. In this method, based on the
GM (1, 1) model of trend section can be gotten as principles of displacement decomposition, the trend
follows. of displacement time series is extracted by Grey Sys-
 tem and the deviation of Grey System is approximated
0.379826 by the new ENN. So, the different method is used to
x̂(1) (k + 1) = x(0) (1) − e0.122959k
−0.122959 model the different section of displacement. And the
whole forecasting capability is very well. At last, one
0.379826
+ (2) real engineering example is used to verify this new
−0.122959 algorithm, and the results are very well.
From above model, the trend series of landslide
displacement can be gotten. And also the deviation
section series can be gotten. The two series are all REFERENCES
showed in Table 1. By some pre-disposal to devia-
tion section series, the samples of ENN can be gotten, Feng, X.T. 2000. Guide to intelligent rock mechanics.
which is also showed in Table 1. Beijing: Science Press.
After computing, we can get the follow results. Gao, W. & Zheng, Y.R. 2000. Study on Some Forecast-
The number of input node is 12. The number of hid- ing Methods in Geo-technical Engineering. In Proc., 6th
ing node is 14. The parameters in MBP algorithm are Conf. of Chinese Rock Mechanics and Rock Engineering
η = 0.226 and α = 0.948. Society: 90–93. Beijing: Science Press.
Gao, W. 2002. Study on neural network method for displace-
By the above results, the computing displacements
ment forecasting in geo-technical engineering. Geotech-
can be gotten as in Table 1 and Figure 3. nical Engineer 14(1): 8–12.
From the above results, the follow conclusions can Gao, W. & Zheng, Y.R. 2003. An New Evolutionary
be drown. The new intelligent algorithm proposed in Back Analysis Method in Geo-technical Engineering.
this paper can reveal the essential rule of landslide J. Chinese Rock Mechanics and Rock Engineering 22(2):
displacements. It has not only the good approximated 192–196.
capability, but also the good forecasting capability, Huang, Z.Q. 1999. Study on non-linear mechanism of slope
and is a very good method for modeling the landslide evolution and forecasting of landslide. Beijing.
displacement. Liu, H.W. & Fan, J.W. 1992. Forecasting and analysis of
landslide displacement based on Grey System theory.
J. Chengdu Science and Technology Univ. (2): 57–64.
Liu, S.F., Guo, T.B., Dang, Y.G., et al 1999. Grey system
4 CONCLUSIONS theory and its applications. Beijing: Science Press.
Shi, Y.S. & Xu, D. J. 1995. Application of time series analysis
The new intelligent algorithm proposed in this paper method in slope displacement forecasting. Rock and Soil
can solve the problem of landslide displacement Mechanics 16(4): 1–7.

709
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

A new approach to in Situ characterization of rock slope

discontinuities: The ‘‘High-Pulse Poroelasticity Protocol’’ (HPPP)

Y. Guglielmi, F. Cappa, S. Gaffet & T. Monfret

Geosciences Azur, Sophia Antipolis, France

J. Virieux
LGIT, Grenoble Cedex 9, France

J. Rutqvist & C.F. Tsang

LBNL-ESD, Berkeley, CA, USA

ABSTRACT: The High-Pulse Poroelasticity Protocol is an alternative approach for in situ rock-slope-
properties estimation. It relies on an innovative probe that allows simultaneous pressure and deformation
measurements in boreholes. The method consists of short-duration hydromechanical pulse tests to estimate
local hydraulic and mechanical properties of fractures (normal, shear stiffnesses and permeability). Then, a
long-term injection induces a large slope deformation, measured at the injection point with the HPPP probe and
in the near field with tiltmeters. Fully coupled, hydromechanical, numerical elastic models are then used to match
all pressure, deformation, and tilt measurements by adjusting discontinuity properties. Applied to a rock slope
of fractured limestone, the Coaraze slope in southern France, the method evidenced a hyperbolic relationship
between hydraulic apertures and stiffness of fractures, and enabled an estimation of fracture compressive and
shear strengths. Compared to other approaches based on joint roughness analyses, this approach seems less
subjective and more accurate.

1 INTRODUCTION the derived Mohr-Coulomb parameters of the jointed

rock slope.
Complex interactions among numerous pre-existing This approach is an efficient method for a first
discontinuities primarily determine the potential areas estimation of jointed distinct elements and continuum
of instability in fractured rock slopes. However, the numerical analyses, to assist in the design of tunnel and
stability analysis and monitoring of any rock struc- cavern reinforcement and support. However, applying
ture faces the major problem of determining both the this method to rock-slope stability analyses is less reli-
properties and constitutive stress-displacement mod- able, because the state of stresses close to the surface is
els characterizing the discontinuities. Several criteria highly heterogeneous and with low values, making the
have been proposed to identify the strength of a natu- estimation of Mohr-Coulomb parameters (c, cohesion,
ral rock joint (Patton, 1966, Ladanyi & Archambault and φ, friction angle) from the derivation of the non-
1970, Barton & Choubey 1977, Amadei et al., 1998, linear Barton’s equation less than precise. Moreover,
Saeb & Amadei 1992, Plesha, 1987). In practice, conversion of JRC numbers into the ISRM rough-
Barton’s model is assumed (ISRM, 1978), and in gen- ness descriptions is subjective and unambiguous only
eral used based on parameters describing the joint for some roughness profiles (Barton & Bandis 1990).
morphology called the JRC (Joint Roughness Coef- Measuring joint roughness also requires separating the
ficient) and the JCS (Joint Compressive Strength). discontinuity surfaces, which causes cohesion break,
These aparmeters are measurable on joint laboratory damage, and mismatching of the two surfaces, all of
samples and in situ through profiling methods, the which can alter any estimation of joint properties.
Schmidt hammer method (Aydin & Basu 2005) and Moreover, the mechanical and hydraulic proper-
visual estimation (Laubscher, 1990). They are then ties of joints are estimated independently. Coupled
related to joint strength, stiffness, and aperture through hydromechanical effects are seldom integrated into
empirical relations (Bandis et al. 1983), and intro- slope stability analyses, even though water is often
duced into stability models using the Barton-Bandis or mentioned as a triggering mechanism for failure in

711
rock with well-developed preexisting fractures net-
works (Erisman & Abele, 2000). The mechanisms
operating in the discontinuity behavior are complex,
and involve fully coupled hydromechanical effects
(Rutqvist & Stephansson, 2003). Compressive effec-
tive normal stresses (σn = σn − Pf ) press opposing
discontinuity walls together, resisting sliding motion
along the discontinuity surface possibly induced by
shear stresses (τ ) acting parallel to the discontinu-
ity plane. A reduction in the effective normal stress
state leads to the normal opening of discontinuities,
inducing a reduction of the inherent shear strength of
these discontinuities. When the slope discontinuities
are deformed, their hydraulic properties are modi-
fied, mainly through dilatancy and the crushing of
discontinuities asperities (Gentier et al. 2000). As a
consequence, there is a close relationship between the
way mechanical and hydrauli properties evolve within
discontinuities. This relationship has been studied in
the laboratory (Li et al. 2007, Gentier et al. 1997, Ols-
son & Brown 1993 for examples) but hardly ever in
the field.
We present the first results from an innovative geo-
physical method for in situ quantification of hydraulic
and mechanical properties within fractures, through
coupled pressure/deformation measurements in bore-
holes combined with surface tiltmeter measurements.
This method, called the High-Pulse Poroelasticity
Protocol (HPPP), relies on high-accuracy hydrome-
chanical pulse injections analyses, which enable in Figure 1. (a) Three-dimensional view of Coaraze fractured
situ estimation of joint properties. This method is slope; (b) Pole plots showing brittle faults and bedding-plane
orientations (lower hemisphere); (c) JRC values of slope’s
then applied to characterize the stability of a fractured fractures.
slope. Specifically, we show (1) the principles and the
sensitivity of the HPPP test, (2) an example of an in situ
estimation of joint properties in a natural slope, and (3)
ridges (low values in dashed circle, Figure 1c) and
the results in terms of joint strength estimation.
perpendicular to that direction (high values in con-
To evaluate the usefulness of the HPPP method,
tinuous circle in Figure 1c). Joint’ surfaces are not
it was applied to the Coaraze site on the lower 15
weathered.
m section of a 40◦ –60◦ dipping slope comprised of
a thick sequence of fractured limestone (Figure 1a).
The slope is cut by 12 parallel bedding - planes, with
a 40◦ trend dipping 45◦ SE, and two sets of approxi- 2 THE HIGH PULSE POROELASTICITY
mately orthogonal, near-vertical faults, with 50◦ /70◦ PROTOCOL
trends dipping 70◦ to 90◦ NW and 120◦ /140◦ trends
dipping 75◦ to 90◦ NE (Figure 1b). The mechanical The ‘‘High-Pulse Poroelasticity Protocol’’ (HPPP) is
properties of the rock matrix, previously determined a new in situ approach developed for a very large
from laboratory testing, indicated Young’s modu- broadband geophysical monitoring of rock HM defor-
lus values ranging from 44.4 to 70 GPa, Poisson mations within boreholes. In this protocol, the rock
ratios of 0.29 to 0.34, and intact-rock permeabil- is subjected to a controlled source corresponding to a
ity of 10−17 m2 . Nine corings were performed at fast (i.e. few seconds) hydraulic pressure pulse local-
different locations in the slope to investigate local ized within a short injection chamber (0.4 to 3 m),
fracture properties. There is a broad range of joint isolated between two inflatable packers in a bore-
surface JRC values (from 2 to 20) (Figure 1c). hole. In the chamber, measurements conducted with
In general, fault surfaces display higher average JRC fiber-optic sensors allow simultaneous monitoring of
values than bedding planes (respectively 16 and 8). changes (high frequency [120 Hz] and high accuracy)
Nevertheless, a large JRC variation occurs on the same in fluid pressure (± 1 kPa) and displacement normal
fault plane parallel to the direction of the tectonic to the walls of the tested joint (±10−7 m) (Cappa et al.,

712
Figure 2. (a) HPPP probe (1: upper and lower packers, 2: extending anchors, 3: displacement sensor, 4: valve, 5: pressure
sensors); (b) typical pressure-deformation loop-shaped HPPP curve; (c) in situ installation of HPPP protocol; (d) pressure-
deformation-tilt responses to a two-pressure-steps injection experiment.

2006). The displacement sensor is fixed to the bore- chamber was extensively analyzed through numeri-
hole walls by two extending anchors located on both cal modeling (Cappa et al., 2006). In figure 2b-c,
sides of the tested discontinuity (Figure 2a). we show an example of a test conducted at the
To conduct the HPPP tests, we increased the fluid Coaraze slope on a 80◦ dipping fault. The resulting
pressure in the injection chamber to the required normal-displacement-versus-pressure curve shows a
pressure-pulse magnitude using a volumetric water characteristic loop behavior, in which the paths for
pump, allowing the pressure to increase from 10 to pressure increase and decrease are different. The ini-
120 kPa (chosen to be lower than the ambient state of tial rising portion of the curve is highly dependent on
stress on the joint to prevent hydraulic fracturing). The the tested fracture’s permeability, normal stiffness, and
injection chamber is connected to a valve leading to intact rock stiffness close to the fracture walls. For
an upstream volume used to perform a pressure pulse. example, a lowering by a factor of 10 in permeabil-
The fluid pressure is first increased upstream of the ity or an increase by a factor of 2 in normal stiffness
valve. Thereafter, a pulse is initiated when the valve is respectively induce a 50% and 25% lowering of the
opened to allow water to enter the injection chamber. curve slope (Figure 2b). The falling portion of the
The pressure first increases in the packed-off section, curve is influenced by mechanical processes within
and then it begins to decrease as a result of fluid flow a larger portion of the surrounding fractures rock. In
into the fractures and rock. this example, shear along bedding planes and, as a
The transient poroelastic response of the fractures result, bedding-plane shear stiffness strongly affect
when the pressure pulse is applied in the injection the results, an increase by a factor of 10 inducing a

713
20 to 45% variation in this part of the curve (Figure 2b).
Then, if the fracture network geometry and the state of
stress are known close to the test, it is possible to esti-
mate the hydraulic permeability (kh ) and the normal
stiffness (kn ) of the tested fracture, as well as those of
adjacent bedding planes.
The application of HPPP to rock slope stability
characterization includes two successive steps:

1. The HPPP tests are conducted to estimate the local

poroelastic properties of joints. This is a short-
duration test (about 15 seconds long) meant to
induce a local deformation of the tested joint and
to estimate its local properties. In the example,
nine tests were conducted in small diameters (Ø =
0.07 m) boreholes, as well as in five faults and four
bedding-planes.
2. An injection test is conducted in one borehole using Figure 3. Numerical model of the Coaraze slope (points
1 to 9 are HPPP tests points).
the HPPP probe while simultaneously monitoring
the slope surface tilt (Figure 2c). This is a long-
duration test (about 2000 seconds long) meant to
induce an extended deformation in the slope and
to test slope behavior. In the example, the injection
was performed at 5 m depth in the slope, in two 1 to 9 and tilt responses at the slope surface matched
increasing pressure steps to produce a clear signal the experimental responses.
both at the injection and at the near-field tiltmeter Close agreement (5 to 25% discrepancy depend-
points (Figure 2c). ing on the points) for both slope surface and internal
measurements was obtained when a large variability in
slope-element properties was introduced into the mod-
els. Specifically, we had to define a strong contrast
3 RESULTS between faults and bedding-plane normal stiffnesses
(knfaults = 10−3 × knbedding−planes ) and hydraulic aper-
Field data were analyzed with different coupled tures (bhfaults = 10× bhbedding−planes ), There was a
hydromechanical (HM) codes (ROCMAS, FLAC3D, highly nonlinear correlation between normal stiffness
and UDEC). Considering the discontinuities and the and the hydraulic aperture bh of fractures, which was
rock matrix to be linearly elastic, we developed a 2D fitted to a hyperbolic function (Figure 4a). This cor-
model that represented the topography and joint net- relation was the result of both the significant property
work geometry of the EW cross section (Figure 3). contrast between the faults and bedding planes, and
A 25 m × 21 m model was chosen to minimize bound- the high variability of properties within a given frac-
ary effects. Stress concentration at the valley base, ture type. Since there are several parameters that affect
calculated from large-scale models, was applied to the the hydraulic aperture of a given fracture (mechanical
bottom 2 meters of the topographical left boundary. aperture, roughness, and tortuosity), it is hard to for-
The remainder of the topographical surface was free mulate a simple explanation for the high bh of such
to move. Fixed X and fixed Y displacement condi- a fracture. In the very special case of shallow dis-
tions were imposed on the right and basal boundaries, continuities in slopes, it was commonly observed that
respectively. All model hydraulic boundaries were discontinuities almost parallel to the slope direction
impervious. HPPP tests were analyzed one by one. The and dip (roughly the case of vertical faults at Coaraze)
model was loaded by imposing the time-dependent were in general widely opened. We can assume that
pressure pulse measured in the injection chamber. their tortuosity was low and their roughness small
Hydraulic and mechanical properties were estimated compared to their mechanical aperture. Elsewhere,
after matching calculated and measured pressure and a few centimetric contacts were observed between
normal displacement curves at the point. Second, the the two walls of those discontinuities, as a result of
injection test was analyzed. The model was loaded by the progressive failure of those contacts due to slope
imposing the time-dependent pressure curve measured decompression. The small number of contacts, and
at Injection point 4 (Figure 3). Properties from all the hence relatively small contact area could explain the
local tests were introduced as input values within the relatively low stiffness of discontinuities, which would
model and adjusted until both local responses at points thus have a large hydraulic aperture.

714
Figure 4. (a) correlation between normal stiffness and hydraulic aperture of discontinuities; (b) localization of damage in
the slope; (c) proposed joint model.

Such a mechanism could explain the property lower bh . Nevertheless, there is a high variability in
variation from Point 3 to 1 on Fault F1, a factor-of- bedding-plane properties corresponding to bedding-
10 lowering of the normal stiffness and a factor-of-2 plane damage characterized by a tenfold increase
increase of bh , respectively. This mechanism could of the bh value and a threefold decrease of normal
also explain the property variation between Points 4, stiffness, from points 8 and 9, located at 2 m and 5 m
5 to 2, a factor-of −2.4 lowering of normal stiffness depths, respectively, to points 6 and 7, located at 1 and
and a factor-of-10 increase in bh (Figure 4a). Points 0.5 m depths, respectively (Figure 4a and b).
1 and 2 are located on shallow segments of the faults The shear stiffness of bedding planes and faults was
and are more disturbed than points 3, 4, and 5, which estimated to be 0.1 to 0.02 times lower than normal
are located deeper (Figure 4b). On the other hand, bed- stiffnesses.
ding planes slightly affected by decompression effects
and submitted to higher compressive stresses (mainly
because of their inward dipping orientation) are more 4 DISCUSSION AND CONCLUSION
closed than faults. In this case, with more contacts
between the two walls of the discontinuity, bedding The HPPP method, based on combining surface tilt
plane would display higher normal stiffnesses and with localized internal coupled pressure/deformation

715
measurements, can help to characterize the influence direct estimation of Ks through in situ measurement of
of discontinuities properties on slope destabilization. both normal and tangential displacements of fracture
For a given slope geometry, the HPPP test enables an walls.
in situ estimation of the hydraulic and mechanical
properties of discontinuities at the single fracture
scale, as well as the coupling relationship between
those key properties. First, an empirical hyperbolic REFERENCES
relationship is proposed between normal stiffness and
the hydraulic aperture of joints. Second, a rough esti- Amadei, B., Wobowo, J., Sture, S. & Price, R.H. 1998.
mation of fracture normal/shear stiffness ratio of 1 to Applicability of existing models to predict the behav-
5 is given. This agrees well with laboratory studies in ior of replicas of natural fractures of welded tuff under
which such a relation is obtained from fracture shear different boundary conditions. Geotech. Geo. Eng. 16,
tests performed under a state of stresses sufficiently pp. 79–128.
low to favor dilation when sliding along discontinuity Aydin, A. & Basu, A. 2005. The Schmidt hammer in rock
planes occurs. material characterization. Engineering Geology 81, pp.
The established permeability-stiffness relationship 1–14.
Bandis, S.C., Lumsden, A.C. & Barton, N.R. 1983.
reduces the number of unknown properties to be intro- Fundamentals of Rock Joints Deformation. Int. J; Rock
duced in stability numerical models. Thus, those Mech. Min. Sci. & Geomech. Abstr. 20, No. 6, pp.
two properties are no longer unknown, but can be 249–268.
used as input data to calculate shear processes hap- Barton, N. & Choubey, V. 1977. Shear strength of rock joints
pening in natural slopes. Further, the compressive in theory and practice. Int. J. Rock Mech. Sci. & Geomech.
and shear strengths of discontinuities can be car- Abstr. 10, pp. 1–54.
ried out as follows. First, the initial normal stiffness Barton, N. & Bandis, S. 1990. Review of predictive capa-
Kni and initial hydraulic aperture bhi can be esti- bilities of JRC-JCS model in engineering practice. Rock
mated using the hyperbolic normal stiffness-hydraulic Joints. Barton & Stephansson (eds). Balkema, Rotterdam.
ISBN 90 6191 109 5. pp. 603–610.
aperture relashionship. If we refer to laboratory exper- Cappa F., Guglielmi Y., Rutqvist J., Tsang C.F. & Thoraval A.
iments described in the literature (Bandis et al., 1983), 2006. Hydromechanical modeling of pulse tests that mea-
a good fitting of Kni can be obtained, mainly with the sure both fluid pressure and fracture-normal displacement
uniaxial compressive strength σc and bh, and secondly at Coaraze Laboratory Site, France. Int. J. Rock. Mech.
with joint surface roughness. This can be illustrated Min. Sci. 43, pp. 1062–1082.
from equation (1): Erisman T. & Abele G. 2000. Dynamics of rockslides and
rockfalls. Springer Verlag, Berlin, 316 p.
Gentier, S.S., Lamontagne, E., Archambault, G. & Riss, J.
Kni = σc /bh (1) 1997. Anisotropy of flow in fracture undergoing shear
and its relashionship to the direction of shearing and injec-
Second, an estimation of shear strength can be tion pressure. Int. J. Rock Mech. Min. Sci. Geomech.
made by considering that the fracture dilation peak Abstr. 34.
is reached at the maximum inflection point of the Gentier, S.S., Riss, J., Archambault, G., Flamand, R. &
normal stiffness-hydraulic aperture curve (Figure 4c). Hopkins, D.L. 2000. Influence of fracture geometry on
sheared behaviour. Int. J. Rock Mech. Min. Sci. 37, pp.
The shear stiffness estimated from the measured 161–174.
normal/shear stiffness ratio is then related to the shear ISRM 1978. Suggested methods for the quantitative descrip-
strength in Equation (2): tion of discontinuities in rock masses. Int. J. Rock Mech.
Min. Sci. & Geomech. Abstr. 15, pp. 319–368.
Ks = σn tan φ (2) Landanyi, B. & Archambault, G. 1970. Simulation of shear
behaviour of a jointed rock mass. Pro. Of the 11th Symp.
On Rock Mech. (AIME), pp. 105–125.
where σn = normal stress and φ = friction angle. Laubscher, D.H. 1990. A geomechanics classification
In our example, the estimated values for JCS and system for rating of rock mass in mine design. J. South
friction angle are 108 Pa and 90◦ (for an average σn = African Inst. Of Mining and Metallurgy 90, No. 10, pp.
4 104 Pa), respectively. This is in good accordance 257–273.
with values extrapolated with the Barton-Bandis Li B., Jiang, Y., Koyama, T., Jing, L. & Tanabashi, Y. 2007.
method: JCS = 108 Pa, JRC = 19, and residual fri- Experimental study of the hydro-mechanical behavior of
ction angle = 40◦ . The HPPP approach is then par- rock joints using a parallel-plate model containing contact
areas and artificial fractures. Int. J. of Rock Mech. & Min.
ticularly effective for it relies on a reduced number Sci., in press.
of parameters (Kn , Ks and bh ) that can be measured Olsson, W.A. & Brown, S.R. 1993. Hydromechanical
in-situ with the HPPP probe, making the estimation response of a fracture undergoing compression and shear.
of in situ fracture-strength parameters less subjective. Int. J. Rock Mech. Min. Sci. Geomech. Abstr. 30, pp.
Moreover, further development of the probe will allow 845–51.

716
Patton, F.D. 1966b. Multiple Modes of Shear of Failure in Rutqvist J. & Stephansson O. 2003. The role of hydromechan-
Rock. Proc. 1st Cong. Int. Soc. Rock Mech., Lisbon, pp. ical coupling in fractured rock engineering. Hydrogeology
509–513. Journal 11, pp. 7–40.
Plesha, M.E. 1987. Constitutive models for rock discontinu- Saeb, S. & Amadei, B. 1992. Modelling rock joints under
ities with dilatancy and surface degradation. Int. J. Numer. shear and normal loading. Int. J. Rock Mech. Min. Sci.
Anal. Meth. Geomech. 11, pp. 345–362. Geomech. Abstr. 29, pp. 267–278.

717
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Fuzzy prediction and analysis of landslides

Yong He, Bo Liu, Wen-juan Liu, Fu-qiang Liu & Yong-jian Luan
Qingdao Agricultural University, Shandong Qingdao, China

ABSTRACT: There are internal and external reasons for the occurrence of landslip. Through the analysis on a
large number of investigation materials of landslip, the growth inducements of landslip are found out. By using
theories in blur maths, we conduct order arrangement of these inducements in accordance their significance, put
forward blur judgment rule of landslip estimation and present applicable examples.

1 INTRODUCTION model, the probability of the landslip’s occurrence

is determined. In the following part, the writer will
Landslip, a serious geological disaster, with seri- make a brief introduction of fuzzy forecast problems
ous harmfulness, greatly threatens human beings’ of landslip raised by the writer and the scientific team.
existence. The inducements of landslip which are com-
plex, include artificial elements and natural elements.
A great deal of facts on landslip shows that landslip 2 BASIC PRINCIPLE OF FUZZY FORECAST
is often the result of mutual effect of many elements. OF LANDSLIP
There is a high degree of obscurity of the relationship
between different elements and the action mechanism, 2.1 Establishment of landslip forecast index system
so it is hard to present a quantitative mathematic model
with common methods. It is one of blur maths’ (Kamal At present, the chain of causation model is commonly
et al. 2001, Duzgoren et al. 2002, Furuya et al. 1999, used to analyze the reasons of various kinds of acci-
Lee et al. 2002, Mauritsch et al. 2000, Raetzo et al. dents. The writer and the scientific research team,
2002) strong points to research and deal with the non- making abundant materials on landslip and the actual
clearness of differences of objective things in their in situ motoring materials as the basis, has formed the
transitions. If blur theories can be applied to settle landslip forecast index element system which can be
the forecast problem of landslip possibility, it will seen in Table 1.
be helpful for the forecast of landslip accidents and
reducing the harm of landslip disasters to the lowest.
2.2 The determination of index weight
The writer, together with the leader of the scientific
research team, has studied various geological disasters The determination of index weight is in accordance
(including landslip, ground sedimentation and so on) with the AHP method (Analytic Hierarchy Pro-
for many years, and in the research process, we have cess Method) put forward by American operational
accumulated a mass of materials on landslip and the research master A.L. Saaty in the 1970s.
in situ motoring data of landslip. On the base of com-
prehensive analysis and research, the writer and the 2.2.1 The determination of index, element
scientific research team has summed up many asso- and sub-element weight
ciated elements (Duzgoren et al. 2002, Furuya et al. The lowest layer of the writer’s analysis of hierarchy
1999), using theories in blur maths and making basic process on landslip forecast is the element collec-
principles of fuzzy evaluation method as the guidance, tion of subsidiary appraisal index, the middle layer
we scientifically set down the subjection degree of is the element collection of leading appraisal index
each variable in the fuzzy set of landslip and confirm and the highest layer, also called the target layer, is
the relative weight of appraisal index elements by use the stabilization condition of the landslip body. There-
of analysis of hierarchy process. Through the analysis fore, the hierarchical structure model is established,
of every associated index of landslip, we find out their which can be seen in Table 1. Then, establish the
subjection degree in the fuzzy set. By use of estimate pairwise comparison matrix (that is judgment matrix)

719
Table 1. Appraisal index system of landslip forecast.

The middle
layer The bottom layer Appraisal matrix (poll result of the appraisal team)

Extremely unstable
Basically stable

Critical stage
Target Layer

Sub-stable

Unstable
Weight

Stable
Main appraisal Subsidary appraisal V1= V2= V3= V4= V5= V6=
elements and code Weight elements and code 0.000 0.20 0.40 0.60 0.80 1.00

Composing of landslip
body materials 0.44 0 1 6 3 1 0
A1
Internal elements 0.41 Thickness of landslip
A body 0.19 1 2 6 2 0 0
A2
Form of the leading
slipping surface 0.37 0 1 4 4 1 1
A3
Coverage condition of the
surface of landslip body
B1 0.06 0 2 4 4 1 0
Free surface condition of
Stabilization condition of landslip body

landslip body 0.20 0 1 5 4 1 0

B2
Change condition of
water table 0.08 0 1 4 4 1 1
B3
Erosion condition of
surface water 0.13 0 1 3 4 2 1
B4
External 0.59 Drainage condition of the
environmental surface of landslip body
elements B5 0.07 1 2 3 4 1 0
B
Fragmentation degree of
the surface of landslip
body 0.05 1 1 5 2 2 0
B6
Loading condition of the
surface of landslip body
B7 0.06 1 1 5 3 1 0
Precipitation condition 0.20 0 1 3 4 2 1
B8
Vibration-bearing
condition of the surface
of landslip body 0.15 1 7 3 0 0 0
B9

720
according to Saaty’s nine-rank rating scale method. Table 2. Average random coincidence index.
See Table 1 to acquire the weight of index, element
and sub-element. n 2 3 4 5 6 7

RI 0.00 0.58 0.90 1.12 1.24 1.32

2.2.2 Calculation of each level’s index weight n 8 9 10 11 12
First of all, the level simple sequence is conducted. RI 1.41 1.45 1.49 1.52 1.54
According to judgment matrix, the weight of the sim-
ple sequence is calculated. That is to say, to the
elements of the last level, calculate this level’s weight
of all the elements which has association with them.
According to the principle of analysis of hierarchy pro- When CR ≤ 0.10, the judgment matrix can have sat-
cess, through mathematic calculation, the eigenvector isfying coincidence, otherwise, it should be adjusted.
which is the result of judgment matrix’s corresponding According to the above calculation methods, carry
with the maximum eigenvalue λmax , is gained and each on the calculation of the coincidence index of judg-
element of eigenvector (that is weight value) is the sim- ment matrix in different levels of target—index,
ple sequence result. By use of MATLAB software, the index—element and element—sub-element. See the
writer calculates the eigenvector and gains the calcula- calculation result in Table 1. None of the CR is
tion result of level simple sequence of corresponding more than 0.10 and all the judgment matrix meet the
elements which can be seen in Table 1. coincidence requirement.
Then, the coincidence test of single taxis is car- Next, the level total sequence is conducted. Taking
ried on. The main advantage of AHP method is that it advantage of all the results of the level simple sequence
can quantify decision-maker’s fixed thoughts. Due to in the same level and all the weight of elements in the
the complexity of the appraised objects and experts’ last level, calculate all the elements’ weight value in
diversity and unilateralism on cognition, the nine-rank this level aiming at the total target.
rating scale can’t guarantee the complete coincidence Finally, the coincidence test of the level total
of each judgment matrix. Therefore, the coincidence sequence is conducted. Like the level simple sequence,
test must be used to check if there are contradictions the level total sequence also needs the coincidence test.
between each index’s weight and the check step is as The test result all accord with CR ≤ 0.10 and also
follows: meets the coincidence requirement.
1. Calculate the maximum latent root of judgment
matrix λmax 2.3 Determination of the appraisal element base
 n  The writer chooses six grades to form the appraisal

λmax = [(AW )i /Wi ] /n (1) collection V of landslip body’s stabilization condition.
i=1 V = {V1 (stable), V2 (basically stable), V3
(sub-stable), V4 (critical state), V5 (unstable), V6
In formula (1), λmax is the maximum latent root; (extremely unstable)}.
n is the linage of the judgment matrix which is also
the number of index in the arrangement subsystem;
A is the judgment matrix; W is the eigenvector of 2.4 Establishment of appraisal matrix
the judgment matrix; and (AW)i is the number i Many experts on landslip and local residents who have
element of vector AW gained by the multiplication experienced the landslip disasters form the appraisal
of the judgment matrix A and eigenvector W. team. After the comprehensive investigation (includ-
2. Calculate the coincidence index CI ing drilling materials, experiment materials and motor-
ing materials) towards the landslip bodies that are to
CI = (λmax − n)/(n − 1) (2) be appraised and analysis, poll on conditions of every
second-grade index elements which are divided into
3. Calculate the random coincidence ratio CR six appraisal collections are hierarchically conducted.
See Table 1 to find out the appraisal result. Form the
CR = CI /RI (3) judgment matrix according to appraisal results. Every
element in the judgment matrix is the result gained by
RI is the average random coincidence index. See the ballot possessed by this element to divide the total
Table 2. ballot.

721
2.5 Establishment of landslip forecast model 3.1 Relative calculation of main appraisal
element A
The landslip forecast model established by the writer
and the scientific team is as follows: The area of A which is composed by three sub-
sidiary appraisal elements is: A = {A1 (Composing
of landslip body materials), A2 (Thickness of landslip

n
body), A3 (Form of the leading slipping surface )}. In
P= [aj bj ] (4)
appraisal systems, the area which is composed by 6
j=1
appraisal classifications is V: V = {V1 (stable), V2
(basically stable), V3 (sub-stable), V4 (critical state),
In formula (4) V5 (unstable), V6 (extremely unstable)}. According to
the subjection degree which is gained by 11 experts’
votes, (for example, in Table 1, the appraisal of sub-

n sidiary appraisal element A1 (Composing of landslip
aj = xj / xj (5) body materials) is that the vote for stable is 0, the pro-
j=1 portion is 0/11 and the subjection degree is 0.00; the
vote for basically stable is 1, the proportion is 1/11
and the subjection degree is 0.09; the votes for sub-
In formula (4), when element j is not contained in stable are 6, the proportion is 6/11 and the subjection
the objective elements, bj = 0; when element j is degree is 0.55; the votes for critical state are 3, the
contained in the objective elements, bj = λj . In for- proportion is 3/11 and the subjection degree is 0.27;
mula (4) and (1), aj is the probability coefficient of votes for unstable are 1, the proportion is 1/11 and the
the number j element; n is the number of the main subjection degree is 0.09; votes for extremely unstable
determined influential elements; xj is the number of are 0, the proportion is 0/11 and the subjection degree
accidents that has relationship with element j; λj is is 0.00) we establish the appraisal matrix RA which is
the degree of subjection of element j determined by composed by three subsidiary appraisal elements as:
experts in accordance with fuzzy appraisal method.
From formula (4), we can find out that the more the ⎡ ⎤
0.00 0.09 0.55 0.27 0.09 0.00
accidents caused by certain accident element j, the
RA = ⎣0.09 0.18 0.55 0.18 0.00 0.00 ⎦
more the probability coefficient aj and the element j
0.00 0.09 0.36 0.36 0.09 0.10
accounts for a more weight comparing with other ele-
ments. At this time, if the element j is contained in the
objective conditions of forecast objects, the probabil- Three subsidiary appraisal elements weight fuzzy
ity of the landslip occurrence P is higher, which can vectors AA which are determined by integrat-
be seen from the model (formula (1)). The more the ing with existing landslip data and utilizing the
objective elements that the landslip causes accidents, analysis of hierarchy process is: AA = (0.44, 0.19,
the more the items which is not zero and the higher 0.37). Integrated appraisal BA of main appraisal
the summation value. Therefore, the possibility of its element A is: BA = AA RA = (0.0171, 0.1071, 0.4797,
occurrence is higher. 0.2862, 0.0729, 0.0370). The subjection degree of
main appraisal element A is calculated by λA . Endow-
ing value K = (0.00, 0.20, 0.40, 0.60, 0.80, 1.00) res-
pectively into every appraisal value of main appraisal
3 APPLICATION EXAMPLES OF LANDSLIP area A, the subjection degree λA of main appraisal
MODEL element A, λA = KBAT = 0.4803.

Table 1 is the fuzzy appraisal that the writer and

3.2 Relative calculation of main appraisal
scientific team has made towards the landslip body
element B
in Cuobu Mountain. The appraisal team consists of
11 experts. After thorough investigations and analy- The area B which is composed by nine subsidiary
sis of landslip scene of Cuobu Mountain and whole appraisal elements is B = {B1 (Coverage condition
experiments and data of supervision and measurement, of the surface of landslip body), B2 (Free surface
these 11 experts make a classification vote for each condition of landslip body), B3 (Change condition
subsidiary appraisal index in accordance with 6 eval- of water table), B4 (Erosion condition of surface
uation collections. After the analysis of fuzzy forecast water), B5 (Drainage condition of the surface of land-
model, they make a prewarning of landslip disaster slip body), B6 (Fragmentation degree of the surface
of Cuobu Mountain and the landslip occurs sooner. of landslip body), B7 (Loading condition of the sur-
According to Table 1, the calculating process of fuzzy face of landslip body), B8 (Precipitation condition),
appraisal of landslip body in Cuobu Mountain is. B9 (Vibration-bearing condition of the surface of

722
landslip body)}. The appraisal area V which is com- 4 CONCLUSIONS
posed by six appraisal classifications in appraisal
collection is: V = {V1 (stable), V2 (basically sta- The key of landslip forecast fuzzy appraisal is the
ble), V3 (sub-stable), V4 (critical state), V5 (unstable), determination of subjection degree which depends on
V6 (extremely unstable)}. According to the subjection experts’ experiences in most degree, so the experts
degree which is gained by 11 experts’ votes, we estab- who are chosen must have very abundant spot experi-
lish the appraisal matrix RB which is composed by nine ences and know the occurring rules of landslip disaster
subsidiary appraisal elements as: very well. To settle the problems of landslip forecast
by fuzzy mathematical methods is to number landslips
⎡ ⎤ which are a fuzzy system by mathematical methods.
0.00 0.18 0.36 0.36 0.10 0.00
Landslip forecast model which is established by the
⎢0.00 0.09 0.45 0.36 0.10 0.00⎥
⎢0.00 0.10⎥
author and scientific research team with fuzzy mathe-
⎢ 0.09 0.36 0.36 0.09 ⎥
⎢0.00 0.09 0.27 0.36 0.18 0.10⎥
matical theory is a good method to solve the problem
⎢ ⎥ of landslip forecast, with the characteristics of simple
RB = ⎢
⎢0.09 0.18 0.27 0.36 0.10 0.00⎥
⎥
⎢0.09 model structure, strong pragmatic, fast calculation and
⎢ 0.09 0.46 0.18 0.18 0.00⎥
⎥
⎢0.09 good operation. Due to the writer’s level, landslip fore-
0.09 0.46 0.27 0.09 0.00⎥
⎣ ⎦ cast model which is established by the author and sci-
0.00 0.09 0.27 0.36 0.18 0.10
entific research team with fuzzy mathematical theory
0.09 0.64 0.27 0.00 0.00 0.00
and the design of its appraisal index must have many
blemishes, the choosing of its elements collection is
Nine subsidiary appraisal elements weight fuzzy not very proper and reasonable and some important
vectors AB which are determined by integrating with landslip elements maybe be left out, so we hope that
existing landslip data and utilizing the analysis of you can complement and perfect it in applications.
hierarchy process is: AB = (0.06, 0.20, 0.08, 0.13,
0.07, 0.05, 0.06, 0.20, 0.15). Integrated appraisal
BB of main appraisal element B is: BB = AB RB = ACKNOWLEDGMENTS
(0.0297, 0.1842, 0.3395, 0.2916, 0.1140, 0.0410).
The subjection degree of main appraisal element B The research project discussed in this paper has been
is calculated by λB . Endowing value K = (0.00, funded by the National Natural Science Fund of
0.20, 0.40, 0.60, 0.80, 1.00) respectively into every China (No.83200655) and Construction Science Fund
appraisal value of main appraisal area B, the subjec- of Shandong Province (No.LJK-20020312). Sincere
tion degree λB of main appraisal element B, λB = gratitude is presented to them.
KBBT = 0.4798.
REFERENCES
3.3 Forecast of occurrence probability of landslip Kamal M., Al-Subhi & Al-Harbi. 2001. Application of the
According to a great number of landslip data statisti- AHP in project management. International journal of
cal results, we chose probability coefficient value a1 project management, 19:19–27.
Duzgoren Aydin N.S., Aydin A. & Malpas J. 2002. Reassess-
(that is aA ) = 0.41; a2 (that is aB ) = 0.59. Accord-
ment of chemical weathering indices: case study on
ing to Landslip Forecast Model (formula (4)) we can pyroclastic rocks of Hong Kong. Engineering Geology,
get the probability P of landslip occurrence of Cuobu 63(1/2):99–119.
Mountain as Furuya G., Sassa K., Hiura H., et al. 1999. Mecha-
nism of creep movement caused by landslide activ-

n ity and underground erosion in crystalline schist,
P= [aj bj ] = aA λA + aB λB = 0.481276 Shikoku Island, southwestern Japan. Engineering Geol-
j=1
ogy, 53(3/4):311–325.
Lee H.S. & Cho T.F. 2002. Hydraulic Characteristics of
Rough Fractures in Linear Flow under Normal and
Generally when P is more than 0.25, small omen of Shear Load. Rock Mechanics and Rock Engineering,
landslip will appear, if more than 0.50, the probability 35(4):299–318.
of landslip occurrence will be much higher. Due to the Mauritsch H.J., Seiberl W., Arndt R., et al. 2000. Geo-
probability P of landslip occurrence of landslip body physical investigations of large landslides in the Carnic
Region of southern Austria. Engineering Geology, 56
in Cuobu Mountain has reached 0.481276 which is (3–4):373–388.
near to dangerous value of 0.50, so scientific research Raetzo H., Lateltin O., Bollinger D. et al. 2002. Hazard
team makes a prewarning that landslip body in Cuobu assessment in Switzerland-codes of practice for mass
Mountain will slide. It is proved to be right by later movements. Bulletin of Engineering Geology and the
facts. Environment, 61:263–268.

723
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

LPC methodology as a tool to create real time cartography

of the gravitational hazard: Application in the municipality of Menton
(Maritimes Alps, France)

M. Hernandez, T. Lebourg & E. Tric

University of Nice—Sophia Antipolis, Géosciences Azur, UMR 6526,
Valbonne, France

M. Hernandez & V. Risser

Cabinet Risser, Saint Laurent du Var, France

ABSTRACT: This paper presents the LPC (Landslide Predictive Cartography) methodology for evaluating
slope failure in wide study area. This deterministic methodology is based on the limit equilibrium theory and
combined with a dynamic hydrogeological model. It assesses, for each time laps, the slope stability in relation
with real or modelled climatic events. This model was described and applied on the Menton municipality
(Maritimes Alps, France). This studied area is densely populated in the thalweg zone and have very steep slope.
The Menton landscape morphology comes from the erosion of the sedimentary formation. The results obtained
indicate that, for a 10 m DEM scale resolution and a 20-year return rainfall modelling, 21.4% of the studied area
should be unstable and 70.7% of the landslide occurred have a FS inferior to 1.

1 INTRODUCTION The equation (1) of the Factor of Safety (FS) is com-

monly used for the study of the equilibrium conditions
Emergency management planning requires prediction of the soil mass. This equation is based on Mohr—
of the damage associated to the landslide occur-
rence. In this paper a new deterministic model, using
dynamic hydrogeological models, is presented: the
LPC (Landslide Predictive Cartography) methodol-
ogy. In order to assess the efficiency of this methodol-
ogy, it is applied on the Menton area (Maritimes Alps,
France) where a landslide mapping is well defined.
The origin of this methodology comes from the study
of the Rucu Pichincha volcano near the town of Quito
in Ecuador (Risser, 2000). Throughout an operational
model, this methodology assesses, for each time laps,
the soil stability in relation with climatic events.

2 LPC: A SHALLOW LANDSLIDE SLOPE

STABILITY MODEL

2.1 Infinite slope stability model

Data is organized by grids depending on the reso-
lution of the Digital Elevation Model (DEM). The
LPC methodology uses the limit equilibrium the- Figure 1. Infinite slope model, where the resisting force
ory to analyse the stability state (Spencer, 1967) for holding the block in place is given by W (which the mass *
each pixel (figure 1). The slope stability is based the gravita-tional force) multiplied by the cosines of the slope
on infinite slope model including several simplifying angle. The driving force is Wsinβ (inspired by Selby et al.
assumptions (Selby et al. 1993). 1993).

725
Coulomb failure criterion: τmax = C + σ  · tanϕ  , intensity-duration (I-D) rainfall. The infiltration pro-
where τmax is the maximum shearing stress and σ the cess is founded on the empirical Horton equations
normal stress (Costet and Sanglerat, 1981). (Horton, 1933). After the infiltration step, the one-
dimensional vertical flow in the unsatured zone is
C based on the simplified equations used in Fuentes et al.
FS = (1992).
ρs.g. sin β. cos β Once the accumulation of the water begins at the
(ρs.h − ρw.hw(t)) × tan ϕ  base of the soil column, the accumulated zone is con-
+ (1) sidered like saturated. The saturated flow is based on
ρs.h. tan β
the Darcy’s equation (Darcy, 1856) and more precisely
on the Bernoulli’s theorem.
where ϕ’ is the effective angle of internal friction The hydrogeological model depends on the quality
(deg), C  is the effective cohesion (KPa), h the ver- of the landuse mapping for the infiltration rate and on
tical soil depth (m) which is calculated with electrical the permeability mapping for the water flow in the soil.
resistivity measurements, hw (t) the vertical dynamic In order to have the most critical situation of
saturated soil depth (m) depending on the hydroge- the slope stability we kept the hydrogeological map
ological model, ρs is the dry soil density (kg.m−3 ), with the maximum soil saturation coming from a
ρw is the density of water (kg.m−3 ), β is the slope 20 year-return rainfall modelling with one hour
angle (deg), and g is the gravitational acceleration duration (Desbordes, 1987).
(m.s−2 ).
A sensitivity study of this equation was realised.
The results are similar to those obtained by Borga
et al. (2002). The following parameters are quoted
in the decreasing order of influence on the safety fac-
tor equation: slope angle, effective angle of internal
friction, soil depth and groundwater-soil ratio.
In this study the pixel resolution of the DEM is
10 m × 10 m (extrapolation of the DEM 25 × 25 m).
The other maps are also organized by grid scale of
10 m. The whole maps are composed by 672 × 530
pixels i.e. 6,720 m × 5,300 m area. Only the Menton
municipality zone is well mapped (geological map),
hence the surface of the effective stability study zone
is 35.6 km2 .
Figure 2. Hydrogeological model.

2.2 Hydrogeological model

In order to evaluate the parameter hw (t), a hydrogeo-
logical model plays an important role in the FS calcu-
lation. The groundwater response following a rainfall
event can be modelled by many hydrologic mod-
els like TOPOG (O’Loughlin, 1986), TOPMODEL
(Beven, 1997), and DYNWET (Wilson and Gallant,
2000).
The LPC methodology uses its own hydrogeolo-
gical dynamic model. This model assumes that the
shallow subsurface flow has the same behaviour than
a perched water table. In case there is a homogenous
soil thickness the shallow subsurface flow downs-
lope follows the topographic gradient. But, contrary
to the other methods, in case there is a variation in
the soil thickness, the subsurface flow follows the
substratum gradient. The flow direction is calculated
with the D∞ algorithm (Taborton, 1997). The rainfall Figure 3. Example application, Menton, (Maritim Alps,
event could be whether modelled with the Desbor- France) for a 20-year return rain, 60 min after the begin-
des (1987) methodology whether came from a rain ning of the rainfall. The white zone at the South-East is the
gauge measured data. These two methods provide Mediterranean Sea. The white zone is Italy.

726
3 STUDY AREA AND LANDSLIDES DATA

3.1 Localisation of the study area

The Menton municipality is localised in the ‘‘Alps
Maritimes’’ (SE, France) touching the Italian border
and the Mediterranean Sea. The geomorphology of
the studied area is compound by very steep slope.
This zone is densely populated in the thalweg zone;
the landslide factor is one of the most important
risks for this municipality. The landslide data and the
geological map are provided by the ‘‘C.E.T.E Méditer-
ranée’’ (In French: ‘‘Centre des études techniques de
l’équipement’’, in English: Facility Technical Study
Center).

3.2 Geological setting

The field study is localised in the Alps external
sedimentary cover units. The figure 4 displays the
localisation of the different lithology. The great part of
the map shows the importance of the colluviums and
eluviums deposit. Theses colluviums and the eluviums
are classified according to their geological origin and
their thickness. Three levels of thickness correspond- Figure 5. Landslide localisation on the Menton landscape.
The black polygons represent the landslides occurred during
ing to the vertical soil depth (h) are used: low (0.5 to
the period 1999–2000.
1 m), medium (1 to 3 m), and high (more than 3 m).
The South-East part of the study area is constituted
by Jurassic limestone and dolomite cliffs. In the north
part, large outcrops of Cretaceous marl are present. show on the surface. The Quaternary formation, con-
In the central part, Oligocene sandstones and flyschs stituted of alluvium and colluvium, covers the thalweg
and the coast zone.

3.3 Landslides localisation

On the field 191, landslides were counted and mapped.
In order to compare the FS map obtained with the
LPC methodology and the landslides map, the land-
slide zones are redrawn keeping only the failure zone
and deleting the accumulation zones. The landslides
surface corresponds to 1872 pixels i.e. 0.18 km2 or
1.34% of the total studied surface. Some of these
events caused several deaths. The origin of these land-
slides can be a natural slope destabilisation or an
anthropogenic perturbation (bad management of the
embankments). They are mainly localised in the cen-
tral part of the map, i.e. in the Oligocene formations
constituted by flyschs, sandstones, and their colluvium
and eluvium.
The photo 1 is an example of the landslide occur-
rence in the Menton municipality. This destabilisation
occurred after a heavy rain estimated as being a 20-year
return rainfall. In this example, the steep slope and the
anthropogenic action (deforestation and habitations)
Figure 4. Geological map of the Menton municipality. had support the landslide occurrence.

727
 Figure 6. FS map of the Menton municipality.

Photo 1. Landslide occurred in 2000.

Figure 7. Factor of Safety distribution in the studied area.

The line graph represents the cumulated percentage and the
4 RESULTS bar chart represents the percentage for each class.

Database and field tests were used to obtain the

mechanical and hydrogeological parameters. Depend-
ing on theses, a ‘‘Factor of Safety’’ map was realised
(Figure 6).
According to the FS map, 21.4% of the studied
area should be unstable for a 20-year return rainfall
(Figure 7).
The realisation of the DEM (1995) is posterior to
the landslides database (1999–2000), consequently the
DEM shows the altimetric data repartition before the
landslides occurrence. In the landslide areas:

– 70.7% of these zones are FS < 1,

– 26.6% are contained between 1 and 1.8, Figure 8. Factor safety distribution in the landslides zones.
– 2.7% are Fs > 1.8 (Figure 8). The line graph represents the cumulated percentage and the
bar chart represents the percentage for each class.
The FS values superior to 1 can have several dif-
ferent origins: a bad calibration of the model, and/or In a densely populated zone, in order to have the
a landslide mapping problem, and/or unexpected most efficiency map on a long period of time, a
conditions variation (geomorphological, mechanical, re-evaluation of the several parameters is necessary:
hydrogeological, anthropological . . . conditions). (topography, anthropogenic perturbations . . .).

728
5 CONCLUSION Darcy, H. 1856. Les fontaines publiques de la ville de Dijon.
V. Dalmont, Paris.
This paper presents the LPC approach for evaluating Desbordes, M. 1987. Contribution à l’analyse et à la
slope failure in wide study area. This determinis- modélisation des mécanismes hydrologiques en milieu
tic model was described and applied on the Menton urbain. Thèse de Doctorat d’Etat, Université Montpellier
2, Montpellier, France, pp. 242.
municipality (Maritimes Alps, France). This method- Fuentes, C., Haverkamp, R. & Parlange, J.-Y., 1992. Param-
ology requires numerous calibrations: mechanical eter constrains on closed-form soilwater relationships.
and hydrogeological parameters are necessary. The J. Hydrol., Vol. 134, pp. 117–142.
dynamic hydrogeological modelling is an important Horton, R.E., 1933. The role of infiltration in the hydrologic
aspect of this approach to model the FS map. Never- cycle: EOS, Transactions, American Geophysical Union,
theless the database accessibility remains possible and Vol. 14, pp. 446–460.
permits to apply this methodology on several types of Montgomery, D.R. & W.E. Dietrich, 1994. A Physically
regions. Based Model for the Topographic Control on Shallow
The results obtained in this study indicate that, for Landsliding, Water Resources Research, 30 (4): pp.
1153–117.
a 10 m DEM resolution, the LPC methodology is a Musy, A. & Soutter, M., 1991. Physique du sol. Laussane.
successful application on the Menton area. Indeed, Presses Polytechniques et Universitaires Romandes, 335p.
for a 20-year return rainfall, 70.7% of the landslide O’Loughlin, E.M., 1986. Prediction of surface saturation
occurred have a FS inferior to 1. zones in natural catchments by topographic analysis.
A classification of the FS map can gives a haz- Water Resources Research, Vol. 22: pp. 794–804.
ard map depending on the model. Nevertheless, the Risser, V., 2000. Mouvements de terrains sur le versant ori-
expert’s evaluation stays an important step to refine ental du volcan Rucu Pichincha (Quito, Equateur). Projet
on the field and to conclude the hazard maps. Sishilad—EMAAP-Q/INAMHI/IRD. 67p.
Selby, M.J., 1993. Hillslope material and processes. Edition
Oxford. Second edition. 445p.
Spencer, E., 1967. A method of analysis of the
REFERENCES stability of embankments assuming parallel inter-slice
forces. Geotechnique; Vol. 17. pp. 11–26.
Beven, K. 1997. Topmodel: A critique. Hydrological Pro- Tarboton, D.G., 1997. A new method for the determination of
cesses. 11:1069–1085. flow directions and upslope areas in grid digital elevation
Borga, M., Dalla Fontana, G., Gregoretti, C. & Marchi, L., models. Water Resources Research, 33 (2): p. 309–319.
2002. Assement of shallow landsliding by using a phys- Wilson, J.P. & Gallant, J.C., 2000. Terrain Analy-
ically based model of hillslope stability. Hydrological sis—Principles and Applications. Wiley, New York,
Processes, Vol. 16, pp. 2833–2851. pp. 87–132.

729
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Back-analyses of a large-scale slope model failure caused by a sudden

drawdown of water level

G.W. Jia, Tony L.T. Zhan & Y.M. Chen

MOE Key Laboratory of Soft Soils and Geoenvironmental Engineering, Zhejiang University, Hangzhou, China

ABSTRACT: Slope failure caused by a sudden drawdown of water level was simulated in a large-scale model
box (15 m by 5 m by 6.5 m). The model test revealed a retrogressive multiple rotational sliding in a loose
silty soil slope. This paper presented the back-analysis work of the large-scale model test. Transient seepage
analyses were conducted to simulate the seepage field induced by rapid drawdown. Slope stability analyses were
conducted to simulate the retrogressive sliding observed in the model test. The results from the back analyses
on the seepage field and slope stability were basically consistent with the observation.

1 INTRODUCTION calculated from the transient seepage analyses. The

main objective of the analyses is to investigate the
Sudden drawdown of water level usually occurs to an seepage field and slope stability of the slope during
earth dam or an embankment along a river or a coastal the rapid drawdown.
line due to the climate change, sudden breaching or
periodical tide. The sudden drawdown usually affects
the stability of the slope on the verge of water and even
2 LARGE-SCALE MODEL TEST SIMULATING
causes slope failures. Slope failures induced by water
THE SUDDEN DRAWDOWN
level drawdown are often reported (Morgenstern,
1963; Nakamura, 1990; Liao et al, 2005). Jones (1981)
The large-scale model slope experiment was con-
investigated the landslides occurred in the vicinity of
ducted in Zhejiang University. The model slope was
Roosevelt lake from 1941 to 1953 and found about
constructed with sandy silt. The geometry of the model
30% of the landslide happened due to the drawdown
slope is illustrated in Figure 1. The initially constructed
of the reservoir. Nakamura (1990) reported that about
slope model is 15 m in length, 5 m in width and 6 m
60% of the landslides around reservoirs occurred
in height. The model slope consists of homogenous
under a drawdown condition in Japan.
sloping ground that was 6 m thick upslope and 2 m
Numerous desk studies have been conducted to
thick downslope with the slope angle of 45◦ , giving a
investigate the performance of slope subjected to a
sudden drawdown (Lane et al. 2000; Zhang et al. 2005; net model slope height equal to 4 m. The crest and toe
Liao et al. 2006), but most of the previous numerical of the slope are 5 m and 6 m long, respectively.
studies usually focused on a specific assumed slope
and no experiments were carried out to verify the anal-
ysis results. In these analyses, the slip surface was
assumed and the slope failure mode can’t be simulated.
Because it is difficult to obtain the accurate input data,
such as slope geometry, phreatic lines and the slip sur-
face, so few back-analyses of slope failures caused by
rapid drawdown have been carried out.
This paper presents the back-analyses work on
a large-scale model slope experiment conducted in
Zhejiang University. In this paper, transient seep-
age and limit equilibrium analyses have been con-
ducted for the slope subjected to rapid drawdown.
The stability of the slope during the drawdown pro-
cess is analyzed incorporating the pore water pressures Figure 1. Three dimensional view of the initial slope model.

731
 6 elevating steps and each elevation increment was
1 m and sustained for a period of typically 24 h.
The initial constructed slope surface collapsed dur-
ing the water level rising process and the slope angle
become approximately 33◦ (Zhan et al, 2007). Sec-
ondly, the sudden drawdown of water level outside the
slope surface was started after the final step of water
rise had lasted for 72 hours. The water level outside
the slope was shown in Figure 2. The drawdown rate
is about 1 m/s. A typical retrogressive multiple rota-
tional slide occurred to the slope model during the
sudden drawdown process (Fig. 3). There were totally
three displaced masses taken place in succession and
three slip surfaces were observed (Fig. 4).

Figure 2. Change of water level outside the slope during the 3 BACK-ANALYSES OF THE SLOPE FAILURE
drawdown process.
3.1 Analyses procedures
For investigating the pore water pressure distributions
in the model slopes and then on the slope stability,
firstly, transient seepage analyses for saturated and
unsaturated soils are carried out using a computer pro-
gram SEEP/W (Geo-slope, 2003). After obtaining the
pore water pressure distributions from the transient
seepage analyses, limit equilibrium analyses are then
carried out to determine the factor of safety of the
slopes using SLOPE/W (Geo-slope, 2003). The shear
strength of unsaturated soils can be represented by an
extended Mohr-Coulomb failure criterion (Fredlund &
Rahardjo, 1993).
The finite element mesh of the slope is shown in
Figure 5. The downstream boundary subjected to the
rapid drawdown is defined as known total head func-
tion which is shown in Figure 2. The drawdown process
Figure 3. The retrogressive multiple rotational landslide. (156 minute) is divided into 156 time step with one
minute for one time step in the numerical analyses. The
downstream boundaries is specified as review bound-
ary which means that if the total head is smaller than
the corresponding elevation the flux will be zero at
the node. All the other boundaries are specified as
zero flux boundaries. The initial groundwater condi-
tions for transient seepage analyses are established by
conducting a steady state analysis, in which the initial
water level is specified at the elevation of 5.6 m.

3.2 Hydraulic properties for seepage analyses

Figure 4. The observed slip surface of the slide.
To simulate transient seepage in unsaturated soils, it
is essential to specify hydraulic parameters includ-
ing soil-water characteristic curves (SWCC), saturated
A water-level control system was developed to sim- water permeability and permeability functions for the
ulate the rise and drawdown of water level in the slope sandy silt. The soil-water characteristic curve of the silt
model (Zhan et al. 2007). Firstly, the water level inside was obtained from volumetric pressure plate extractor
and outside the slope was elevated by the water-level and the results are shown in Figure 6. The saturated
control system. The elevating program comprised of permeability of the soil material was calibrated by

732
Figure 5. The finite element mesh of model slope for
seepage analyses.

Figure 7. Permeability functions for the sandy silt.

new slope, the slope surface is the critical slip sur-

face. In the analysis, the shear strength resistance
of the previous slide debris was neglected. Saucer
et al. (1983) divided the retrogressive landslide into
several blocks and each successive block combina-
tion is considered to move as a coherent unit. In this
study, Saucer’s method was adopted to analyze the
retrogressive landslide.

Figure 6. Soil-water characteristic curve for the sandy silt.

4 BACK-ANALYSES OF THE SLOPE FAILURE

4.1 Results from transient seepage analyses

varying the hydraulic permeability until the computed
phreatic surface matched the observation during the Phreatic lines and pore water pressure field for every
water level rising process. The calibrated permeabil- minute can be obtained from the seepage analyses.
ity for the soil was 1.0 × 10−5 m/s. The permeability Figure 8 shows phreatic lines changing with time
function was predicted from the SWCC by using the during the drawdown process. It can be seen that
method proposed by Fredlund et al. (Fredlund et al, the phreatic lines inside the slope show a signifi-
1993). The calculations were assisted by the program cant delay to the water level outside the slope. It is
SEEP/W. The estimated permeability function for the unfavorable to the stability of the slope. The delayed
soil is shown in Figure 7. responses of phreatic lines inside the slope are consis-
tent with the phenomenon that outward seepage was
observed on the slope surface during the drawdown
process.
3.3 Approach for analyzing retrogressive In order to verify the back-analysis results, the
slope failure numerical results of pore-water pressure at the position
Limit equilibrium method of slices is commonly used of Piezometer-4 were compared with the observation.
for calculating the factor of safety of a slope. In this Figure 9 shows the comparison. Due to the matches
study, the Morgenstern-Price method of slices has between the computed and the field-measured pore-
been adopted for the calculation. water pressure to a certain extent, the results from
Though there have been many methods for the anal- seepage analyses can be in accordance with the fact.
ysis of slope stability, few numerical investigations
on landslides with retrogressive mode is available in
4.2 Calculation of factor of safety
the literature. Only few simplified approaches have
been proposed to analyze such failures (Saucer 1983; In this section, the pore-water pressure distributions
Li et al, 2005). Li et al (2005) assumed that the obtained from the seepage analyses were incorporated
soil mass above the critical slip surface is slipped into a slope stability analyses to find the factor of
down completely and the soil mass below forms a safety. It can be seen from Figure 3 that sliding distance

733
Figure 8. Phreatic lines changing with time during the
drawdown process.

Figure 11. Back analyses of the combination of blcok 1 and

block 2.

Figure 9. The comparison between the pore-water pressures

from numerical analyses and experiment.

Figure 12. Back analyses of the combination of blcok 1,

block 2 and block 3.

When the water level outside the slope drawdown

for 0.7 m at the 40th minute, the first slide block
(named block 1) formed and slide downward with a
Figure 10. Back analyses of block 1. rapid speed. The factor of safety value corresponding
to slip surface 1 calculated by SLOPE/W was shown
in Figure 10. In the slope stability analyses, both the
for block 2 and block 3 is relatively small to the slope matric suctions above the phreatic line and the thrust
dimensions, so the Saucer’s method can be applied of water outside the slope surface were taken into
in the slope stability analyses. The slope failure sur- account. It can be seen that the factor of safety obtained
faces were specified to the three observed slip surfaces from the back analysis was slightly lower than 1. The
observed from the observation windows (Fig. 4). slope safety factor is consistent with the observation.

734
When the factor of safety was lower than 1, the slope REFERENCES
failure initiated.
When the water level was drawdown for 1.7 m at the Fredlund, D.G. & Rahardjo, H. 1993. Soil Mechanics for
100th minute, the second block and the third block ini- Unsaturated Soils, New York: John Wiley & Sons, Inc.
tiated nearly at the same time. Figure 11 shows the back GEO-SLOPE International Ltd. 2003. SEEP/W and
analyses of the combination of blcok 1 and block 2. SLOPE/W for finite element seepage analysis, vol.5.
Users’ Manual. Calgary, Alberta, Canada, 2003.
The factor of safety of the combination of blcok 1 and Lane, P.A. & Griffiths, D.V. 2000. Assessment of stability of
block 2 was slightly lower than 1 when the slip sur- slopes under drawdown conditions. J Geotech Geoenviron
face 2 is the critical slip surface. Figure 12 shows the Eng, ASCE 126 (5): 443–450.
back analyses of the combination of blcok 1, block 2 Li, S., Yue, Z.Q., Tham, C.F. et al. 2005. Slope failure in
and block 3. The factor of safety of the combination under consolidated soft soils during the development of a
of blcok 1, block 2 and block 3 was 0.85, which is port in Tianjin, China. Part 2: Analytical study. Canadian
much lower than 1. The difference between the calcu- Geotechnical Journal 42: 166–183.
lated factor of safety and practical slope failure may Liao, H.J., Sheng, Q., Gao, S.H. et al. 2006. Influence of
be caused by the sliding mass stacking at the foot of drawdown of reservoir water level on landslide stability.
Chinese Journal of Rock Mechanics and Engineering, 24
the slope. (19): 3454–3458. In Chinese.
Liao, Q.L., Li, X. & Dong, Y.H. 2005. Occurrence, geology
and geomorphy characteristics and origin of Qianjiang-
5 CONCLUSIONS ping landslide in Three Gorges Reservoir area and study
on ancient landslide criterion. Chinese Journal of Rock
In this paper, the back-analyses of a large-scale slope Mechanics and Engineering 24 (17): 3146–3153. In
model failure caused by a sudden drawdown of water Chinese.
level were carried out. The phreatic lines inside the Morgenstern, N. 1963. Stability charts for earth slopes during
slope show a significant delay relative to the water rapid drawdown. Geotechnique 13 (2): 121–131.
Nakamura, K. 1990. On reservoir landslide. Bulletin of Soil
level outside the slope. The phreatic lines and pore- and Water Conservation, 10 (1): 53–64. In Chinese.
water pressure of the back-analyses were basically Saucer, E.K. 1983. The Denholm landslide, Saskatchewan.
consistent with the observation. Part II: Analysis. Canadian Geotechnical Journal, 20:
The procedures that the successive sliding blocks 208–220.
combination were assumed to move as a coherent was Zhan, L.T., Zhang, W.J. & Chen, Y.M. 2006. Influence of
employed to analyze the retrogressive landslide. The reservoir level change on stability of a silty soil bank.
calculated values of factor of safety were consistent The fourth international conference on unsaturated soils,
with the observation. It is demonstrated that procedure ASCE: 463–472.
was feasible to analyze retrogressive landslides. Zhan, L.T., Jia, G.W. & Chen, Y.M. 2007. A large scale
model test simulating a slope failure caused by a sudden
drawdown of water level. Proceeding of the 3rd Asian
conference on unsaturated soils: 531–536.
ACKNOWLEDGMENTS Zhang, W.J., Chen, Y.M. & Ling, D.S. 2005. Seepage and
stability analysis of bank slopes. Chinese Journal of
The authors would like to acknowledge the financial Hydraulic Engineering 36 (12): 1510–1516. In Chinese.
support from the Key Technology R& D Program for
the eleventh five years provided by the Ministry of
Science and Technology of the People’s Republic of
China.

735
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Effect of Guangxi Longtan reservoir on the stability of landslide at Badu

station of Nankun railway

Riguang Jiang, Rongguo Meng, Aizhong Bai & Yuliang He

Guangxi Survey Institute of Hydrogeology and Engineering Geology, Liuzhou 545006, China

ABSTRACT: The landslide at the Badu station of Nankun railway is an ancient landslide. The volume of the
main landslide is about 3.772 million m3 , and that of the secondary is about 0.8344 million m3 . The landslide
was activated by construction of Nankun railway. However, the ten-year monitoring indicates that the landslide
did not have any deformation after maintained. If the Longtan reservoir is running, the stored water would have a
negative effect on the stability of the landslide. It is necessary to reinforce the existing stabilization engineering,
which is calculated and analyzed in terms of the water level change.

have no obvious characteristics of slipping. The

Figure 1. Overlook of Badu landslide. bedding is made up of light gray, gray green cal-
careous, siliceous quartz sandstone and gray black
mud, shale of the Triassic EC Bianyang Group
(T2 b). In front of the formation, there are allu-
1 CHARACTERISTICS OF THE LANDSLIDE vial sand and gravel layers, which indicate that the
ON BADU STATION OF NANKUN RAILWAY landslide has moved forward and covered fluvial
sediments. Three reverse faults and one transla-
1. The landslide lies on the bank of Nanpan river with tion fault has developed in the bedrock of the slide
the altitude of about 367 m in the front and 556–568 bed (the width of the broken belt is 10–60 m), and
m in the rear. The main slide body extends at the intersected each other.
direction of 145◦ –175◦ with the length of about 300 3. It is an ancient landslide. The emergency manage-
m, the width of 460–570 m, the thickness of about ment was taken on the main landslide after partially
23 m, the area of 16.40 × 104 m2 and the volume revived, including two rows of anchor piles, ground
of about 377.20 × 104 m3 . Whereas the secondary and underground drainage engineering measures.
landslide is about 280–310 m in length, 320–440 The controlling measures were adopted on the
m in width, and about 7 m in thickness, with the secondary landslide, such as prestressed chain,
area of 11.92 × 104 m2 and the volume of about cantilever piles, ground and underground drainage.
83.44 × 104 m3 (Figure 1). According to monitoring data, there was no defor-
The landslide in the front was badly collapsed, mation observed in the secondary landslide during
the front and the middle of which were revived. The last 10 years.
new body of landslide has the length of about 280
m, the width of 460–570 m, the area of 13.24 ×
104 m2 , the average thickness of about 25 m, and
the volume of about 331.00 × 104 m3 . Because 2 MAIN OBJECTIVES
of Nanning-Kunming railway built between 1996
and 1997, as well as the change of the geologic It is necessary to evaluate the effect of Longtan
condition such as hydrogeological condition, the reservoir on the stability of the landslide, based on
secondary landslide was revived with the volume investigation of the Badu station landslide’s character-
of about 83.44 × 104 m3 . istics. As long as the reservoir runs, the water table
2. The body of the landslide is mainly made up of will arise and accordingly affect the stability of the
detritus, angular boulder and clayey soil. The old landslide. Therefore, the characteristics of the stored
sliding zone is mainly composed of gray green, water on Longtan reservoir would determine the sta-
brown, yellow brown soft-plastic to plastic silty bility of the landslide, which can be grouped into 5
clay containing gravel which was visibly burnished, cases shown as follows (Ministry of Construction of
while the new deformation sliding zone is weak and the People’s Republic of China, 1999).

737
 Case 1: The present state with the water level of 368
m, including before the building of railway, Figure 2. The sketch map of Badu landslide’s profile for
between after the building of railway and evaluating.
the maintenance of the slide, and after the
maintenance of the slide;
Case 2: The present state in case of the flood once 3.3.1 Limit equilibrium analysis
every hundred year with the water level of Three vertical axial profiles (Profile 1, Profile 2 and
387.01 m, including before the building of Profile 3) were selected as a model. Because the
railway, between after the building of railway ancient landslide was revived to a new body deforma-
and the maintenance of the slide, and after the tion after the excavation for the railway, the stability of
maintenance of the slide; each profile is separately calculated by the following
Case 2–1: The present state in case of the flood once sections: original main landslide (C’-A’ section); the
every hundred year, after the slide was main- original secondary landslide (A-B Section); the whole
tained, with the water level between 368 and landslide composed of the main landslide and the sec-
387.01 m; ondary landslide (A-A ’section); the new deformable
Case 3: The reservoir operation after the slide is main- body as a result of the railway excavation, in which
tained for the railway building, with the water Profile 1 and Profile 2 are divided into two new
level of 375 m in the near future and 400 m in deformable bodies (IV1 , IV2 ), IV1 is the B’-A’ section,
the future; IV2 is the C’-A’ section, and Profile 3 is composed of
Case 4: The reservoir operation in case of the flood one new deformable body (IV) (Figure 2).
once every hundred year after the slide is main-
tained for the railway building, with the water 3.3.2 Finite element analysis
level decreasing 375 m from to 392.50 m in Geological analysis and mode of the Badu Landslide
the near future and from 407.33 m to 400 m in geological characteristics are used to simulate the
the future. groundwater flow, slope stress field and deformation
Therefore, it can be seen that Cases 1 and 3 are the field of the Badu landslide in two- and three-dimension
normal conditions, while Cases 2 and 4 are exceptive. in three conditions, including the natural condition, the
railway construction condition and the stored water
condition, in order to estimate the stability of Badu
3 METHODS landslides and to forecast deformation of the landslide.

3.1 Data collection

The first step is to collect related data obtained from 4 RESULTS AND DISCUSSIONS
the predecessors. The data include results of vari-
ous studies and investigation, information on design It is reasonable to synthetically confirm the parameters
and construction completion, and ten-year monitoring of slip soil depending on the lab test with the samples
data.Then do further research to the information. from the adits, the large in-situ shearing test and the
result of the anti-count on the untreated condition after
railway excavation.
3.2 Complementary reconnaissance
Comprehensive approaches was used to further iden-
4.1 Limit equilibrium analysis
tify the basic features of the landslide and to investigate
the effectiveness of the accomplished project, such as 1. Results of the limit equilibrium analysis show that
engineering geological mapping, four adits, trench- the ancient landslide was at stable state prior to
ing, geophysical exploration, large shear test in-situ, railway construction.
drilling, laboratory tests, the rigidity test, water- 2. Between the railway construction and the treatment,
injecting test and dating of sliding zone (Ministry of the landslide was stable or less stable, while it was
Construction of the People’s Republic of China, 2002). less stable or instable when the flood took place
once every one hundred-year.
3. Treatment increased the stability of the landslide,
3.3 Analytical methods
which was stable or less stable in the flood once
Both limit equilibrium analysis (Ministry of Construc- every one hundred-year.
tion of the People’s Republic of China, 2002) and finite 4. Because of the reservoir running, the landslide
element analysis were used to evaluate the effect of the became less stable at diverse extends: the stability
reservoir (Shibiao Chang et al.1994) on the stability of of the secondary landslide was less affected, while
the landslide. the new deformable body on the main landslide

738
 was significantly affected. The original main land- level up to 400 m. The displacement will reach 47
slide and the original secondary landslide were at mm, mainly in the front edge of the landslide below
steady state, while the whole landslide was stable the water level.
or less stable. The new deformation body of the
main landslide was less stable in Case 3. In summary, the reservoir has impacted effects on
5. When the water level drops from 392.5 to 375 m or the landslide as follows.
from 407.33 to 400 m due to the flood once every Firstly, the reservoir increased groundwater water
100-year, the landslide in every profile was less table and caused the change of the natural groundwater
stable. flow in the landslide, which would negatively affect the
stability of the landslide.
Therefore, local parts of the landslide will possibly Secondly, after the reservoir run, the project pro-
slide after the reservoir functions. tecting the landside from sliding would play a float-
ing role, which would decrease the stability of the
landslide.
4.2 Finite element analysis Thirdly, the descent of the water level in the reser-
1. Three-dimension Modflow software was used to voir for flooding protection led to higher penetration
simulate the groundwater flow and dynamic char- pressure. This increased the downward stress imposed
acteristics of the Badu landslide in natural condition on the landslide.
and on the conditions of water level of 375 m and Fourthly, the dammed water and the change of
400 m in the reservoir. The pressure of pore water the water level acting on the bank reduced the skid
and the change of the hydraulic gradient were cal- resistance, while increased the downward stress.
culated in terms of the water level change in the
reservoir.
2. The result of the Badu landslide stress analysis 4.3 Overall merit
indicates that the Badu landslide is stable in nat- 1. There is a drainage system on the land surface,
ural conditions, where there is tensile stress only at which need be repaired in time.
local slope surface. The landslide is easily revived 2. The underground drainage works have put into
by induced factors, such as rainstorm, excavation, effect in the secondary Landslide, while there are
and reservoir. There was tensile stress on the exca- less engineering measures in the main landslide. It
vated slope, and partial parts were damaged. With is proposed to build underground drainage channel
the excavation of railway, the stress and the ten- instead of the adits.
sile stress of the landslide increased in case of 3. The project to counteract sliding has been built in
heavy rain, and the value of destroyed degree η the secondary landslide. Some of anti-sliding pegs
increased dramatically, which made the landslide are slightly shallow in the rock in the main landslide
plastic damage. After the comprehensive repair, the (below the rail), and few prestressed anchors and
landslide was generally stable. However, its sta- chain are damaged. If the reservoir is filled to a high
bility would decline because of Longtan reservoir, level or the high water level drops, the shore will
especially at the water level up to 400 m, which serious collapse. That has a serious effect on the sta-
would make local part be damaged. Therefore, the bility of the main landslide. Therefore, it is essential
reinforcement must be carried out in order to ensure to accomplish the bank protection and reinforce-
the safety of the landslide. ment works before the reservoir runs, in order to
3. The results of the Badu landslide deformation protect the landslide and to avoid endangering the
and the stability status analysis indicate that the Badu part of the Nanning-Kunming railway.
Badu landslide was stable in natural conditions.
It was deformed only in the initial stages, and
gradually tended to slow creep state and eventu-
ally became stable while the time went by. The ACKNOWLEDGEMENTS
displacement of the landslide was less, primarily
occurred in the angle of excavated slope after the The authors would like to express their thanks to
railway had been excavated. In case of the heavy experts and professors in Longtan Hydropower Devel-
rain, the displacement increased rapidly (up to opment Co., Ltd., Southeast Electric Power Design
120 mm). Accordingly, the landslide was revived Institute, Liuzhou Railway Bureau, Railway Sur-
and instable. With the comprehensive repair, the vey and Design Institute, the Ministry of Land and
deformation tended to slow down and eventually to Resources, the Ministry of Railways, the former Min-
stabilize. However, the deformation will increase istry of Water and Electricity, Guangxi University,
and the landslide stability will decline because of Chengdu University of Technology and Chongqing
Longtan reservoir, especially in case of the water University for their guidance and helps during the

739
reconnaissance, analysis and the modification of the Ministry of Construction of the People’s Republic of
disquisition. China.2002.Technical code for building slope engineer-
ing, GB50330–2002. Beijing: China Archicture & Build-
ing Press.
Ministry of Construction of the People’s Republic of
REFERENCES China.2002.Code for investination of geotichincal engi-
neering, GB50021–2001.Beijing: China Archicture &
Ministry of Construction of the People’s Republic of Building Press.
China.1999.Code water resoure and hydropower engi- Shibiao Chang et al.1994. Handbook of geologic engi-
neering geological investigation, GB50287–99. Beijing: neering, the third edition. Beijing: China Archicture &
China Archicture & Building Press. Building Press.

740
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Application of SSRM in stability analysis of subgrade embankments

over sloped weak ground with FLAC3D

Xin Jiang
Key Laboratory of Road & Traffic Engineering of the Ministry of Education, Tongji University, Shanghai, P.R. China

Yanjun Qiu
School of Civil Engineering, Southwest Jiaotong University, Chengdu, P.R.China

Yongxing Wei
No. 2 Civil and Architecture Design Institute, China Railway Eryuan Engineering Group Co. Ltd., Chengdu,
P.R. China
Jianming Ling
Key Laboratory of Road & Traffic Engineering of the Ministry of Education, Tongji University, Shanghai, P.R. China

ABSTRACT: Subgrade embankments over sloped weak ground exhibit different deformation behavior due to
downhill sliding potential compared to filling over flat ground. The stability analysis of subgrade embankments
over sloped weak ground plays an important role in embankment design process. Some numerical simula-
tion software packages including Shear Strength Reduction Method (SSRM) to analysis slope stability were
summarized. Based on Finite Difference Method (FDM) software package FLAC3D, 3D stability analysis of
embankments over sloped weak ground was performed. The influences of some model geometry parameters,
such as the weak subsoil layer longitudinal length and vertical thickness, on potential slip surface behavior and
Factor of Safety (FS) were discussed. The research results present the considerable differences between the slip
surface behavior and Factors of Safety (FS) of an embankment over sloped weak ground estimated from 2D and
3D numerical calculations. The slip surface behavior depend on the weak subsoil layer longitudinal length and
vertical thickness seriously. It seems that FS obtained from 2D calculations may be underestimated when the
weak subsoil layer longitudinal length is limited. FS obtained from 2D calculations were lower than from 3D.
With the rapid development of computer hardware and available commercial numerical simulation software,
application of Shear Strength Reduction Method (SSRM) in 3D seems to be a reasonable alternative to 2D
analysis to take the complexity of geology under consideration—especially the presence of thin and weak strata.

1 INTRODUCTION the special case of sloped weak ground (the sur-

face weak layer thickness is zero) as illustrated in
The surrounding mountains around Sichuan basin pro- Fig. 1.
duce many geotechnical challenges for highway and Two major concerns arise in highway and railway
railway engineers such as deep cut, high fill, cut-fill construction over sloped weak ground: embankment
transition, sloped weak ground, and other uncon- stability and lateral deformation. Embankment sta-
ventional subgrade design in southwest China. For bility may cause severe problems without sufficient
example, there are about 80 sections and total length design consideration and full understanding of struc-
11.8 km sloped weak ground in Yuhuai (Chongqing- tural mechanism. Literature reported that embankment
Huaihua) railway project, a newly built main railway and pavement failure in railway and highway construc-
route in southwest China. Due to its downhill slid- tion with fills over sloped ground as shown in Fig. 2
ing potential, subgrade embankments over sloped and Fig. 3, including sliding failure, pavement surface
ground exhibit different deformation behavior as com- longitudinal crack due to intolerable uniform deforma-
pared fills over flat ground and the trend will be tion. Anti-slide piles or pile net foundation in sloped
strengthened when there is a weak layer over sloped weak ground were designed in original plan to increase
ground. So sloped ground could be regarded as slope stability in Fig. 4.

741
Figure 1. Engineered embankments over natural sloped
ground.

Figure 4. Anti-slide piles (photo by Yongxing WEI).

sloped ground. You et al. (2002) studied the construc-

tion process of railway fill over sloped weak ground in
plateau area. Later Jiang et al. (2006, 2007a) explored
the stability of railway fills over sloped weak ground
using 2D strength reduction method (SRM) with com-
mercial software Plaxis. Qiu et al. (2007) studied
the deformation and stability behavior of highway fill
over sloped weak ground and influence on pavement
responses.
It can be concluded based on literature review that
Figure 2. Embankment failure (photo by Yongxing WEI). embankments over sloped ground need more research
efforts to have a better understanding of the defor-
mation and stability behavior in order to design a
sound upper structures, such as pavement or rail track.
The stability analysis of embankments over sloped
weak ground plays an important role in embankment
design process without doubt. Traditional 2D limit
equilibrium method and FEM were used to perform
the stability analysis in the existing research and the
limit longitudinal length of the weak subsoil will be
ignored, however. 3D shear strength reduction method
(SSRM) was conducted in this paper to investigate
potential sliding mode and more realistic value of fac-
tor of safety (FS) of embankment over sloped weak
ground.
Figure 3. Pavement longitudinal crack (photo by Xin
JIANG).
2 BRIEF INTRODUCTION TO SSRM
AND SOFTWARE
Most research efforts on mechanical behavior of
embankments over sloped weak ground. Wei (2001)
2.1 Shear strength reduction method
established the concept of engineered fill over sloped
ground. Jiang et al. (2002, 2003) simulated defor- The shear strength reduction method (SSRM) is also
mation and stability behavior of railway embankment known as strength reduction method (SRM). By auto-
over sloped ground using commercial software GEOS- matically performing a series of simulations while
LOPE, based on plane strain FEM and 2D limit changing the soil strength properties, the factor of
equilibrium method (LEM), while Luo et al. (2002) safety (FS) can be found to correspond to the point
conducted geo-centrifuge testing to investigate defor- of stability, and the critical failure (slip) surface can
mation behavior, stability characteristics and effective be determined (Griffiths et al., 1999). Actual shear
of engineering measures of railway subgrades over strength properties, cohesion (c) and friction (ϕ),

742
are reduced for each trial according to the following
equations:

ctrial = c/Ftrial (1)

ϕ trial = arctan(tan ϕ/Ftrial ) (2)

The SSRM provides advantages over traditional

limit equilibrium solution as follows (Itasca Consult-
ing Group, Inc, 2005b):
1. Any failure mode develop naturally; there is no Figure 5. FLAC3D model.
need to specify a range of trial surface in advance.
2. No artificial parameters (e.g., functions for inter-
slice force angles) need to be given as input. Table 1. Mechanical properties of soil layers.
3. Multiple failure surface (or complex internal yield-
ing) evolve naturally, if the conditions give rise Soil type E/MPa μ γ /(kN/m3 ) c/kPa ϕ/◦
to them.
Embankment 30 0.35 19 25 25
4. Structural interaction (e.g., rock bolt, soil nail
Sloped weak layer 5 0.40 18 10 10
or geogrid) is modeled realistically as fully cou- Stiff layer 300 0.25 22 200 50
pled deforming elements, not simply as equivalent
forces.
5. The solution consists of mechanisms that are kine-
matically feasible. (Note that the limit equilibrium
method only considers forces, not kinematics). (Itasca Consulting Group, Inc., 2005a). FALC3D is
widely used for slope stability analysis recently (Varela
Suarez et al., 2003; Cala et al., 2006). The FLAC3D
2.2 Software including SSRM analysis model can be seen in Fig. 5. In this model,
Due to the rapid development of computing effi- the embankment top width is assumed to be 28 m,
ciency, some geotechnical numerical simulation codes embankment height is 5 m, side slope ratio is 1:1.75
were developed to perform the slope stability anal- (V:H) and the ground inclination ratio is 1:5. The thick-
ysis using SSRM directly. Now FLAC2D, FLAC3D, ness t and the longitudinal length L of the weak subsoil
FLAC/SLOPE developed by Itasca Consulting Group, layer were changed to investigate the model geometry
Inc. in USA, Phase 2 V6.0 developed by Rocscience design parameter sensitivity. It is noted that the full
Inc. in Canada, Plaxis, Plaxis 3D Tunnel and Plaxis longitudinal length of embankment is assumed to be
3D Foundation developed by Plaxis B.V. in Nether- 100 m.
lands, Z-SOIL.PC developed by Zace Service Ltd.
in Swzerland, Geo FEM developed by FINE Ltd. in
Czech, MIDAS/GTS developed by POSCO Group 3.2 Material properties
in Korea, RFPA (Realistic Failure Process Analysis) Table 1 shows the mechanical properties assumed in
developed by Dalian Mechsoft Co., Ltd. and VFEAP the paper for the soil units.
developed by Zhejiang University in China are some In Table 1, E is elastic modulus, μ is Poisson’s
mainstream numerical simulation codes that can be ration, γ is soil unit weight, c is cohesive force and
performed the slope stability analysis using SSRM ϕ is inner friction angle. E and μ should be trans-
directly. FLAC2D, FLAC3D and FLAC/SLOPE are formed to bulk modulus K and shear modulus G
developed using finite difference method (FDM) and in FLAC3D input process. More information can be
the others are developed based on finite element found in FLAC3D manual (Itasca Consulting Group,
method (FEM) among the above codes. More infor- Inc., 2005).
mation about these software features and comparisons
can be found in Jiang et al. (2007b).
4 RESULTS AND DISCUSSION
3 FLAC3D NUMERICAL ANALYSIS MODEL
The comparison pictures of potential failure mode for
the longitudinal length of the weak subsoil L = 20 m,
3.1 Geometry dimensions
100 m (weak subsoil thickness t is 1 m) are pre-
FLAC3D developed by Itasca Consulting Group, sented in Fig. 6. The FEM computer code, Phase 2
Inc. was chosen to perform analysis in this paper V6.0 was used for 2D slope stability analysis using

743
 (a) 3D analysis using FLAC3D (L=20m, t=9m)

(b) 3D analysis using FLAC3D (L=100m, t=9m)

(c) 2D analysis using Phase2 V6.0 (t=9m)

Figure 7. Potential slip surface behavior (t = 9 m).

Figure 6. Potential slip surface behavior (weak layer thick-

ness t = 1 m). surface looks like to be a circle line in embankment
body and a beeline in weak subsoil layer, in other words
the failure slip surface is a composite slip surface.
The comparison pictures of potential failure mode
SSRM. The contours of shear strain rate and direction for the longitudinal length of the weak subsoil L =
of velocity vectors are clearly identifying failure sur- 20 m, 100 m (weak subsoil thickness t is 9 m) are pre-
face in FLAC3D. However, the potential failure mode sented in Fig. 6. The FEM computer code, Phase 2
is presented using the maximum shear strain or dis- V6.0 was used for 2D slope stability analysis using
placement vectors in Phase 2 V6.0. The failure surface SSRM. The contours of shear strain rate and direction
is only local when the longitudinal length of the weak of velocity vectors are clearly identifying failure sur-
subsoil L is 20 m but the failure surface is turned to be face. However, the potential failure mode is presented
global when the longitudinal length of the weak sub- using the maximum shear strain or displacement vec-
soil L is 100 m. The failure mode of 2D analysis is tors in Phase 2 V6.0. The failure surface is only local
only close to 3D analysis when the longitudinal length when the longitudinal length of the weak subsoil L
of the weak subsoil L is 100 m. It seems that the failure is 20 m but the failure surface is turned to be global

744
when the longitudinal length of the weak subsoil L is of the weak subsoil stratum. The value of FS is con-
100 m. The failure mode of 2D analysis is only close to stant up to the length of the weak subsoil stratum in
3D analysis when the longitudinal length of the weak 2D analysis. Increasing the weak subsoil layer result
subsoil L is 100 m. The failure slip surface is a circle in decreasing of FS value. The FS values of 3D anal-
slip surface but not a composite slip surface. The rea- ysis are larger than those of 2D analysis obviously
son is the weak subsoil layer thickness is thicker than especially when the weak subsoil thickness is thin rel-
before. So reasonable engineering measures should be atively. These means the 2D analysis will magnify
adopted to improve guidance of embankment design. the effect of weak subsoil layer because the longitudi-
Fig. 8 shows the potential slip surface behavior nal length in 2D analysis is considered as unlimited.
using SLIDE V5.0 developed by Rocscience Inc. FS obtained from 3D calculations slowly tends to the
based on limit equilibrium method (LEM). It can be factor of safety value obtained from 2D calculations.
seen that slip surface behaviors are similar to those Fig. 10 shows the relationship between FS and
obtained from numerical simulation. sloped weak layer thickness t. It can be seen that
Twenty five series of analysis were performed using increasing the sloped weak subsoil layer thickness will
FLAC3D and Phase 2 at the same time. Fig. 9 shows lead to the decreasing of FS value.
2D and 3D FS values for several longitudinal length Table 2 shows the 2D and 3D FS value comparisons
when the weak subsoil layer thickness vary (the longi-
tudinal length L = 20 m). It can be seen the largest FS
value difference reach to 0.57. The FS value difference
will be relatively flat when the weak layer thickness
t increase. In case of the longitudinal length of weak
subsoil layer, FS obtained from 2D calculations may
be seriously underestimated. Application of SSRM in
3D may produce a reasonable value of FS for most
cases.

Figure 8. Slip surface obtained from 2D LEM.

Figure 10. Factor of safety-sloped weak layer thickness t.

Table 2. FS comparisons between 3D model and 2D model.

Factor of safety (FS)

Weak layer
thickness 3D model 2D model FS
t/m (FLAC3D) (Phase 2) difference

1 1.68 1.11 0.57

3 1.37 0.95 0.42
5 1.26 0.90 0.36
7 1.22 0.88 0.34
Figure 9. Factor of safety-sloped weak layer longitudinal 9 1.21 0.87 0.34
length L.

745
5 SUMMARY AND CONCLUSIONS Itasca Consulting Group, Inc. 2005a. FLAC3D-Fast
Lagrangian Analysis of Continua in 3 Dimensions, Ver.3.0
Embankment over sloped weak ground is still a great User’s Guide. Minneapolis: Itasca Consulting Group, Inc.
challenge for geotechnical and highway engineers due Itasca Consulting Group, Inc. 2005b. FLAC/Slope User’s
to its complexity. Sufficient safety reserve is one major Guide. Minneapolis: Itasca Consulting Group, Inc.
Jiang, X., Wei, Y. & Qiu, Y. 2002. Numerical simulation of
concern in design and construction. Using FLAC3D subgrade embankment on sloped weak ground, Journal of
as the major simulation tool, this paper conducted Traffic and Transportation Engineering, 2(3): 41–46.
preliminary research on stability behavior of high- Jiang, X., Wei, Y. & Qiu, Y. 2003. Stability of subgrade
way embankments over sloped ground in 3D SSRM. embankment on slope weak ground, Journal of Traffic
Conclusions can be summarized as follows: and Transportation Engineering, 3(1): 30–44.
Jiang, X., Qiu, Y. & Wei, Y. 2006. Research on the
1. It seems that SSRM can be conducted to perform subgrade embankments engineering on sloped weak
slope stability analysis of embankment over sloped ground, Journal of Railway Engineering Society. (1):
weak ground. 32–35, 39.
2. Application of SSRM in 3D may produce a rea- Jiang, X., Qiu, Y. & Wei, Y. 2007a. Engineering behav-
sonable value of FS and potential slip surface ior analysis of subgrade embankments on sloped weak
behavior. ground based on strength reduction FEM, Chinese
Journal of Geotechnical Engineering, 29(4): 622–627.
3. SSRM in 3D often cost more computer time and Jiang, X. & Qiu Y. & Ling, J. 2007b. Comparisons of strength
need more computer hard disk space to store the reduction method software for slope stability analysis.
result files than 2D analysis and the effect of 3D Chinese Journal of Underground Space and Engineering.
is often considered as an additional safety reserve (in press).
so one must find a reasonable equilibrium between Luo, Q. & Zhang, L. 2002. Centrifuge testing of rail-
safety and economy. way embankments over sloped weak ground, Research
Report, School of Civil Engineering, Southwest Jiaotong
University, Chengdu, China.
ACKNOWLEDGEMENTS Qiu, Y., Wei, Y. & Luo, Q. 2007. Highway embankments
over sloped ground and influence on pavement responses.
International Conference of Transportation Engineering
This research is supported by Chinese Postdoctoral 2007, Chengdu, China, 22–24 July, China 2007. ASCE
Science Foundation (20060390654). The authors are Press. pp1615–1620.
grateful to Mr. Yongsheng ZHU of ITASCA Consult- Varela Suarez, A. & Alonso Gonzalez, L.I. 2003. 3D
ing China Ltd for his valuable assistances. slope stability analysis at Boinas East gold mine. In
Richard Brummer, Patrick Andrieux, Christine Detour-
nay & Roger Hart (eds), FLAC and numerical modeling
REFERENCES in geomechanics. Proceedings of the third international
FLAC symposium, Sudbury, Ontario, 21–24 October
2003. A.A. Balkema Publishers.
Cala, M., Flisiak, J. & Tajdus, A. 2006. Slope stability
Wei, Y. 2001. Stability evaluation methods of embankments
analysis with FLAC in 2D and 3D. In Pedro Varona
over sloped weak ground, Journal of Geological Disaster
& Roger Hart (eds), FLAC and numerical modeling in
and Environment Protection, 12(2): 73–79.
geomechanics. Proceedings of the fourth international
You, C., Zhao, C., Zhang, H. & Liu, H. 2002. Study on con-
FLAC symposium. Madrid, Spain, 29–31 May, 2006.
struction test of embankment on soft clay of plateau slope,
(CD-ROM). Minneapolis: Itasca Consulting Group, Inc.
Chinese Journal of Geotechnical Engineering, 24(4):
Griffiths, D.V. & Lane, P.A. 1999. Slope stability analysis by
503–508.
finite elements. Géotechnique. 49(3): 387–403.

746
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Strength parameters from back analysis of slips in two-layer slopes

J.-C. Jiang & T. Yamagami

Department of Civil and Environmental Engineering, The University of Tokushima, Japan

ABSTRACT: For a two-layer slope with a known geometry, unit weight and pore water pressure distribution,
the position of critical slip surface from a limit equilibrium stability analysis will remain unique for a given set
of (λ1 = c1 / tan φ1 , λ2 = c2 / tan φ2 , λ3 = c1 /c2 ) values regardless of the magnitude of individual strength
parameters (c1 , φ1 , c2 , and φ2 ) of two soil layers. This theoretical relationship between c and φ  and the critical
slip surface is used to develop a back analysis method by which the shear strengths of soils can be estimated from
a known position of the failure surface observed in the field two-layer slope. In this method, an objective function
is first defined to describe the difference between the actual failure surface and theoretical critical slip surface,
and then a nonlinear programming technique is used to minimize the difference (i.e. the objective function) so
as to determine an optimal solution of (λ1 , λ2 , λ3 ) that corresponds to the failure surface. When the magnitude
of (λ1 , λ2 , λ3 ) are obtained, a unique set of (c1 , φ1 , c2 , φ2 ) values can easily be computed by considering the fact
that the factor of safety is equal to unity. Results back-calculated from a failure surface involved in a two-layer
slope are presented to demonstrate the effectiveness of the proposed back analysis method.

1 INTRODUCTION tribution in a two-layer slope are given, the position

of the critical failure surface from a limit equilib-
Back analysis of slope failures has been a useful rium theory will remain unique for a particular set of
tool to determine soil strengths along the slip surface (λ1 = c1 / tan φ1 , λ2 = c2 / tan φ2 , λ3 = c1 /c2 ) values
especially in connection with remedial stabilization regardless of the magnitude of individual parameters
works. Existing back analyses have been performed (c1 , φ1 , c2 , φ2 ) of two soils. Based on this finding, a
only for failed slopes in homogeneous soils. In prac- back analysis method is established to determine the
tice, empirical methods are often used to estimate c four strength parameters of two soil layers involved. In
and φ by assuming one of these parameters and back- this method, an objective function with three indepen-
calculating the other for a factor of safety of unity dent variables (λ1 , λ2 , λ3 ) is first defined to describe
(e.g., Duncan and Stark 1992, Japan Road Associa- the difference between the actual failure surface and
tion, 1999). On the other hand, it has been suggested theoretical critical slip surface. Then, this difference
that the magnitude of both c and φ  can be deter- (i.e. the objective function) is minimized using a non-
mined by considering the position of the actual slip linear programming technique to obtain an optimal
surface together with the fact that the factor of safety solution of (λ1 , λ2 , λ3 ) that corresponds to the fail-
should be equal to unity (Saito 1980, Li and Zhao ure surface. When the magnitude of (λ1 , λ2 , λ3 ) are
1984, Yamagami & Ueta 1989, Greco 1996, Wesley & obtained, a unique set of (c1 , φ1 , c2 , φ2 ) values can
Leelaratnam 2001, Jiang & Yamagami, 2006). easily be computed by considering that the factor of
When shear failure occurs in slopes consisting safety is equal to unity.
of multi-layered soils, stability analysis requires the The proposed method satisfies the two essential
determination of the strengths of each layer involved. requirements for back analysis: i) the actual failure sur-
This paper presents a method to back-calculate the face must be consistent with a theoretical critical slip
strength parameters from the slip surface in a two- surface, and ii) the factor of safety should be equal to
layer slope. As estimation of the shear strengths is unity. Back analysis satisfying these conditions takes
based on the information provided by a failure sur- full advantage of the information provided by an exist-
face, the theoretical relationship between the (c , φ  ) ing landslide which represents the field large-scale
and the location of critical slip surface in a two-layer shear test conducted by nature. Therefore, the (c , φ  )
slope is studied. It is shown that when the slope values obtained in such a way can be expected more
geometry, unit weight and pore water pressure dis- reliable.

747
2 THEORETICAL RELATIONSHIP BETWEEN A
STRENGTH PARAMETERS AND LOCATION
OF CRITICAL SLIP SURFACE
layer 1
In limit equilibrium methods of slope stability analy- c1 , tan 1
sis, the factor of safety, F, is commonly defined as
B layer 2
shear strength of soil
F= (1) C c 2 , tan 2
shear stress required for equilibrium F
Failure surface
The shear strength of the soil is usually described
by the Mohr-Coulomb failure criterion as a function Figure 1. Failure surface in a two-layer slope.
of two parameters: cohesion c and internal friction
angle φ  . When the soil mass above a slip surface is constant). This means that if the strength parameters
divided into vertical slices, the definition of F can be (c1 , tanφ1 , c2 , tanφ2 ) are multiplied by μ, respectively,
expressed as: then the factor of safety for the same slip surface is also
   multiplied by μ in order to maintain a limit equilibrium
 Rf (c l + N tan φ  ) state for the same slip surface.
T= = Now let us consider an arbitrary two-layer slope
F F
  c with the (c1 , tanφ1 , c2 , tanφ2 ) soils where the critical
tan φ  
= l+ N (2) slip surface location with a minimum factor of safety,
F F denoted by F0 , is known. Equations (3) and (4) indicate
that if the soil layers in the slope are substituted by the
where Rf -the available shear resistance on the base of two artificial materials having the strength parameters
a slice, T-the shear force mobilized on the slice base,  
(c̄1 = μc1 , tan φ̄1 = μtanφ1 , c̄2 = μc2 , tanφ̄2 =
N -the effective normal force on the slice base, l-length 
of the slice base. T and N can be obtained by solv- μtanφ̄ 2 ) and the problem is reanalyzed using the
ing some or all of the equations of equilibrium. c /F same method of slices, the factor of safety defined
and tanφ  /F in Equation (2) are sometimes called the for each of the trial slip surfaces will be multiplied by
strength parameters that are necessary only to maintain μ, but the critical slip surface with the lowest factor of
the slope in limit equilibrium. In other words, F can safety, F̄0 (= μF0 ), will remain at the same position.
be defined as the factor by which the shear strength of It is of interest to note that in the cases men-
the soil would be reduced to bring the slope into a state tioned above the ratios of the strength parameters keep
of barely stable equilibrium (Michalowski, 2002). unchanged although individual strength parameter val-
For a two-layer slope such as the one shown in ues of the two soils are different. That is, the following
Figure 1, Equation (2) can be rewritten as relationships between the (c1 , tanφ1 , c2 , tanφ2 ) and c̄1 ,
 
tan (φ̄1 , c2 , tanφ̄2 ) are held.
   c tan φ  

T= l+ N c1 / tan φ1 = c̄1 / tan φ̄1 (5.1)
F F
  c tan φ1 

c2 / tan φ2 = c̄2 / tan φ̄2 (5.2)
= 1
l+ N
F F
AB c1 /c2 = c̄1 /c̄2 (5.3)
 c  
tan φ2 
+ l+ tan φ1 / tan φ2 = tan φ̄1 / tan φ2
2
N (3) (5.4)
BC
F F
When any three of these four ratios are fixed, the
From Equation (3) it follows that rest one will automatically be determined. In other
   c̄ tan φ̄1 
words, only three of the four ratios are independent of
T= 1
l+ N each other. For the sake of convenience, the following
AB
F̄ F̄ three ratios are chosen and used in this study.
 
c̄2 tan φ̄2 
+ l+ N (4) λ1 = c1 / tan φ1 (6.1)
BC
F̄ F̄


λ2 = c2 / tan φ2 (6.2)
where c̄1 = μc1 , tan φ̄1 = μtanφ̄1 , c̄2 = μ c2 ,
tan φ̄2 = μ tan φ2 , and F̄ = μ F (μ is a positive λ3 = c1 /c2 (6.3)

748
 Summarizing the above discussions, we can con- 0 5 10 1 5 (m) 20 25 30 35
clude that different combinations of (c1 , φ1 , c2 , φ2 ) 0
c1′ ( kPa ) tan 1′ c′2 (kPa ) tan ′2 Fmin (safety factor)
with a same set of (λ1 , λ2 , λ3 ) values will result in the 9.8 0.1 82 3.92 0.337 0.9
1 9.6 0.364 7 .84 0.675 1 .8
identical critical slip surface. That is, when all other 5 29.4 0.546 1 1 .7 6 1 .01 2 2.7
conditions except for (c , φ  ) are the same (the same 1 = 53.85( kPa ), 2 = 11.62( kPa )
laye r 1
slope geometry, the unit weight of soils, the pore water (m) 1 0
and 3 = 2.50 for all the
le
pressure distribution), the position of the critical slip three cases r tab c1′ , tan 1′
wate layer 2
surface in a given two-layer slope depends only on a c′2 , tan ′2

set of (λ1 , λ2 , λ3 ) values regardless of the magnitude 15

critical slip surf aces
of individual strength parameters. Moreover, while the =17.64 kN / m 3 , = 18.62kN / m 3
1 2
value of the minimum factor of safety varies as the 20
strength parameters change from (c1 , tanφ1 , c2 , tanφ2 )
Figure 2. Critical slip surfaces for different (c, φ) combi-
into (c̄1 , tanφ̄ 1 ,c̄2 , tan φ̄ 2 ), the following relationships
nations associated with a same set of (λ1 , λ2 , λ3 ) values.
exist.

c1 /F0 = c̄1 /F̄0 , tan φ1 /F0 = tan φ̄1 /F̄0 (7.1) 0 5 10 1 5 (m) 20 25 30 35
0
c2 /F0 c̄2 /F̄0 , tan φ2 /F0 tan φ̄2 /F̄0
critical slip Fmin (safety factor)
= = (7.2) surfaces 1 (kPa) 2 (kPa) 3

53.85 1 1 .62 2.50 0.9

5 1 38.96 1 0.49 5.00 1 .3
The above-mentioned relationship between the 1 0.49 1 38.96 0.20 1 .2
strength parameters and the location of critical slip
layer 1
surfaces is found directly from the definition of the (m) 1 0
r tablec′ , tan ′
factor of safety, and therefore, it is commonly valid for wate 1 1
layer 2
any existing method of slices. As an illustrative exam- c′2 , tan ′2
15
ple, the stability analysis of a two-layer slope such as
the one shown in Figure 2 was carried out using the =17.64 kN / m3 , = 18.62kN / m3
1 2
same unit weight and pore water pressure distribution 20
but different strength parameter values. Three different
combinations of (c1 , φ1 , c2 , φ2 ) associated with a same Figure 3. Critical slip surfaces for different sets of
set of (λ1 , λ2 , λ3 ) values were considered. For each of (λ1 , λ2 , λ3 ) values.
these cases, the critical slip surface was located using
the search approach by Baker (1980) which combined
A
A′
the Spencer method (1967) with dynamic program- A
ming technique. The critical slip surfaces obtained are
illustrated in Figure 2 together with their values of min- layer 1
imum factor of safety (Fmin ). It is seen from Figure 2 cc′1 , tan 1′
that the three critical slip surfaces completely coin-
yi layer 2
cide with each other. Also, the relationships shown in
Equation (7) are hold for all the cases considered. c′2 , tan 2′
B′
When the problem shown in Figure 2 was solved B O
O′
O
y1
using combinations of (c1 , φ1 , c2 , φ2 ) corresponding y2

to different sets of (λ1 , λ2 , λ3 ) values, the positions

of critical slip surfaces located were also different, as failure surface Critical slip surface
shown in Figure 3. Consequently, it may be concluded
that for a given two-layer slope with the known unit Figure 4. Difference between failure surface and critical
weight and pore water pressure distribution, the crit- slip surface (DIS = | yi |).
ical slip surface associated with a particular method
of slices corresponds uniquely to a set of (λ1 , λ2 , λ3 )
values. This signifies that it is possible to estimate the is denoted by F0 which is usually taken to be unity.
magnitude of (λ1 , λ2 , λ3 ) from a known position of the As mentioned previously, the position of the failure
actual slip surface observed in a field two-layer slope. surface can be used to determine the magnitude of
(λ1 , λ2 , λ3 ) that correspond to the true field values of
(λ10 , λ20 , λ30 ). In other words, by carrying out a con-
3 BACK ANALYSIS PROCEDURE ventional stability analysis with (c1 , φ1 , c2 , φ2 ) values
that satisfies (λ10 = c1 /tanφ1 , λ20 = c2 /tanφ2 , λ30 =
Figure 4 shows a two-layer slope in which a fail- c1 /c2 ), it is possible to obtain a critical slip surface that
ure plane, AOB, has occurred. Its factor of safety is consistent with the actual slip surface, AOB.

749
 Because the position of the failure surface, AOB, relationships in Equation (6), so that a set of (c1 , φ1 ,
corresponds uniquely to the λ10 , λ20 , λ30 ) values, the c2 , φ2 ) values are obtained.
stability analysis using another set of (λ1 , λ2 , λ3 ) will 2. Compute the factor of safety (the Spencer method
yield a critical slip surface, A O B (Figure 4), which is used in this study), F̄0 , of the failure surface using
is different from AOB. Now let us take note of the the (c1 , φ1 , c2 , φ2 ) values obtained in 1).
difference in the positions of these two slip surfaces, 3. Determine the magnitude of (c10 , φ10 
, c20 , φ20

) for
defined by DIS = | yi |, in which yi stands for the factor of safety of F0 (= 1.0) by substituting the
vertical distance between the failure surface and the (c1 , φ1 , c2 , φ2 , F̄0 , F0 ) values obtained in 1) and 2)
critical slip surface at each slice dividing line, as shown into the relationships in Equation (7).
in Figure 4. DIS can be regarded to be a function of
three variables (λ1 , λ2 , λ3 ) because the position of crit-
ical slip surface varies with the change in values of 4 VERIFICATION
(λ1 , λ2 , λ3 ). When this function reaches a minimum
(i.e. DIS = 0.0), a required solution of (λ1 , λ2 , λ3 ) is The accuracy of the proposed back analysis method
obtained that should be equal to (λ10 , λ20 , λ30 ), respec- was verified using a theoretical (hypothetical) slope
tively. Therefore, if DIS is defined as an objective failure in a two-layer slope such as the one shown in
function, the present back analysis can be mathemat- Figure 5. When the soil unit weight, pore water pres-
ically described as a minimization problem, as shown sures (zero) and (c1 , φ1 , c2 , φ2 ) values shown in the
below. figure are given, the critical slip surface with a min-
imum factor of safety of F0 = 1.06, as shown in the
Minimize DIS(λ1 , λ2 , λ3 ) = | yi | (8) figure, was located using the Baker’s (1980) dynamic
programming search. It is seen that the slip surface
Subject to inequality constraints: passes through the two soil layers that have totally
0 ≤ λ1 ≤ λ1max , 0.0 ≤ λ2 ≤ λ2max , different strength parameter values.
In order to demonstrate the effectiveness of the pro-
0.0 ≤ λ3 ≤ λ3max posed method, the slope profile in Figure 5 is now
treated as a post-failure problem in which the (criti-
where DIS (λ1 , λ2 , λ3 )—the objective function, and cal) failure surface and its factor of safety (F0 = 1.06)
λ1max , λ2max , λ3max are estimated maximum values of are known but the strength parameters of the two soil
(λ1 , λ2 , λ3 ), respectively. layers are assumed to be unknown. Thus, the pro-
The minimization problem shown in Equation (8) posed method can be used to back-calculate (c1 , φ1 ,
can generally be solved by a nonlinear mathematical c2 , φ2 , ) values based on the information provided by
programming method. The SUMT technique called the ‘‘failure’’ surface in Figure 5, and the results can be
the interior point method (Jacobi et al., 1972) is used in compared with their known (correct) values to verify
this paper to transform DIS (λ1 , λ2 , λ3 ) into a modified the accuracy of the method.
objective function without the constraints. Then, the Numerous tests performed herein have illustrated
Nelder and Mead simplex method (1964) is employed that the simplex technique provides a systematic and
to obtain an optimal solution of (λ1 , λ2 , λ3 ) that gives efficient tool to solve the minimization problem as
the minimum value of the objective function. The shown in Equation (7). The convergence is reached
details of the solution procedure are omitted due to rapidly in all the cases considered. However, the sim-
the limitation of space. plex iterations may converge to a local minimum when
A critical slip surface search is needed for each
of the sets of (λ1 , λ2 , λ3 ) in performing the simplex
(m)
iterations. In this study, we used the search method 0 5 10 15 20 25 30 35 40 45 50 55 60
by Baker (1980) which integrated the Spencer slope 0
critical slip surf ace
stability analysis (1967) with dynamic programming 5
(minimum factor o f safety =1 .06)
to locate the critical slip surface for each set of
(λ1 , λ2 , λ3 ). 10

When the optimum solution denoting by (λ10 , λ20 , 15 layer 1 layer 2

λ30 ) is obtained, all the four parameters (c1 , φ1 , c2 , φ2 ) (m) c1′ , tan 1′ c′2 , tan 2′
20
can be uniquely determined by considering the fact
of F = F0 . This has been done using the following 25 c1′ = 19.6 kPa , 1′ = 10.0°
procedure. c′2 = 4.9 kPa , ′2 = 20.0°
30
=17.64 kN / m 3 , = 19.6kN / m 3
1. Specify a value for one of (c1 , φ1 , c2 , φ2 )
appro- 35
1 2

priately and calculate the values of the other three

parameters by substituting (λ10 , λ20 , λ30 ) into the Figure 5. Hypothetical failure in a two-layered slope.

750
Table 1. Results back analyzed from hypothetical failure (Correct solution: λ1 = 111.16 kPa, λ2 = 13.46 kPa, λ3 = 4.0).

Initial values Back calculated values Objective function


No λ1 (kPa) λ2 (kPa) λ3 λ1 (kPa) λ2 (kPa) λ3 | yi | (m)

1 98.0 9.8 10.0 109.16 14.71 0.0003

2 9.8 9.8 10.0 18.81 27.13 28.167
3 45.0 180.0 5.0 102.49 311.45 13.23 30.667
4 180.0 9.8 4.0 164.71 19.98 4.55 21.367

the objective function is multimodal (Nguyen, 1985).

That is, the optimum solution is dependent upon the
selection of initial values of (λ1 , λ2 , λ3 ). This is also
true for the present problem. A number of solution
sequences are therefore needed to obtain an overall
optimum, each starting from different initial values of
(λ1 , λ2 , λ3 ). By running the proposed method with dif-
ferent initial values for (λ1 , λ2 , λ3 ), the sequences will
yield DIS values associated with different (λ1 , λ2 , λ3 )
combinations as potential solutions. The smallest one
of these DIS values gives a correct (optimal) solution
and the corresponding set of (λ1 , λ2 , λ3 ) is therefore
taken as the required values of (λ10 , λ20 , λ30 ).
Several different initial values for (λ1 , λ2 , λ3 )
Figure 6. Plan view of moving sliding blocks.
were randomly selected for the problem shown in
Figure 5, and for each case the back analysis procedure
described in Section 3 was executed. Table 1 shows Sliding block 2
the obtained values of (λ1 , λ2 , λ3 ) and DIS as well Sliding block 1
as the associated initial values for typical four cases.
Results in Table 1 indicate that the best solution is
obviously given by the (λ1 , λ2 , λ3 ) values for the No.1
case which resulted in the smallest difference (lowest Colluvial
Colluvial
Observed deposit
Observed
DIS value) between the theoretical critical slip surface sliding
sliding
surface
surface
and the actual failure plane. Using λ10 = 109.16 kPa,
λ20 = 14.71 kPa, and λ30 = 3.65, four strength param-
eters are calculated for F0 = 1.06 and the results were
found to be c10 = 19.40 kPa, φ10 
= 10.1◦ , c20 = Shale
Shale Serpentinite
Serpentinite
 ◦
5.29 kPa, and φ20 = 19.9 . These solutions agree well Serpentinite
Conglomerate
Conglomerate
0 55 10101515
20m
20m
with their correct values.

Figure 7. Geological cross section of slope and sliding

5 BACK ANALYSIS OF A FAILED SLOPE blocks.
IN TWO-LAYER SOILS

The proposed method is herein used to back-analyze The strong precipitation for the storm was centered
a rainfall-induced landslide in a two-layer soil slope. on a very narrow area of 5 to 6 km in the east-west
Heavy rain fell on the Shikoku area during Typhoon direction, and 10 to 20 km in the south-north direction.
Namtheun (the 10th tropical storm in the western Within the area, many landslides and debris flows were
Pacific in 2004). The total precipitation from July 30 triggered during the first two days of August (Wang
to August 2 was more than 2,000 mm in Kisawa village et al. 2005). One of the rainfall-induced landslides
and Kaminaka town, Tokushima Prefecture, Japan. was a landslide in Shiraishi town. This slide occurred
This is several times the normal precipitation for the near the top of the Furon valley in a small residential
months of July and August in this area. Hourly precipi- area in the town. The sliding mass from the cover-
tation reached more than 120 mm and the highest daily ing deposits overlying on weathered bedrock moved
precipitation of 1,317 mm was recorded on 1 August. downslope and transformed into a large debris flow,

751
 0 As the sliding block 1 moved slowly, its factor of
Water table safety was estimated to be 1.0. A number of combina-
5 tions of (λ1 , λ2 , λ3 ) are appropriately assumed as initial
Slope surface Layer 1 values, and then the proposed method was run for
10
each of the combinations. The back calculated values
of (λ1 , λ2 , λ3 ) and the magnitude of the correspond-
y (m)

15
Layer 2 ing objective functions are summarized in Table 2 for
20
Estimated slip surface typical four cases.
It is seen from Table 2 that when λ1 = 14.3 kPa,
25
λ2 = 10.1 kPa, and λ3 = 1.2, the difference between
30 the theoretical critical slip surface and the failure plane
0 5 10 15 20
x (m)
25 30 35 40 45 reaches a minimum of DIS = 5.6 m. Thus, these val-
ues of (λ1 , λ2 , λ3 ) are taken as the correct solution.
Figure 8. Cross section and failure surface of sliding The four strength parameters are calculated to meet
block 1. the condition of F0 = 1.06 and the relationships of (c1 /
tanφ1 = 14.3 kPa, c2 /tanφ2 = 10.1 kPa, c1 /c2 = 1.2).
As a result, c1 = 9.7 kPa, φ1 = 34.2◦ , c2 = 8.3 kPa,
traveling about 800 m. The debris flow destroyed a and φ2 = 37.8◦ are obtained.
number of houses which were built on the right side
and near the exit of the valley.
Around the top of the valley, two unstable blocks
with many cracks were found to move slightly after 6 CONCLUSIONS
the landslide. The ground deformation observed by
the installed extensometers indicates that the deform- It has been shown that when the soil unit weight and
ing landslides show nearly translational movement. pore water pressure distribution in a two-layer slope are
Figures 6 and 7 show a plan view and a representa- given, the location of the critical slip surface from a
tive geological cross section of the landslides. At this limit equilibrium stability analysis will remain unique
site, the slope is underlain by the weathered bedrock for a particular set of (λ1 = c1 / tanφ1 , λ2 = c2 / tanφ2 , λ3 =
that is mainly composed of a serpentinite. The cover- c1 /c2 ) values regardless of the magnitude of individual
ing materials overlying on the bedrock are colluvial strength parameters (c1 , φ1 , c2 , φ2 ). This theoretical
deposits. Field investigation showed that the collu- relationship between the (c , φ  ) and the critical slip
vial deposits are unconsolidated mixtures of clay and surface is found directly from the definition of the
weathered fragments of sandstone and claystone. The factor of safety and thus is available for any existing
serpentinite is commonly fissured, crumbling easily. limit equilibrium slope stability method.
Bore-hole data and inclinometer observations indi- Based on the findings of the relationship between
cate that the failure surfaces of the both unstable blocks (c , φ  ) and the critical slip surface, a straightforward
cut through the covering deposits and the serpenti- back analysis method has been presented to estimate
nite. As a very large portion of the sliding block 2 the magnitude of (λ1 , λ2 , λ3 ) from a known position
is located in the underlying bedrock, the sliding of the failure surface observed in the field slope. In
block 1 is chosen and used to back calculate the the proposed method, back analysis of (λ1 , λ2 , λ3 ) is
strength parameters of the two different materials. described as a problem to minimize the difference
Figure 8 shows a cross section of the failure sur- between the actual failure plane and theoretical critical
face of sliding block 1. The water table in the slip surface. Although the optimal solution is affected
figure was determined after the debris flow event by initial values, it is possible to obtain a correct com-
using the highest groundwater level observed in bore- bination of (λ1 , λ2 , λ3 ) using the solution sequences
holes. suggested in this paper. By running the method with

Table 2. Results back-calculated from the slope failure shown in Figure 8.

Initial values Back calculated values Objective function


No λ1 (kPa) λ2 (kPa) λ3 λ1 (kPa) λ2 (kPa) λ3 | y1 |(m)

1 10.0 10.0 1.0 14.3 10.1 1.2 5.6

2 30.0 120.0 1.0 29.0 47.0 2.7 56.0
3 150.0 40.0 10.0 109.3 74.3 11.4 61.5
4 500.0 250.0 1.5 391.4 253.2 2.4 64.5

752
different initial values for (λ1 , λ2 , λ3 ), several DIS val- Jacoby, S.L.S., Kowalik, J.S. & Pizzo J.T. 1972. Iterative
ues can be obtained as potential solutions. The smallest methods for nonlinear optimization problems. Prentice-
one among the DIS values gives the correct solution of Hall.
(λ1 , λ2 , λ3 ) that corresponds to the actual failure sur- Japan Road Association, 1999. Guidelines for slope stabi-
face. When the magnitude of (λ1 , λ2 , λ3 ) is obtained, a lization works and slope stability. (in Japanese).
Jiang, J.-C. & Yamagami, T. 2006. Charts for estimating
unique set of (c1 , φ1 , c2 , φ2 ) can easily be determined strength parameters from slips in homogeneous slopes.
by considering the fact that the factor of safety of the Computer and Geotechnics, 33, 294–304.
actual failure surface is equal to unity. Li, T.D. & Zhao, Z.S. 1984. A method of back analysis
Results for two examples have been presented to of the shear strength parameters for the first time slide
demonstrate the proposed method. For the shear failure of the slope of fissured clay, Proc., 4th Inter. Symp. on
in a hypothetical slope, back calculated (c1 , φ1 , c2 , φ2 ) Landslides, Toronto, 2, 127–129.
values of two soil layers agreed well with their cor- Michalowski, R.L. 2002. Stability charts for uniform slopes.
rect solution. The back analysis of a rainfall-induced ASCE J. Geotech. Geoenvir. Engrg., 128 (4), 351–355.
landslide resulted in a set of (c1 , φ1 , c2 , φ2 ) that prob- Nelder, J.A. and Mead, R. 1964. A simplex method for
function minimization. Computer Journal, 7, 308–313.
ably reflected the actual situation of the failure and Nguyen, V.U. 1985. Determination of critical slope failure
could be used for the remedial work design of the surfaces. ASCE J. Geotech. Engrg., 111, 238–249.
slope. These results show the potential of the proposed Saito, M. 1980. Reverse calculation method to obtain c and
method for practical use. Future research planned is φ on a slip surface. Proc. 3rd Inter. Symp. on Landslides,
to carry out laboratory failure tests for model slopes New Delhi, 1, 281–284.
to further verify the accuracy of the proposed back Spencer, E. 1967. A method of analysis of the stabil-
analysis method. ity of embankments assuming parallel inter-slice forces.
Géotechnique, 17 (1), 11–26.
Wang, G.H., Suemine, A., Furuya, G., Kaibori, M. & Kyoji
Sassa. 2005. Rainstorm-induced landslides at Kisawa
REFERENCES village, Tokushima Prefecture, Japan, August 2004,
Landslides, 2, 235–242.
Baker, R. 1980. Determination of the critical slip surface Wesley, L.D. & Leelaratnam, V. 2001. Shear strength param-
in slope stability computations. Int. J. Numer. and Anal. eters from back-analysis of single slips. Géotechnique,
Meth. in Geomech., 4, 333–359. 51 (4), 373–374.
Duncan, J.M. & Stark, T.D. 1992. Soil strengths from back Yamagami, T. & Ueta, Y. 1989. Back analysis of average
analysis of slope failures. Proc. Specialty Conf. Stability strength parameters for critical slip surfaces. Proc. Com-
and Performance of Slopes and Embankments-II, ASCE, puter and Physical Modelling in Geotech. Engrg. (eds.
Berkeley, CA, 1, 890–904. A. S. Balasubramaniam, et al.), Balkema: Rotterdam,
Greco, V.R. 1996. Back-analysis procedure for failed slopes. 53–67.
Proc. 7th Int. Symp. on Landslides. Balkema, 1, 435–440.

753
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Development characteristics and mechanism of the Lianhua Temple

Landslide, Huaxian County, China

Jia-yun Wang, Mao-sheng Zhang, Chuan-yao Sun & Zhang Rui

Xi’an Institute of Geology and Mineral Resources, Xi’an, China

ABSTRACT: The Lianhua Temple landslide is located in Huaxian county of Weinan City, Shaanxi province,
China. On the basis of the analysis of its development characteristics and mechanism of the landslide, it is
concluded that the destruction mode is a progressive process of sliding-tension along the joints and gneissosities,
and that the contribution factors to the landslide include triggering and controlling factors.

1 INTRODUCTION to the Mihu valley on the east, and connected with the
Baila valley on the west. At the foot of the mountains
The Lianhua Temple landslide is located in Huax- are the alluvial and prolvial fans of the Baila valley
ian county of Weinan City, Shaanxi Province, and and the Mihu valley, of which the terrain is flat and
becomes one of the famous rock landslides with high wide, while in the landslide deposit zone the terrain is
speed and long slip distance because of its complicated uneven and disorderly.
geological background, the unique development char-
acteristics, the perplexing mechanism and the long slip
distance etc. Previous historical documents and sur- 2.2 Formation lithology
veys on the landslide show that the Lianhua Temple
landslide was not the so-called avalanche triggered by The parent rock of the landslide is gneiss in the archean
the intensive earthquake in Huaxian county in 1556, era. Well-developed gneissosities down the slope are
instead, it is a super huge rock landslide with long slip favorable to the occurrence of landslide. In the fault
distance and high speed which happened before this zones, rocks are fragments consisted of crushed rock
earthquake. Based on the engineering geological map- and tectonite. Due to poor shear resistance, the fault
ping, pitting and trenching , the authors expounded the zone is the favorable site for shear failure of the slope.
terrain and physiognomy, the formation lithology, the The hill at the foot of the mountains is covered by
characteristics of the tectonic, the rock mass discon- five to six meters loess. Paleosol develops along the
tinuity, the shape of the landslide, and the landslide hill, and the vertical joints and the joints along the hill
bed, and discussed the characteristics of the landslide are well developed. During the accumulation, loess
in the physical structure and the movement. In the end, suffered so much dislocation that about 1 meter loess
the authors analyzed the mechanism of the landslide, at the bottom of the loess slope has been immingled
including the deformation and destruction mode, as with the crushed rock, and that the arc interface, which
well as the triggering and controlling factors. is formed by the shearing of loess to the crushed rock,
can be seen obviously.

2 THE GEOLOGICAL CONDITIONS

OF THE LANDSLIDE 2.3 Tectonic characteristics
The Lianhua Temple landslide is located in the Baoji-
2.1 The terrain and physiognomy
Tongguan fault zones, consisted of East-West or
The Lianhua Temple landslide, Huaxian county lies in NWW complex folds and faults. Faults experienced
the north of the Qinling mountains, which are located several tectonic movements and their types trans-
in the steep transition belt that crosses the Qinling formed many times. In the early stage, their types were
mountains and the Guanzhong basin. The elevation compression or compression-shearing faults. Later,
ranges from 300 m to 2500 m, so the hill has large rel- strong tectonic movements transformed the faults into
ative elevation difference and steep terrain. The profile tensional or shearing faults (Xu Ren-chao et al. 1987)
of the valleys is in ‘‘V’’ shape. The landslide is adjacent Due to multiple tectonic movements, rocks are crushed

755
strongly and tectonic joints down the slope develop, million cubic meter. The head of the landslide is 1300
which provide potential slip plane for the landslide. m above the sea level, and the shear-outlet is 550 m
above the sea level. So the relative height difference
between them is 750 m, and the one from the shear-
3 DEVELOPMENT CHARACTERISTICS outlet to the foot of the hill is 120 m. Therefore, there
OF THE LANDSLIDE is high potential energy existing caused by the obvious
relative height difference. After landing, the slip mass
3.1 The shape characteristics slides along the direction of 350◦ .
The shape of landslide is like a fan, which spreads out
3.2 The characteristics of rock mass discontinuity
from the rear to the middle and the front part (Figure 1).
The slip mass is 5 km long from the south to the north, The parent rock of the landslide is gneiss with well-
which is adjacent to the Mihu valley on the east and developed gneissosities. The occurrence of the gneis-
adjacent to the Baila valley on the west. The rear of sosities is 355◦ ∠34◦ , which is almost parallel with
the landslide is 1 to 1.5 km wide, the middle is 1.5 to the hill. And the rock mass discontinuity causes
1.8 km wide, and the front part is 1.8 to 3.25 km wide. anisotropy and reduces the strength of the rocks. The
It covers an area of 7.5 km2 , and its volume is 300 faults experienced several tectonic movements and
their types were transformed many times from the
compression or compression-shearing faults in the
early stage to tension or shear faults in the late stage.
Due to multiple tectonic movements and transforma-
tion of faults types, tectonic joints of the rocks are
well developed down the hill. The occurrence of the
joints in the right sidewall of the landslide is 335◦ ∠40◦ ,
and the one in the left sidewall is 0◦ ∠40◦ . The well-
developed gneissosities and tectonic joints along the
hill reduce the stability of the slope, in the mean time,
provide potential slip plane.

3.3 The characteristics of the landslide bed

When the rock mass slides out of the landslide bed
with high speed, only small part of slip mass is left
on the landslide bed. So most of the landslide bed is
bare, which is 2500 m long and 500 m wide and can be
seen obviously. The landslide bed with curving shape
is narrow in the rear and wide in the middle. It con-
stringes at the shear-outlet. The shape of the landslide
bed is like fold line, and the average gradient of slope
is 20◦ (Figure 2). The sidewall of the landslide bed
is gentle in the west and steep in the east, of which
the slope angel is greater than 70◦ in the east and is
less than 60◦ in the west. And in the west, huge rocks
scatter on the surface. There are large amount of mix-
tures of rocks and loess on the landslide bed. In the
front, the diameter of the rock blocks is 5–50 cm with
certain psephicity. And in the rear part, the diameter
is over 1 m with poor psephicity. The slip direction of
the slip mass changes several times from 320◦ to 355◦
and then to 310◦ , and the scrape on the right sidewall
of the landslide can prove this.

Figure 1. The shape of the Lianhua Temple land- 3.4 The physical structure characteristics
slide. 1—the perisporium of the landslide; 2—the of the landslide
boundary of the slip mass; 3—the contour line and alti-
tude; 4—the zoning boundary of the slip mass accumulation; The region of the slip mass accumulation can be
5—railway; 6—highway; 7—channel; 8— residential area. divided into three zones: chipping flow zone, mass

756
 3.5 The movement characteristics of the landslide
The large difference between the head and the shear-
outlet of the landslide is 750 m. The great height
difference causes high potential energy, which can
be transformed into kinetic energy for landslide in
high speed. During sliding nearly 2500 m along the
landslide bed, the rocks impact the landslide bed con-
tinuously; at the same time, the potential energy is
being converted into kinetic energy. So the accelerat-
ing effects like the elastic impulse a near the landslide
bed and drop of peak-residual strength (Hu, 1988)
appear. The sliding speed of rocks is accelerated and
the rocks get high speed when they slide away from
the shear-outlet. Because of the alluvial and proluvial
deposits at the foot of the mountains, the tremendous
Figure 2. The profile of the landslide bed. 1—the slope impact makes the pore-water pressure in the deposits
deposit in Holocene; 2—the landslide deposit in Holocene; increase sharply and makes the deposits fluidified
3—Taihua group in archean; 4—mixture of rock blocks and when the slip mass gets to the ground. So the resis-
loess; 5—rock block; 6—gneiss; 7—the landslide bed and tance decreases and the slip mass can keep sliding in
slip direction; 8— fault and tectonite; 9—the original terrain high speed. The friction caused by the slide in high
line inferred.
speed makes the pore-water vapored and the air cush-
ion effect forms on the interface between the slip mass
and the alluvial and proluvial deposits, which keeps
slide zone and rock scattering zone. The rock scat-
the slip mass slide in high speed. Because of the above
tering zone is below the shear-outlet. The surface is
reasons, the mass slide body can slide 2.5 km away
concave which was the Baiya lake, and it is covered
from the shear-outlet. Due to wide and gentle terrain
by rock blocks with average diameter of more than
of the alluvial and proluvial fans and due to the effects
1 m and the largest diameter of 10 m. Below it is the
of inertia and the pore-water, the mixtures of loess,
0.5 to 1 m mixture of loess and rock blocks and the
crushed rocks and chippings flow down along the gen-
loess has the evidence of extrusion. The mass slide
tle slope, and the slip distance gets to more than 2 km.
zone is covered by 3–5 m loess. The waved pale-
The longest slip distance of the slip mass is 5 km away.
osol are visible, though the loess is disturbed. The
From the above analysis it can be seen that the landslide
middle part of the mass slide body is the mixture of
has the movement characteristics of high speed, long
rock blocks and loess with several meters to more
slip distance, the air cushion effect and the chipping
than tens meters thick. The rock blocks in the mix-
flow effect.
ture have clear angularities and no psephicity. And
the diameters of the rock blocks range from several
centimeters to several meters. Loess in the mixture is
4 THE MECHANISM OF THE LANDSLIDE
brick red and compact, and it has signs sintered. In
partial zones it can be seen that the rock blocks and
4.1 The mode of landslide destruction
chippings are turned on the loess. Below them is the
gneiss. The gneiss is still continuous in general, and Well-developed gneissosities in parent rocks and over-
even the gneissosities can be seen, although it experi- developed tectonic joints for multiple tectonic move-
ences strong disturbance. White belt of materials in the ments provide the potential slip plane. The occurrence
gneiss is sinter formed in the high temperature environ- of landslide bed is close to those gneissosities and
ment caused by slide friction. In the mass slide zone, tectonic joints, which indicates that the landslide is
the bottom boundary can not be seen clearly despite of caused by deformation and destruction of sliding-
the largest exposure thickness of over 30 m. The chip- tension of the gneissosities and tectonic joints for
ping flow zone is located in the gentle terrain to the gravity. The historical records of Song dynasty indi-
north of the Yishan temple. Rock blocks in this zone is cated that the inhabitants in the mountains said that
smaller and the diameters of the rock blocks are from there were always clouds on the peak in a few years
several centimeters to more than tens centimeters, and the rocks sent out sounds when it rained. This is
which are immingled with the loess and chippings. The the evidence that the deformation and destruction of
mixtures of loess, chippings and rock blocks move for- the landslide is the progressive process. From the
ward several kilometers due to the Inertia and ground above evidence, it is proved that the deformation
water (Figure 3), and the chipping flow zone comes and destruction is the progressive process of sliding-
into being. tension of the gneissosities and tectonic joints.

757
Figure 3. The profile of the slip mass in major slip direction. 1—tectonite; 2—mixture of rock block, chipping and loess;
3— loess; 4—gneiss; 5—proluvial deposit in late pleistocene; 6— the fault; 7—present terrain; 8—the original terrain inferred.

4.2 The triggering factors structures of the slope and the stress was adjusted sev-
eral times. The stability of the slope decreased and the
4.2.1 Rainfall
possibility of the destruction increased along with it.
The historical records indicate that it always rained
and the rocks in the hill sent out sounds a few years
before occurrence of the landslide. So the rainfall is
the active factor that triggers the happening of the land- 4.3 The controlling factors
slide. It includes two aspects: one is the decrease of 4.3.1 Great relative height difference
rocks strength. The rainfall makes the water content and steep terrain
of the rocks increase, even saturated. For the materials The altitude of the landslide zone ascends sharply from
exchange of water and rocks, the mechanical pro- 300 m in the northern Fen-Wei basin to over 2 km in the
perties of the rocks are changed, especially the shear north of Qinling mountains and the terrain is steep. The
resistance is decreased. The decrease of rocks strength great relative height difference (750 m) from the head
causes the decrease of the stability of the slope. The to the shear-outlet provides enough potential energy
other aspect is the effect of the pore-water pressure. for the slip in high speed.
The rainfall causes the joint network saturated and the
pore-water pressure increase, so that the horizontal
thrust and the buoyancy force to rocks increase. The 4.3.2 Shear-outlet in crushed fault belt with high
horizontal thrust causes the increase of the pushing free surface
force of rocks, and at the same time, the buoyancy The shear-outlet of the landslide lies in the fault belt
force reduces the sliding friction force, so the increase and the lithology is made of crushed rocks with infe-
of pore-water pressure reduces the stability of the rior cementation and low strength. So the shear-outlet
slope. zone becomes prone to destruction. The relative height
difference of the free surface is 120 m, which provides
4.2.2 The effect of early earthquakes favorable condition for slip of the slope.
The Lianhua Temple landslide is located in the down-
faulted belt and earthquake belt of the Fen-Wei basin,
and the borders of the basin are controlled by the 4.3.3 Well-developed gneissosities and tectonic
normal faults with intense movements. In this area joints along the slope
the tectonic movements of faults cause earthquakes, The gneissosities is well developed in parent rocks, and
so the earthquakes epicenter always distributes along the discontinuity of the gneissosities causes anisotropy
the faults belt. The historical records shows that the and reduces the strength. Multiple tectonic move-
Lianhua Temple landslide happened in 1072 before ments cause the tectonic joints overdeveloped along
the Huaxian county earthquake (1556), so the land- the slope, which aggravates the anisotropy of rocks
slide has no relationship with this earthquake. While and reduces the stability of the slope. Well-developed
multiple earthquakes appeared in the eastern area gneissosities and tectonic joints along the slope not
of the Guanzhong Basin before the Lianhua Temple only reduce the stability of the slope, but also provide
landslide (He, 1990). The intense shake changed the the landslide with potential slip plane.

758
5 CONCLUSIONS landslide include great relative height difference
and steep terrain, shear-outlet in crushed fault
According to the surveys, analyses and studies of belt with high free surface, and well-developed
the Lianhua Temple landslide, Huaxian county, the gneissosities and tectonic joints along the slope.
following conclusions can be obtained: 5. Further surveys and studies of the Lianhua Tem-
ple landslide are needed to find out the dynamic
1. The Lianhua Temple landslide has the physical
mechanisms.
structure characteristics of chipping flow, mass
slide and rock scattering.
2. The Lianhua Temple landslide has the movement
REFERENCES
characteristics such as high speed, long slip dis-
tance, the air cushion effect and the chipping flow Xu Ren-chao, Zhong Li-xun, Pan Bie-tong & HU Guang-tao.
effect. 1987. The Stability Research of Slopes in the Southern
3. The deformation and destruction of the Lianhua Mountainous Area of Shaanxi Province.
Temple landslide is the progressive process of slid- Hu Guang-tao. 1988. Dynamical Landslideology. Xi’an:
ing—tension along the gneissosities and tectonic Shaanxi Science and Technology Press.
joints. He Ming-jing. 1990. The Hazards Research of the Earth-
4. The triggering factors of the landslide include quakes in Huaxian County. Xi’an: Shaanxi People Edu-
the effect of early earthquakes and the rainfall, cation Press.
which reduces the strength of rocks and increases Cheng Qian-gong & Peng Jian-bing et al. 1999. Dynamics of
Rock Landslide with High Speed. Chengdu: the Southwest
the pore-water pressure to cause horizontal thrust Communication University Press.
and buoyancy force; the controlling factors of the

759
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Modeling landslide triggering in layered soils

R. Keersmaekers
Department of Architecture, Provincial University College Limburg, Hasselt, Belgium
Department of Civil Engineering, Building Materials Division, Katholieke Universiteit Leuven, Leuven, Belgium

J. Maertens & D. Van Gemert

Department of Civil Engineering, Building Materials Division, Katholieke Universiteit Leuven, Leuven, Belgium

K. Haelterman
The Authorities of Flanders, Geotechnics Division, Ghent, Belgium

ABSTRACT: The Flemish Government (represented by AMINAL; Administration of Environment, Nature,

Land and Water management) commissioned a research project to study the triggering of landslides in the
Flemish Ardennes. A few representative sites subject to landslides were studied from a geotechnical point of
view. Several sites were selected for geotechnical calculations in order to predict the conditions necessary to
trigger a landslide and to verify the predictions with the observations on site. Two of those sites are discussed
in this paper. A hypothetical collapse mechanism, possibly responsible for many landslides in the Flemish
Ardennes, was numerically verified. The presence of a sand layer (high water permeability) between two clay
layers (low water permeability) causes the building up of pore water overpressures, decreasing the effective
stresses, eventually resulting in the uplift and/or collapse of the slope. The understanding of the mechanisms
responsible for the studied landslides resulted in specific recommendations to prevent future landslides on these
sites, but also on similar sites in the region.

1 INTRODUCTION Another important parameter is the piezometric

height of the water in the different soil layers, or the
1.1 Landslides in layered soils: causes ground water level. The shear resistance is determined
and parameters (in drained circumstances) by the effective stresses
Insufficient safety of the global stability of a slope in the soil, which are directly correlated with the
can cause a landslide that occurs along a slip surface piezometric heights in the soil layers. When dealing
(circular or not), or a slow displacement (creep) of the with permeable layers with significant thickness, the
entire slope. Landslides can also be triggered by the groundwater level or the piezometric height is easily
seepage of ground water out of the slope, causing local measured using water level tubes. The determination
erosion and caving. of correct piezometric heights becomes much more
An insufficient safety of the global stability of the difficult when thin permeable layers occur within lit-
slope may be induced by several parameters. The tle permeable or impermeable layers. The variation
geometry of the slope is obviously a very important of piezometrc height within an enclosed layer can be
factor. A steeper slope will collapse more rapidly. very high (i.e. during periods with heavy rainfall). In
The shear strength characteristics of the involved Figure 1 a hypothetical collapse mechanism caused by
soil layers play an even more important role. The high piezometric levels is illustrated.
friction angle ϕ[◦ ] and cohesion c [MPa] (cfr. The Seepage of rainwater through the top layer into the
Mohr-Coulomb soil model) must be determined for sand layer will cause high water overpressures in this
every layer when evaluating the global stability of the permeable layer when the entrance water amount (e.g.
slope. When checking the safety of a slope which true cracks) is bigger than the exit water amount (which
already experienced a landslide, it is best to evalu- is very limited because of the inclusion between two
ate the stability also with the residual shear resistance impermeable layers). This high water pressures result
parameters (De Beer, (1979)). in low effective stresses (= total stress minus water

761
 analytical (SLOPE, www.geo-slope.com) mathemati-
cal algorithms.
With these models it is possible to calculate an over-
all factor of safety to evaluate the stability of the slopes.
An hypothetical collapse mechanism as defined in
paragraph 1.1 (See Figure 1), possibly responsible
for many landslides in the Flemish Ardennes, will
be numerically verified. The presence of a sand layer
(high water permeability) between two clay layers (low
water permeability) causes the building up of pore
Figure 1. Typical failure mechanism. A permeable (sand) water overpressures, decreasing the effective stresses,
layer is enclosed by little permeable layers (e.g. loam or clay) eventually resulting in the collapse of the slope.
resulting in high water overpressures and thus low effective Numerical simulations also proved to be very useful
stresses in the sand layer, resulting in collapse during heavy to determine the negative influence on the overall slope
rainfall. stability of adding additional loads to the upper slope
surface (buildings), the slightly positive influence of
vegetation and the negative influence of excavations
pressure) and thus lowers the shear resistance, causing (swimming pools, ponds) at the bottom of the slope.
the landslide to occur. The understanding of the mechanisms responsible
The global stability can also be affected when the for the studied landslides results in specific rec-
load on the slope is altered. This can have a stabilizing ommendations to prevent future landslides on these
or a destabilizing effect. When adding load (buildings, sites, but also on similar sites in the region. For a
soil deposits, . . .) at the top of the slope, this will have more detailed publication in Dutch of this work see
a destabilizing effect. Adding load at the toe of the (Keersmaekers et al. (2005)).
slope will have a stabilizing effect.

2 MODELLING
1.2 Research aim
The Flemish Government (represented by AMINAL; 2.1 Analytical method: SLOPE (Bishop)
Administration of Environment, Nature, Land and
The development of calculation methods to check the
Water management) commissioned a research project
global stability of slopes with arbitrary shapes and
to study the triggering of landslides in the Flemish
materials with cohesion and friction goes back for
Ardennes. This particular rolling region in Belgium
decades. Firstly only circular slide surfaces were con-
has a history of landslides and general stability prob-
sidered (Figure 2). Later on calculation methods were
lems of slopes. The project marked out a part of
developed for irregular slide surfaces. Recently, the
the Flemish Ardennes, then produced an inventory,
use of finite element methods is becoming more and
a classification, a statistical and spatial analysis and a
more common.
methodology for the production of hazard maps (Van
The most used analytical method to evaluate circu-
Den Eeckhaut et al. (2005)). Complementary, a few
lar slip surfaces is the Bishop method. The soil above
representative sites subject to landslides were studied
from a geotechnical point of view. This geotechnical
study is the topic of this paper.
Three sites were selected for geotechnical calcu-
lations in order to predict the conditions necessary to
trigger a landslide and to verify the predictions with the
observations on site. Two of those sites are discussed
in this paper. The study of each site started with the
execution of soil investigation tests (Cone Penetration
Tests, borings and triaxial shear tests on representa-
tive non disturbed soil samples) to draw a geotechnical
profile of the slope, determining the stratification of
layers and their geotechnical parameters. Combined
with topographical data of the collapsed slope and the
assumed profile before the landslide occurred, numer-
ical models of these sites were implemented using Figure 2. Circular slide surface for analytical calculations
two software programmes based on numerical (finite conform Bishop. This analytical method divides the slide
elements method using PLAXIS, www.plaxis.nl) and volume in vertical strips.

762
the slip surface is divided in vertical strips, bounded When evaluating the Slope and Plaxis calculations
by vertical surfaces. A shear stress t is mobilized along in this project, an overall SF of 1.3 is required to
the slip surface, which is supposed to be a factor have a save slope. SF-values lower than one indicate
SF (safety factor) smaller than the maximum possible slopes that will collapse under the given parame-
shear resistance. A 2D-geometry is assumed. ters. SF-values between 1 and 1.3 correspond to
Verifying the stability of the slope consist of the slopes with insufficient safety, but will not necessarily
expression of the momentum equilibrium to a center collapse.
point of the considered circular slip surface. The calcu-
lation of SF is done iteratively (starting with SF = 1)
and must be done for many center points (i.e. many
possible circular slip surfaces) to find the lowest value
3 SITE1: SOCCER FIELD ‘‘KORTE KEER’’
of the SF.
AT MAARKEDAL
The software package used in this research project
to calculate the Bishop method is Slope.
3.1 Situation
The first site is the football field Korte Keer at
2.2 Finite element method: PLAXIS
Maarkedal (Nukerke, Belgium). The site is located
Most calculations in this research project were made above a very large, deep and old landslide. The foot-
with Plaxis, a finite element based software package, ball field itself was built on a slope by rectifying the
especially developed for geotechnical applications. terrain by filling the site with a sandy embankment.
The volume is divided into small elements which are From a geotechnical point of view, this added material
numerically coupled to each other. The stress equi- has similar properties as the top layer of the original
librium and the soil deformation are described by a slope (see hereafter).
system of regular and partial differential equations Figure 3 shows the situation of the site Korte Keer.
which are solved numerically. In this way the soil The line defines the topographic profile used to model
stresses and deformations can be calculated. the geometry of the collapsed and original slope (see
The advantage of using finite elements is that the also Figure 4). The original profile of the football field
real behavior of the soil is better simulated and that had a steep inclination of 1/1.
the real stresses occurring in the soil are taken into
account. There is for example a clear division between
vertical and horizontal stresses which is not the case
in analytical methods.
To model the behavior of the soil a so called Mohr-
Coulomb model has been used. This model assumes a
complete elastic behavior until the shear stress in the
element equals the shear resistance. After this point
the soil behaves completely plastic.
For the determination of the safety factor (SF as
defined above), the so called phi, c (ϕ, c)-reduction
method is used. The shear resistance parameters ϕ 
and c are reduced in the same way until collapse of
the slope. This collapse is verified by the displacement
of one or more well chosen physical points.
Figure 3. Situation of the site Korte Keer. The line defines
2.3 Safety factor the topographic profile used to model the geometry of the
collapsed and original slope.
When slopes are designed with the analytical Bishop
method, a safety factor SF of minimum 1.3 is normally
required. The calculation is then based on the momen-
tum equilibrium of a circular slip surface.
The calculation of the safety factor using finite ele-
ment methods is based on the phi, c (ϕ, c)-reduction
method. Many calculations in the past learned that the
overall safety factor, obtained from analytical methods
like Bishop do not differ a lot from the safety fac-
tor obtained using the phi, c (ϕ, c)-reduction method Figure 4. Profile before and after the event of the landslide.
(finite elements). Notice the location of the soundings.

763
3.2 Characteristics of the soil layers 3.3 Results of the calculations
There were five CPT-tests made on site with a 200 kN 3.3.1 Original profile with inclination 1/1
apparatus, which probed to depths until 25 meters. Figure 6 gives the slip surface of the original slope. The
Also two borings and triaxial tests on undisturbed soil safety factor SF is calculated to be 0.561, meaning that
samples were performed. The information obtained the collapse of the site was inevitable. The inclination
from these tests are than used to determine the strati- 1/1 is too steep. The calculation is made with a deep
fication and characteristics of the different soil layers. water table, which proves the landslide was not due to
The locations of these CPT-tests are given in Figures 3 water overpressures.
and 4. Also a piezometric pipe was installed to monitor
the variation of the water table over one year.
From the on-site investigations it was concluded 3.3.2 Profile after landslide
that the site consists of two major layers. The top layer The measurements of the water level pipe showed a
is a well permeable sand layer which is partly con- maximum piezometric height of +73.37 TAW (TAW
structed from sand deposits to rectify the slope and is the reference level for Belgium) between September
the underlying original sandy material (quaternary ori- and October 2004.
gin). Both layers have very similar characteristics and The SF-value obtained was 1.218, meaning the
are therefore modeled as one top layer. Under this top slope will not collapse, but has an insufficient SF
layer, a clay layer (from tertiar origin, i.e. ‘‘Ieperiaan’’) according to literature (minimum SF = 1.3). When
is found with an almost horizontal orientation. Table 1 the piezometric height of the top layer increases to +76
summarizes the main characteristics of both layers. TAW (i.e. 2.63 m higher than in the above situation),
A Mohr-Coulomb soil model is used to define the the SF-value drops to 0.997, meaning the slope has
two layers in Plaxis. Two models are constructed. The just reached equilibrium, the landslide can occur at
first one with the original slope geometry and one with any moment.
the present profile. This second profile, representative The influence of the growth of plants on the site
for the situation after the landslide, aims to determine is incorporated by giving the first meter of the top
the present safety of the site (see Figure 5). soil layer a cohesion of 5 kN/m2 . The SF increases

Table 1. Ground characteristics of the different soil layers.

Top sand layer Bottom clay layer

Parameter Value Unit Value Unit

γdrv 18 kN/m3 19 kN/m3

γwet 19 kN/m3 19 kN/m3
E-mod 2E + 4 kN/m2 1E + 4 kN/m2 Figure 6. Slip surface of the original slope. The safety fac-
Poisson v 0.3 [-] 0.35 [-] tor SF is 0.561, meaning that the collapse of the site was
Cohesion c 0.0 kN/m2 25 kN/m2 inevitable.
ϕ 27.5 ◦ 23 ◦

Figure 7. Top: Piezometric heights at +73.37 TAW of both

layers. Notice the drop in water level towards the toe of the
Figure 5. Plaxis geometric models for the original profile slope. Bottom: Preferential slip surface, the safety factor SF
and after the landslide occurred. is 1.218.

764
from 0.997 to 1.074, which proves the positive effect
of vegetation.
Comparison with a Slope calculation (analytical
model) resulted in a SF-value of 1.08 (compared
to 0.997), which proves that the overall safety fac-
tor, obtained from analytical methods like the Bishop
method does not differ a lot from de safety factor
obtained using the phi, c (ϕ, c)-reduction method
Figure 9. Profile before and after the advent of the land-
(based on finite elements modeling). Also the slip sur-
slide. Notice the location of the soundings.
face showed a very similar sliding surface, showing a
rather shallow collapse of the slope.
Finally a calculation was made to rebuild the foot- Table 2. Ground characteristics of the different soil layers.
ball field without the danger of triggering a landslide.
The piezometric height of the top layer was chosen one Top Middle Bottom
meter below the surface (this can be done by placing clay sand clay
a drainage system under the new football terrain and
Parameter Value Value Value Unit
a drainage at the toe of the slope). A minimum incli-
nation of the slope of 12/4 resulted in a SF of 1.333, γdrv 18 17 19 kN/m3
reaching the minimum required SF-value of 1.3. γwet 18 20 19 kN/m3
E-mod 1E + 4 1.3E + 4 1E + 4 kN/m2
Poisson v 0.35 0.30 0.35 [-]
Cohesion c 10 1 20 kN/m2
ϕ 25 30 25 ◦
4 SITE2: SCHERPENBERG RONSE

4.1 Situation
The Scherpenberg is a complex landslide on a slope 4.2 Characteristics of the soil layers
without buildings. Therefore it is easily accessi-
ble for cone penetration test and chosen for this There were five CPT-tests made on site with a 200 kN
project. apparatus, which probed to depths until 25 meters.
Figure 8 shows the situation of the site Scher- Also two borings were made and triaxial shear tests on
penberg. The dotted line defines the topographic undisturbed soil samples were performed. The infor-
profile used to model the geometry of the original mation obtained from these tests are then used to
slope. The continuous line defines the topographic determine the stratification and characteristics of the
profile used to model the geometry of the collapsed different soil layers. The locations of these CPT-tests
slope. Notice the location of the CPT-tests (see are given in Figures 8 and 9.
Figure 9). From the on-site investigations it was concluded
that the site consists of three major layers. The mid-
dle layer is a well permeable sand layer which is
enclosed between two low permeable clay layers.
Table 2 summarizes the main characteristics of the
three layers.
A Mohr-Coulomb soil model is used to define the
three layers in Plaxis. Two models are constructed. The
first one with the original slope geometry and the sec-
ond one with the present profile. This second profile,
representative for the situation after the landslide, aims
to determine the present safety of the site (Figure 10).

4.3 Results of the calculations

4.3.1 Presumed original profile
Figure 8. Situation of the site Scherpenberg. The dotted
Figure 11 gives the slip surface of the presumed orig-
line is representative for the original slope (no landslide has inal slope. The piezometric height of +65 TAW (for
occurred there). The continuous line defines the topographic the middle and bottom layers, for the top clay layer
profile used to model the geometry of the collapsed slope. the piezometric height is assumed one meter under the
Notice the location of the soundings. ground surface) is the maximum value measured on

765
 Figure 12. Piezometric heights of the middle and bottom
layers at +68 TAW. Preferential slip surface, SF = 1.026.

Figure 10. Plaxis geometric models for the presumed

original profile and after the landslide occurred.

Figure 13. Plaxis geometric model including the load

representative for a building at the top of the slope.

Figure 11. Top: Piezometric heights of the middle and bot-

tom layers at +65 TAW. Bottom: Preferential slip surface at
the top of the slope, SF = 1.526.
Figure 14. Top: Piezometric heights of the middle and bot-
tom layers at +68 TAW. Bottom: The preferential slip surface
site between November 2004 and January 2005. The shifts from the top to the bottom of the slope, SF = 1.865.
safety factor SF is calculated to be 1.526, meaning
that the site is safe under these conditions. Notice the
preferential slip surface at the top of the slope. top clay layer the piezometric height is set one meter
Raising the piezometric level of the middle and under the ground surface.
bottom layer to +68 TAW results in a SF of 1.026 The safety factor SF is calculated to be 1.865, mean-
(meaning the slope has just reached equilibrium, the ing that the site is safe under these conditions (which
landslide can occur at any moment), and a different are unlikely to occur in reality). For lower water levels
preferential slip surface at the bottom of the slope is the slope gave even higher SF’s and preferential slip
obtained (Figure 12). This means that when a land- surfaces at the top of the slope. Notice the shift from
slide will occur, the slope will collapse according to a high to a low preferential slip surface in Figure 14.
this second slip surface. When the piezometric height for the middle and
When adding a load of 40 kN/m2 at the top of bottom layers is set at +71.5 TAW, SF drops to 1, 212,
the slope (representative for a building on a fill layer, resulting in a slip surface at the bottom of the slope
Figure 13), for the same conditions of Figure 11 (cfr. Figure 15). For a TAW level +72.5, SF = 0.952,
(+65 TAW, SF = 1.526), the SF drops to 1.264. In meaning the slope has just reached equilibrium, the
case the load is 60 kN/m2 , the SF drops to 1.155, landslide can occur at any moment.
proving the negative effect of load at the top of the Figure 15 shows the negative influence of a swim-
slope. ming pool at the bottom of the slope. The empty
swimming pool is lifted, SF drops from 1.212 to 1.025,
4.3.2 Profile after landslide respectively without and with swimming pool.
Figure 14 gives the slip surface of the slope after occur- Finally the influence of the decrease of the strength
rence of the landslide. The piezometric height for the characteristics due to an already occurred landslide
middle and bottom layers is set at +68 TAW, for the is taken in to account. Figure 16 shows the Plaxis

766
 shear strength characteristics for quaternary clay are
given: ϕr = 12.5◦ and cr = 5 kN/m2 . The piezo-
metric heights of the middle and bottom layers are set
at +71, 5 TAW. The SF is 1.037 (compared to 1.212
above) proving the negative influence on the present
stability of the already disturbed zones in the top layer.
This result was verified with Slope, using the Bishop
method (Figure 17), giving a value SFSlope = 1.094
(almost equal to 1.037 above).

Figure 15. Top: Influence of a swimming pool at the bottom 5 CONCLUSIONS

of the slope. Piezometric heights of the middle and bottom
layers at +71, 5 TAW. Bottom: Incremental displacement This study demonstrates that the presence of a sand
vectors: The empty swimming pool is lifted, SF drops from layer between two clay layers may cause the building
1.212 to 1.025, respectively without and with swimming pool.
up of pore water overpressures, decreasing the effec-
tive stresses, eventually resulting in the collapse of a
slope (cfr. Figure 1).
Numerical simulations proved very useful to deter-
mine the negative influence on the overall slope sta-
bility of adding additional loads to the upper slope
surface (buildings), the positive influence of vegeta-
tion and the negative influence of excavations (swim-
Figure 16. Plaxis geometric model including the presumed ming pools, ponds) at the bottom of the slope.
disturbed zone (based on the sounding profiles) of the former
landslide.
REFERENCES

De Beer, E. 1979. Historiek van het kanaal Lei-Ieper; Eigen-

schappen en gedragingen van Ieperiaanse klei. Uittreksel
uit het tijdschrift der openbare werken van België nrs
4, 5 en 6.
Keersmaekers, R., Maertens, J. & Van Gemert, D. 2005.
Verkennende studie met betrekking tot massabewegin-
gen in de Vlaamse Ardennen. Deel II: Geotechnisch
onderzoek van enkele representatieve sites onderhevig
aan massabewegingen. Rapport Laboratorium Reynt-
jens R/30232/04. Rapport in opdracht van de Vlaamse
Gemeenschap, AMINAL, afdeling Land.
Van Den Eeckhaut, M., Poesen, J., Verstraeten, G., Vanacker,
V., Moeyersons, J., Nyssen, J. & van Beek, L.P.H., 2005.
Figure 17. Slope calculation (Bishop method) of a geo- The effectiveness of hillshade maps and expert knowl-
metric model including the presumed disturbed zone of the edge in mapping old deep-seated landslides. International
former landslide, SFSlope = 1.094. journal on Geomorphology 67, p351–363.

model including the presumed disturbed zone (based

on the sounding profiles) due to the occurred land-
slide. In (De Beer, (1979)), values for the residual

767
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Numerical modeling of debris flow kinematics using discrete element

method combined with GIS

Hengxing Lan & C. Derek Martin

Dept. Civil & Environmental Engineering, University of Alberta, Edmonton, Alberta, Canada

C.H. Zhou
LREIS, Institute of Geographic Sciences and Natural Resources Research, CAS, Beijing, P.R. China

ABSTRACT: Numerical modeling of debris flow kinematics was performed using the integration of Discrete
Element Method and Geographic Information Systems (GIS). A discrete element code was used to investigate
the dynamic behavior of debris flows in mountainous terrain. The macroscopic dynamic behavior of the debris
flows was evaluated using a spatial averaging method implemented in GIS. The coupling of the discrete element
method with GIS technology was evaluated at a debris flow site along a railway corridor in the Rocky Mountains
of Western Canada.

1 INSTRUCTION particles, ground surface can also be examined. Dif-

ficulties exist using the discrete element method in
A debris flow represents a mixture of sediment par- interpreting the output at a macroscopic scale due
ticles of various sizes and water flowing down a to the large local divergence during the simulation
confined or unconfined channel (Hutter et al. 1995). (Richard et al 2004). A range of methods can be used
Its physical and mechanical behavior exhibits a distinct to extract the useful quantitative macroscopic infor-
discrete nature which is characterized by the internal mation for the debris flow kinetics such as spatial
interaction between solid particles and between solid averaging or Fourier transform.
and fluid in the mixture of the debris flow body. The In this paper, an integration of discrete element
solid components are characterized by a broad distri- code (PFC3D) and Geographical System Information
bution of grain sizes, from boulders to clay particles (GIS) is used for the numerical modeling of debris
(Pirulli M. & Giuseppe S. 2007). The characteristics flow kinematics. PFC3D is used to investigate the
of flow kinematics also vary widely depending on the microscopic dynamic behavior of debris flow in an
fluid condition and particle size which result in large irregular ground-surface channel. The macroscopic
challenges in numerical modeling. dynamic behavior of the debris flow is inspected using
Continuum mechanics using depth averaging meth- a spatial averaging method implemented in GIS. The
ods is the current principle approach for the dynamic integration methodology is applied to a debris flow site
modeling of debris flows (Chen & Lee 2000; Den- along a major railway near Klapperhorn Mountain in
linger & Iversion 2001; McDougall & Hungr 2005). the Canadian Rockies of Western Canada.
Physically, this approach has difficulties accounting
for the microscopic behaviors of debris flow materials
in terms of transportation kinematics and collision or 2 STUDY AREA
friction interaction. Such an approach can not lead to
the construction of a clear picture about the debris flow The study area is located on Klapperhorn Mountain
dynamics in an irregular channel (Zhu & Yu 2005). near the border between Alberta (AB) and British
The Discrete element method is recognized as an Columbia (BC), Canada (Fig.1). The site is within
effective approach to study the fundamental behaviour the Rocky Mountains Proterozoic Middle Miette
of granular materials (Cundall & Strack 1979). Com- Group, which includes feldspathic sandstone, gran-
pared with the continuum techniques, it has advan- ule and pebble conglomerate, siltstone and argillite
tages of allowing us to elucidate the internal properties (Mountjoy, 1980). Two rail-lines operated by CN
of debris flow such as internal stress distribution (Canadian National Railway), the Albreda and Rob-
(Zhu & Yu 2005). The microscopic quantities such son subdivisions run parallel to each other at the base
as velocities of particles and interaction force between of Klapperhorn Mountain. The 3 km long section of

769
 drainage basin dominates the terrain units. Rock-
falls and snow avalanches occurring frequently in this
region deposit material at the top of a colluvial cone
where most of the debris flows initiate. Debris trans-
port occurs in the unconfined or confined drainage
channel whose morphology is usually bedrock con-
trolled. Debris flows deposit debris on the colluvial
fans at the end of primary or secondary channels which
can be above or below the track level.

3 NUMERICAL SIMULATION OF DEBRIS

FLOW KINEMATICS

3.1 Method
Figure 1. Study area of Klapperhorn Mountain. Five debris The integration of discrete element code (PFC3D)
flow drainage basins are identified. Photo shows the overview and GIS functions was implemented to facilitate the
of study area. The Albreda (upper) and Robson (lower) rail- numerical modeling of debris flow kinematics and to
way Subdivisions along the base of Klapperhorn are also facilitate the interpretation of the results (Fig. 2).
shown. The main creek in the foreground is at Mile 54.3 The geospatial data related to the debris flow
of the Albreda Subdivision. The rockshed at Mile 54.7 is vi- characteristics are managed using the GIS system,
sible (From Davies, 2005). Photo in the top left corner show such as debris flow channel topography, debris flow
the debris flow source area.
mode (confined or unconfined), source and deposition
region delineation. For example, the Digital Elevation
Model (DEM) for the debris flow channel topography
can be prepared in GIS using TIN format (Triangulated
Irregular Network). It is then imported into PFC3D
as irregular walls to indicate the mountain surface by
writing a FISH function. FISH is an embedded pro-
gramming language in PFC3D. Figure 3 shows the
three dimensional topography of debris flow channel
in Klapperhorn Mountain area imported into PFC3D
from a digital elevation model generated in GIS.
Numerical modeling of debris flow was performed
in PFC3D with the fluid option. This approach ensures
that the 3D kinematics of the flowing mass, the force

Figure 2. Integration of PFC3D and GIS (Geographic

Information System).

both railways has been impacted by debris flow events

in the past. Most of these debris flows have reached
the track elevation with varying consequences. Figure
2 clearly shows the feature of distinct debris flows in
the study area, particularly at mileage 54.3. At this
location a major channel has formed and this channel
has been engineered to reduce the impact to the track.
Davies et al (2005) classified the study area into six
drainage basins based on detailed air photograph inter-
pretation and field survey works. The characteristics Figure 3. GIS was used to create the digital elevation model
of the debris flow drainage basins are described in (DEM) for the site and export the topography as a triangulated
Davies’ work including the feature of debris flow irregular network. These were then imported into PFC3D.
creeks, debris source areas, debris deposition areas and The screen shot represents the final DEM in PFC3D covered
their transportation. In general, the bedrock in upper with fluid domain.

770
transmission within the granular mass and in particu- and initial conditions specification; loading, solution
lar the lateral expansions are properly accounted for. and sequential modeling; and finally interpretation of
A ‘‘fixed coarse-grid fluid scheme is implemented in results.
PFC3D for particle-fluid coupling simulation. This According to the field survey, the boundaries for
scheme solves the continuity and Navier-stokes equa- debris flow material source were identified. Four new
tions for incompressible fluid flow numerically in an boundary walls were created in PFC3D to specify the
Eulerian Cartesian coordinate system, and then derives generation region for the assembly of particles. The
the pressure and fluid velocity for each fixed grid particles are statistically uniformly distributed with
(or cell) by including the influence of particles, and radius from 0.5 meter to 2.5 meter. Under gravity,
the corresponding porosity, within each cell. Driving the assembly was compacted and reached equilibrium
forces from the fluid flow are applied to the parti- state on the irregular debris channel wall. In this study
cles as body forces. These forces are also added to area, the assembly has an approximate porosity of
the fluid equations and cause change in momentum, 0.4. Three different source boundaries were specified
as reflected by the change in the pressure gradient in to investigate the effect of magnitude of debris flow
the flow direction (Manual of PFC3D). The present source indicated by the number of particles. In this
formulation in PFC3D only applies to fully saturated, study, we generated three different assemblies of parti-
fixed, fluid domains. Interaction is only between the cles which have 600 particles, 1604 particles and 7309
fluid and particles; fluid interaction with walls is not particles respectively.
included. Also, fluid boundaries can only be spec- The particles were released by deleting the bound-
ified for a rectangular fluid domain; arbitrary fluid ary walls. Different material properties were set before
boundaries cannot be given. the particles release including contact property (elastic
During the simulation, the status of each particle stiffness and Coulomb’s friction), damping ratio and
was recorded at a certain time interval, for example fluid conditions. In order to account for the modulus
every 5 seconds. It includes the position of each parti- of debris flow mass, a modulus-stiffness scaling rela-
cle, velocity, displacement, force, stress, stain rate and tion was employed (Potyondy & Cundall 2004). The
energy. Using FISH, this information can be exported normal and shear stiffness of particles are scaled with
to ASCII files and then import into GIS system for the particle radii to achieve a constant grain modulus.
interpretation. A rectangular fluid domain was setup covering the
Spatial averaging on the numerical result is per- debris flow channel. The fluid pressure boundary con-
formed by means of a spatial mean interpolation dition was assigned based on the assumption that the
approach in the spatial analyst extension of GIS. The fluid flows parallel to the slope. So the fluid pres-
discrete time series data generated by PFC3D are then sure boundary was applied to ensure the fluid pressure
converted into various continuous raster surfaces. The gradient is same as slope gradient.
raster surfaces could represent the spatial distribution
of kinetic property of the flowing mass in terms of
velocity, stress or thickness at a certain time step. 3.3 Parameter calibration
Figure 4 shows an example of converting particle The input parameters for debris flow numerical simu-
velocity calculated from PFC3D into a velocity raster lation cannot be measured easily, since the understand-
surface in GIS. It can be seen that using raster surface, ing of rheology or mechanics of the flowing material
the modeling results from discrete element code can
be easily interpreted and provide useful information
for the further analysis and prediction.
Five energy and work terms are also be calculated
and traced in every PFC3D run including body work,
boundary work, strain energy, kinetic energy and fric-
tion work. In this paper, we focus on the kinetic energy
and friction work. The kinetic energy represents the
total kinetic energy of all particles, accounting for
both translational and rotational motion. The friction
work indicates the total energy dissipated by frictional
sliding at all contacts.

3.2 Procedure
Specific aspects must be considered in the debris flow
model creation and solution including: the assem-
bly of particle generation and compaction; choice Figure 4. Converting particle velocity calculated from
of contact model and material properties; boundary PFC3D into velocity raster in GIS.

771
 (a) (b)

Figure 5. Parameters calibration. (a) Friction coefficient

calibration; (b) Contact properties calibration.

is still limited. In discrete element method, the con-

tact force and displacements of a stressed assembly of
particles are determined mainly by contact properties,
frictional and damping coefficient. The parameter cal-
ibration was conducted using a trail-and-error analysis
method. It takes into account both the debris flow run
out and the shape of debris flow transportation.
From the field work and aerial photo interpreta-
tion, the detailed debris flow channel path can be
delineated indicating the debris flow initiation, trans-
portation and deposition. With proper parameters, the
simulated particles should travel within the defined
debris flow channel. The particles dropping out the
debris flow channel boundary are considered as wrong
particles. The percentage of wrong particle to total par-
ticles is defined as a modeling error. Figure 5 show the
calibration result for the frictional coefficient and con-
tact properties determined by Young’s modulus. The
process was conducted using a smaller assembly of
particles (600 particles). The calibration results show Figure 6. A Series of screenshots at different time indicat-
that using frictional coefficient of 0.5 (friction angle ing the debris flow kinematics.
is about 26 degree) and Young’s modulus of 50 Mpa,
a model could obtain a better result.
The damping coefficient affects the kinematics particles resulted in the various flow mass configu-
of the flowing mass in an evident way. However rations during the flowing process. A quite compact
this parameter is not clearly related to any physical granular mass configuration dominates the rear part
mechanism. The realistic value must be obtained by of the flow while the debris flow mass front is more
back-analysis of the experimental data. Crosta et al dispersed with time.
(1991) found that using damping coefficient = 3%, The most reprehensive component of the flowing
the numerical model can reproduce the experimental kinematics is the velocity. The macroscopic velo-
flow by Hutter et al (1995) with high satisfactory. So city distribution during the flowing process is shown
in this study, the damping ratio of 3% was utilized for in Figure 7. A very non-uniform velocity distribution
all the simulations. was observed. At the initiation stage, the magnitude
of velocity increases gradually away from the source
region indicating the kinetic energy propagates down-
4 RESULT AND DISCUSSION ward the slope. Most of the particles obtain a high
velocity with a range from 3 m/s to 5 m/s when they
4.1 Results reach the central part of the debris flow channel and
then decrease gradually when travel into the lower part
A real scale three dimensional numerical model was
of the channel.
established and performed for debris flow M54.3 in
Klapperhorn Mountain study area. A series of pictures
in Figure 6 show the typical configuration of flow mass
4.2 Sensitivity analysis for debris flow kinematics
at different phases of debris flow dynamics. The trans-
portation of debris flow mass is highly controlled by A preliminary analysis of the influence of the numer-
the irregular channel topography. The interaction of ical parameters on the debris flow kinematics was

772
 of assembly of particles results in a larger energy dis-
sipation and the larger particle tends to obtain higher
kinetic energy during the simulation.

4.3 Velocity profile

Velocity profiles were generated along the longitudi-
nal section along central and left line of debris flow
channel (Figure 9). The velocity along both central line
and left line was vibrating frequently due to the algo-
rithm PFC uses and the influence of irregular surface.
It can be seen for the central line velocity profile, the
velocity vibrates strongly at the upper flow path indi-
cating a high particle interaction at the steeper channel
portion. In the lower part of flowing path, the velo-
city of debris mass decreases gradually until it vibrates
again when reaching the deposition area. The velocity
profile adjacent to the left bank of the channel shows
an irregular vibrating character which might indicate
Figure 7. Velocity distribution during the simulation. the interaction between debris flow mass and the left
boundary of channel.
In order to study the effect of topography on the
6.00E+07 6.00E+07
debris flow kinetics, statistical correlation between
5.00E+07 5.00E+07
10Mpa
velocity and slope of flowing path was conducted.
It can be found that at a certain time period dur-
Kin e t ic E n e r g y (J )

4.00E+07 30Mpa
K in e t ic E n e r g y (J )

0.1 0.3 4.00E+07

3.00E+07
3.00E+07
ing the flowing process, the particles traveling along
2.00E+07

1.00E+07
0.7
0.5
2.00E+07
70Mpa the steeper portion do not necessarily obtain a higher
1.2

0.00E+00
1.0
1.00E+07
50Mpa
velocity compared to those along the gentle portion.
However they show a higher potential to gain the
0 .0 0 E +0 0

1 .0 0 E +0 7

2 .0 0 E +0 7

3 .0 0 E +0 7

4 .0 0 E +0 7

5 .0 0 E +0 7

6 .0 0 E +0 7

7 .0 0 E +0 7

100Mpa
0.00E+00

Time step
1.00E+05 1.20E+05 1.40E+05 1.60E+05 1.80E+05 2.00E+05
Time step
higher kinetic energy since they definitely obtain a
(a) (b) higher velocity changing rate (or strain rate) when they
travel into a steeper terrain (Figure 10a).
Figure 8. (a) Correlation between kinetic energy and fric- Figure 10b show the velocity profiles along a cross
tion coefficient, (b) Correlation between kinetic energy and section located at 300 m far away from the debris
contact properties determined by Young’s Modulus. source at different time. The pattern of the curves at
time = 200 s and time = 400 s show a general trend
that the velocity at the center of the channel has a
performed. The effect of friction coefficient and con- higher magnitude but the curves at time = 300 s and
tact properties on the kinetic energy of flowing mass is time = 500 s show an opposite phenomenon. It, on the
shown in Figure 8a. As expected, the results show the other hand reveals that the discrete element method can
total kinetic energy of flowing mass has distinct depen-
dency on the friction coefficient between particles
and between particles and walls. The kinetic energy
decreases dramatically with the increase of contact
friction coefficient. The lower kinetic energy accounts
for a lower runout distance for the flowing mass. The
high dependency of kinetic energy on the contact stiff-
ness can also be observed from Figure 8b. The contact
stiffness is determined by the Young’s modulus using a
stiffness scaling method. The loose flowing mass with
lower Young’s modulus tends to have a higher mobility
than the dense flowing mass with high Young’s mod-
ulus. Therefore the effect of contact stiffness can not
be neglected for the simulation of flowing mass kine-
matics. The effect of size of particle and assembly on
the kinetic energy and its dissipation was also ana- Figure 9. Velocity profile along longitudinal section in
lyzed. The general trend is that an increase of the size central and left line of the debris flow channel.

773
 The modeling the debris flow kinematics using
the discrete element method accounts for the micro-
scopic behavior of the flowing mass. PFC3D was used
to investigate the microscopic dynamic behavior of
debris flows over an irregular ground surface.
The numerical output of discrete element code was
easier to interpret when integrated with a GIS sys-
tem. A time series of continuous raster surfaces was
Figure 10. (a) Velocity rate vs. Time for flowing mass trav-
eling on different terrains. (b) Velocity profile at different
generated using a spatial averaging approach in GIS
time along cross section which is 300 meter away from source. to represent the macroscopic behavior of the debris
flow kinematics, which provided useful information
for further analysis and prediction for the debris flow
hazard.
The debris flow kinematics is highly dominated
by the characteristics of the irregular ground surface
which has a direct effect on the internal interaction of
the flowing mass. The most reprehensive component
of the flowing kinematics is the velocity. The spatial
distribution of velocity has been studied in detail for a
Figure 11. Velocity profile of flowing mass crossing the
railway track. (a) large assembly of particles; (b) small debris flow in the Klapperhorn Mountain. The poten-
assembly of particles. tial kinetic energy of the flowing mass reaching or
crossing the railway track at the base of the moun-
tain could be evaluated using the predicted velocity
magnitude.
capture the transient feature of the kinetic properties
of the flowing mass.
ACKNOWLEDGMENTS
4.4 Velocity profile crossing railway
This research was supported by the Canadian Railway
A big concern for the debris flow hazard analysis along Ground hazard Research Program http://www.tc.gc.
railway is to predict the velocity or kinematical energy ca/tdc granted by CN, Pacific Railway, Transport
of flowing mass when reaching and crossing the rail- Canada and Natural Sciences and Engineering Re-
way track. The maximum and mean velocity profile search Council of Canada, and partly by the National
were prepared using two different simulations (small Science Foundation of China (40501055). The authors
and large assembly of particles) to indicate the kinetic also wish to acknowledge the contribution of Mr.
property of flowing mass crossing the railway track Michael Davies from BGC Engineering Inc. for pro-
at Albreda subdivision (Figure 11). It can be seen viding data and sharing his experience of field survey
that the mean velocity profiles from the two different of debris flow in the study area.
simulations show a similar tend. They are primarily
changing in the range from 1 m/s to 2 m/s. But the
maximum velocity profiles exhibit a large difference. REFERENCES
The particles in the smaller assembly tend to get higher
maximum velocity due the smaller energy dissipation Chen, H. & Lee, C.F. 2000. Numerical simulation of debris
by interaction of particles. The kinetic energy could be flows. Can. Geotech. J., 37(1): 146–160.
easily calculated when taking into account the mass of Crosta, G.B., Calvetti, F., Imposimato, S., Roddeman, D.,
flowing mass. Frattini, P & Agliardi, F. 2001. Granular flows and numer-
ical modeling of landslides. Thematic report. Debrisfall
assessment in mountain catchments for local end-uses
contract No EVG1-Ct-1999-00007.
5 CONCLUSIONS Cundall, P.A. & Strack, O.D.L. 1979. A discrete numerical
model for granular assemblies. G’eotechnique 29: 47–65.
A debris flow composed of complex mixture of materi- Davies, M.R., Froese, D.G. & Cruden, D.M. 2005. Klap-
perhorn mountain debris flows, Yellowhead Pass, British
als exhibits a distinct discrete nature in its physical and Columbia. Proceedings of International Conference on
mechanical behavior. The continuum theory which Landslide Risk Management, Vancouver. D010.
is widely used for flowing dynamics has difficulties Davies, M.R. 2007. Klapperhorn mountain debris flows, Yel-
accounting for the internal interaction in the solid-fluid lowhead Pass, British Columbia. MSc Thesis, University
mixture of the debris flow mass. of Alberta.

774
Denlinger, R.P. & Iversion, R.M. 2001. Flow of variably Pirulli, M. & Giuseppe, S. 2007. PROPAGATION OF
fluidized granular masses across three-dimensional ter- DEBRIS FLOWS: COMPARISON OF TWO NUMER-
rain: 2. Numerical predictions and experimental tests. ICAL MODELS. In V.R. Schaefer, R.L. Schuster &
J. Geophys. Research, 106: 553–566. A.K. Turner (eds), Proceeding of 1st North American
Hutter, K., Koch, T., Pliiss, C. & Savage, S.B. 1995. The Landslide Conference June 3–8, 2007 Vail, Colorado:
dynamics of avalanches of granular materials from initia- 1542–1551 (CD-Rom).
tion to runout. Part II. Experiments. Acta Mechanica 109: Potyondya, D.O. & Cundallb, P.A. 2004. A bonded-particle
127–165. model for rock. International Journal of Rock Mechan-
Hutter, K., Svendsen, B. & Rickenmann, D. 1996. Debris ics & Mining Sciences 41: 1329–1364.
flow modeling: A review. Continuum Mech. Thermodyn. Richards, K., Bithell, M., Dove, M. & Hodge, R. 2004.
8: 1–35. Discrete-element modelling: methods and applications in
McDougall, S. & Hungr, O. 2005. Dynamic modelling of the environmental sciences, Phil. Trans. R. Soc. Lond. A,
entrainment in rapid landslides. Can. Geotech. J., 42(5): 362: 1797–1816.
1437–1448. Zhu, H.P. & Yu, A.B. 2005. Steady-state granular flow in
Mountjoy, E.W. 1980. Geology, Mount Robson, West of a 3D cylindrical hopper with flat bottom: macroscopic
Sixth Meridian, Alberta-British Columbia. Geological analysis. Granular Matter 7: 97–107.
Survey of Canada. ‘‘A’’ Series Map 1499A, 1: 250 000.

775
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Three dimensional simulation of landslide motion and the determination

of geotechnical parameters

Yuhua Lang & Xianqi Luo

China Three Gorges University, Key Laboratory of Geological Hazards on Three Gorges Reservoir Area,
Ministry of Education, China

Hiroyuki Nakamura
Tokyo University of Agriculture and Technology, Japan

ABSTRACT: A geotechnical model was proposed to do three dimensional simulation of landslide motion.
Coulomb’s law was used to express the cohesion, friction angle and other geotechnical parameters along the
slip surface. The total variation diminishing method was used in data processing to disperse three independent
variables: thickness and velocities of x and y directions. A formula on the relationship between static friction
angles and dynamic friction angle was proposed on the basis of three dimensional simulations of worldwide
landslides.

  
1 GEOTECHNICAL MODEL where M = udz; N = vdz; : calculus from
z to h.
1.1 Introduction
The simulation of landslide motion can be helpful for 1.3 Equation of movement
the prediction of disaster area. Landslide mass can
be regarded as a composition of soil particles with There are 3 kinds of forces acting on individual grid
internal friction. This composition can be considered along x and y directions (Figure 1). Arithmetic ele-
as an incompressible viscous fluid then the laws of ments include stress F1 , internal friction F2 and slip
motion can be expressed by Navier-Stokes equations. friction F3 . According to the equation of movement,
And the friction in sliding surface can be expressed by
Coulomb’s law of friction.
Ground

1.2 Laws of mass conservation

Landslide body is divided to grid of right angle as z
basic calculation cell. The average numerical values of
every element are put in the gravity center of each grid. h
By the laws of mass conservation, the mass volume of
flowing into a finite cell is equal to the volume flowing
out of the cell in finite time. Slip interface

∂u ∂v ∂w
dx + dy + dz = 0 (1)
∂x ∂y ∂z y
s
where u, v, w is the speed along x, y, z directions.
Calculus along vertical direction from the bottom z l
of mass to the ground surface h in equation (1) x
dh ∂M ∂N
= dx + dy (2)
dt ∂x ∂y Figure 1. Three dimensional grid of sliding mass.

777
the internal forces along x direction can be expressed Equation of movement along x direction is:
by equations below:

∂h (ux )n+1
i+1/2,j − (ux )i+1/2,j
n
F1 = ρgh dx (3)
∂x (hρ)ni+1/2,j + conv(x)
t
F2 = cdx + ρgh tan φs (4) n
Hi+1,j − Hi,jn
= −gz (hρ)ni+1/2,j
F3 = ρgh tan φm dx (5) xi+1/2,j

dM ∂h − gz (ρh)ni+1/2,j (tan φm )xi+1/2,j − (fs )n+1

= ρgh + cdx + ρgh tan φs + ρgh tan φm (6) xi+1/2,j
dt ∂x
where ρ = density; g = gravity; h = thickness of + (hfx )n+1
i+1/2,j (12)
sliding column; s = dynamic or apparent friction
angle along slip surface; m = dynamic friction angle
in sliding mass. Because the directions of velocity are parallel or
In formula (6), vertical to the grid, One time disperse modulus can be
calculated as below:
dM ∂M ∂M ∂M
= +u +v (7)
dt ∂t ∂x ∂y
li+1,j (ux2 )ni+1,j − li,j (ux2 )ni,j
With the same pattern, the equations along y conv(x) = (hρ)ni+1/2,j
2si+1/2,j
direction are:
dN ∂h + (hρuy )ni+1/2,j
= ρgh + cdy + ρgh tan φs + ρgh tan φm (8)
dt ∂y li+1/2,j+1/2 (ux )ni+1/2,j+1/2 − li+1/2,j−1/2 (ux )ni+1/2,j−1/2
×
dN ∂N ∂N ∂N si+1/2,j
= +u +v (9)
dt ∂t ∂x ∂y (13)

There are no direct results from equations (2), (6), where the resistance along x direction on slip
and (8), they have to be dispersed in data processing. surface is:

1.4 Total variation diminishing method

uxn+1
Total Variation Diminishing method (Harlow et al. (fs )x = (c + ρghn φs )  n (14)
1970) is used in the data processing operation to dis- ux2 + uy2 + uz2
perse three independent variables: thickness, velocity
of x and y directions. Law of mass conservation can
be dispersed as formula below: where hn is the thickness in the n step of calculation, ux ,
uy , uz is the speed of movement along x, y, z directions.
(hρ)n+1
i,j − (hρ)i,j
n
(10)
+ conv(A) = 0 Equation of movement along y direction is:
t

Conv(A) =
(uy )n+1
i,j+1/2 − (uy )i,j+1/2
n
li+1/2,j (hρ)ni+1/2,j (ux )n+1 n+1
i+1/2,j − li−1/2,j (hρ)i−1/2,j (ux )i−1/2,j
n (hρ)ni,j+1/2 + conv(y)
t
si,j

li,j+1/2 (hρ)ni+1/2,j (ux )n+1 n+1

n
Hi,j+1 − Hi,jn
i,j+1/2 − li,j−1/2 (hρ)i,j−1/2 (uy )i,j−1/2
n
+ = −gz (hρ)ni,j+1/2
si,j xi,j+1/2
(11)
− gz (ρh)ni,j+1/2 (tan φm )yi,j+1/2
where l = length of a calculation cell, s = bottom
area of the cell, i, j are subscript for two calculating
− ( fs)n+1
yi,j+1/2 + (hfy )i,j+1/2
n+1
(15)
steps one after another.

778
 li,j+1 (uy2 )ni,j+1 − li,j (uy2 )ni,j 2 DETERMINATION OF GEOTECHNICAL
conv( y) = (hρ)ni,j+1/2
2si,j+1/2 PARAMETERS

+ (hρux )ni,j+1/2 2.1 Input & output data

Data required for the program are the 3D data of
li+1/2,j+1/2 (uy )ni+1/2,j+1/2 − li−1/2,j+1/2 (uy )ni−1/2,j+1/2 ground surface and sliding surface; physical proper-
×
si,j+1/2 ties of the sliding mass such as unit weight, cohesion,
(16) coefficient of internal friction, and seismic data etc.
The results of calculation are velocity, flow thickness
where the resistance along y direction on slip sur- of each grid from which 3D map, longitudinal and
face is: traverse sections of each time interval are outputted
(Lang & Nakamura, 1998).
The geography data of ground and sliding surface
uyn+1 can be obtained from GIS data or other investiga-
(fs )y = (c + ρghn φs )  n (16)
tion. The static friction angle can be obtained by two
ux2 + uy2 + uz2
dimensional slice methods or soil experiments.

1.5 Structure of the simulation program 2.2 Reliability of the simulation

Figure 2 is the flow chart of the simulation program. Reliability of the model is proved by case studies on
It can be divided to mainly 4 steps: (1) data input 20 worldwide landslides (Table 1). Property parame-
and time setting, (2) landform and friction calcula- ters, which make the simulation most suitable to the
tion; (3) movement and thickness calculation; (4) data real one, are selected for each landslide.
output to other graphical tools. Take the No.18 Nigawa Landslide as a example
of case study. It occurred on Jan. 17th, 1995, by
M7.2 earthquake. Its length was 175 m, width 100 m,
volume 110,000 m3 . Figure 3 shows the calculated
start 1

Table 1. Landslides used for calculating coefficient of

input data friction.
confirm of
repeat times
No. Landslide tan φs tan φm tan φc B A
time interval
determinant 1 Kaerikumo 0.132 0.222 0.613 0.353 0.577
calculation of frictions
2 tukaro 0.087 0.176 0.521 0.264 0.507
3 nata 0.009 0.268 0.344 0.277 0.804
landform calculation calculation onthe relation
4 Frank 0.012 0.466 0.839 0.479 0.570
of each cell between thickness and velocity 5 tuedani 0.044 0.268 0.543 0.312 0.574
6 kinnkoji 0.044 0.268 0.488 0.312 0.639
7 Madison 0.123 0.176 0.445 0.299 0.672
8 nakagi 0.141 0.176 0.613 0.317 0.517
movement calculation
of x direction Recurrent 9 Itinomiya 0.087 0.176 0.374 0.264 0.706
determinant 10 Kobuki 0.070 0.158 0.404 0.228 0.565
11 Nigorisawa 0.105 0.141 0.268 0.246 0.917
12 Sala 0.141 0.176 0.613 0.317 0.517
movement calculation
of y direction
13 Matukosi 0.044 0.176 0.543 0.220 0.405
thickness output 14 Jizukiyama 0.105 0.194 0.466 0.299 0.642
15 Jiaojia 0.141 0.249 0.625 0.390 0.624
1
16 Shangtan 0.141 0.213 0.625 0.353 0.565
parameters update
17 Tuedani 0.158 0.213 0.675 0.371 0.550
18 Nigawa 0.044 0.176 0.325 0.220 0.677
19 Takaratuka 0.070 0.087 0.213 0.157 0.741
time confirm
20 Dangjiacha 0.141 0.194 0.601 0.335 0.557

φs = dynamic friction angle along slip surface;

END
φm = dynamic friction angle in sliding mass;
φc = static friction angle;
Figure 2. Flow chart of simulation program. B = tan φs + tan φm ; A = B/ tan φc

779
 of getting by soil experiments. Two lines up and down
the formula line are dropped to cover all of the data;
The range of friction angle can be obtained to simulate
maximum and minimum extension of landslide mass.
The different shapes of marks in Figure 4 show the
different inducement of landslide. Circle, triangle, and
square represent the inducement of earthquake, rain-
fall, and others respectively. Triangle marks of rainfall
induced landslide situated all on the upper side of the
formula line. It mains that this kind of landslide has
bigger coefficient of friction than that of earthquake
induced landslide.
Figure 3. 3D image of hazard area and calculated area of
the Nigawa Landslide.
3 PREDICT HAZARD AREA BY SIMULATION
0.60

3.1 Process of simulation

+tan4˚
4 The method and process of hazard area prediction are
0.40 15
proposed as following: (1) Calculate the critical slid-
1 -tan4˚
6 5
20
16
17
ing surface and static friction angle in main section
7
3
9 14 2
8,12 of a landslide; (2) Determine the 3D shape of sliding
11
18
10 13 mass according to the critical sliding surface, geology,
0.20
19
and geomorphology of slope; (3) Calculate the max-
:Earthquake imum and minimum dynamic friction angle by using
:Rainfall of the proposed formula; (4) Simulate the movement
:Others
0.00 of landslide by the developed model.
0.00 0.20 0.40 0.60 0.80 1.00

3.2 Recurrence of Tonbi Landslide in Tateyama

Figure 4. Relation between static coefficient of friction and
dynamic coefficient of friction. (Numbers in the figure show Tonbi Landslide was triggered by a M7.1 earthquake
the number of landslides in Table 1). on Feb. 26th, 1858 in Tateyama area, Japan. Its
1.23 × 108 m3 debris formed a nature dam across
the valley. The dam broken after several weeks then
3D image after slide. The calculating mesh is 30 × 30
brought huge damages to the towns of down stream.
with a width of 10 m. The calculated area mach the
The simulated area, volume and speed is most sim-
hazard area when φs = 2.5 degree, φm = 10.0 degree.
ilar with occurred one when the coefficient below
All other case studies do the same kind of calculation
were used (Figure 5): r = 18 kN/m3 , c = 0.10
in order to get suitable parameters.
kN/m2 , φs = 3.0◦ ; φm = 18.0◦ ; φc = 33.0◦ .
It is generally considered that the dynamic coeffi-
cient of friction is around 0.7 times of static one. From
the Table 1 we can see that friction on the interface
plus the friction inside sliding mass (tan φs + tan φm )
is about 0.4–0.7 times of the static one (tanφc ).

2.3 Determination of dynamic friction angle

Figure 4 shows the relation between static coefficient
of friction and dynamic coefficient of friction. Num-
bers in the figure are the same number of landslides
in Table 1. The relation line can be summarized to a
formula below:

tan φs + tan φm = 0.41 tan φc + 0.10 (17)

The correlation modulus of all data is 0.77. It shows

a good relationship between the two coefficient. It can Figure 5. Tonbi Landslide simulation can be used to get
be used to calculate the dynamic friction angle instead parameters of this area for further prediction.

780
All the parameters in this simulation can be used to A formula on the relationship between static friction
do prediction of landslide in the same area. angle and dynamic friction angle is proposed, based on
three-dimensional simulation of worldwide landslides.
The process of hazard area prediction of landslide is
3.3 Prediction of landslide in Tateyama area
proposed with a case study in Tateyama area.
Stability of slopes in this area was investigated and
the possibility of sliding was ranked. Siqirou Slope
was pointed out as the most unstable slope. Critical REFERENCES
sliding surface and other geomorphology data were
determined by investigation and 2D calculation. Yuhua Lang & Hiroyuki Nakamura. 1998. Characteristics
The simulation result shows that the landslide will of slip surface of loess landslides and their hazard area
dam valley and bring huge damages to down streams. prediction. J. Japan Landslide Society, 35–1, 9–18.
Harlow, F.H. & Amsden, A.A. 1970. A numerical fluid
dynamics method for all flow speeds. J. Comp. Phys.,
Vol. 8, 197–213.
4 CONCLUSIONS

A three-dimensional movement simulation model

is developed to predict the extension of landslide.

781
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Stability analysis and stabilized works of dip bedded rock slopes

Jin-yan Leng & Zhi-dong Jing

The Third Railway Design and Survey Institute Limited Corporation, Tianjin, China

Xiao-ping Liao
Fuzhou University, Fuzhou, China

ABSTRACT: In this paper, a typical equivalent unit was brought forward because of inhomogeneity and
discontinuity of bedded rock. Through one engineering case, we obtained orthotropic mechanical parameters of
the equivalent unit by a series of numerical tests, and so the bedded rock had the homogeneity and continuity
characteristics. Then we simulated and analyzed the instability and stabilized works of some dip bedded rock
slopes by finite element method. Finally, a convenient and applied technique to simulate instability and stability
works for some kind dip bedded rock slopes was offered.

1 GENERAL INSTRUCTION 2 TYPICAL EQUIVALENT UNITS

With the construction of large-scale infrastructure, Bedded rock mass is make up of rocks and structure
especially in the west development and the moun- planes. Typical equivalent unit is the large unit that can
tainous construction, there occur many rock slopes. compose characteristics of rock and structure planes.
Among them, the dip bedded rock slope is one of the It is smaller than engineering scale, whereas larger
familiar types of slopes. than structure plane spacing. And so its mechanical
With the characters of large quantity, wide distri- characteristics can represents the wholly bedded rock
bution, diversification, bad condition of engineering mass mechanical characteristics.
geology and instability of works, the dip bedded rock The study rationale of bedded rock mass with ‘‘typi-
slope is analyzed commonly by the theory of rigid body cal equivalent unit’’ is numerical simulation test on it to
limit balance. But coupling action between engineer- study the property of deform and intensive, and obtains
ing and structure is not considered in it. The analogy
technique of engineering geology applied in reinforce
works usually, the works depend on experience of engi- cutting slope
neering mostly, and so either sometimes the works are
stronger or reliability inferiority.
typical equivalent unit
Due to the characteristics of inhomogeneity and dis-
continuity on bedded rock, the typical equivalent unit
was brought forward in this paper. By one engineer- structure planes
ing case, we obtain orthotropic mechanical parameters
of equivalent unit by a series of numerical tests, and
endow the bedded rocks with homogeneity and conti-
nuity characteristics. Then we simulated and analyzed
instability and stabilized works of some a dip bedded
rock slope by finite element method. And so a conve-
nient and applied technique to simulate instability and
stability works for some kind dip bedded rock slopes
was offered in this paper. Figure 1. Dip bedded rock slope sketch.

783
 microfissure

cross joints

structure
planes

Figure 2. Typical equivalent unit sketch.

Figure 4. Shear numerical test.

way it can substitute the physical model test that needs

long time and large cost.
Based on the characteristics of bedded rock, we
simulate and analyze bedded rock slopes with RFPA
software. Through uniaxial compressive tests and
direct shear tests on parallel bedding plane and verti-
cal bedding plane, the mechanical property of bedding
plane are equaled to whole rock mass, and then we can
obtain homologize anisotropic mechanics parameter
of rock mass. Thereby bedded rock with aspectual dis-
continuity grew into orthotropic continuous medium,
and it supply a convenience way to analysis the
stability.

4 STABILITY ANALYSES
Figure 3. Compressive numerical test.
We obtained orthotropic mechanical parameters of
bedded rock by a series of numerical tests. The dip
bedded rock was endowed with orthotropic plastic
typical equivalent unit strength of materials parameter; material. Then we analyzed and simulated cutting
Then the macro layers rock mass is treated as trans- process of dip bedded rock slope by finite element
versely isotropy media, and analysis the stability of method, and summarized instability models of dip
bedded rock slope. bedded rock slope. And so we gave the stabilized
works corresponding to the instability models. Con-
sequently, a convenient and applied technique to
3 NUMERICAL TESTS simulate instability and stability works was offered in
paper. It was testified that it was a feasible and appli-
As a sort of numerical simulation, numerical tests is cable technique by simulated instability and stability
a means of simplicity virtual supplementary, and in a works of an engineering case.

784
5 ENGINEERING CASE Table 2. Parameters of slope rock and structural plane.

5.1 Engineering general situation Mud-sandstone (strongly weathering)

Certain speedway was in construction, after cutting E (Mpa) υ φ

two-stage aback, one segment in the same direction
as layer slope all at once occurred slippage following Vertical 190 0.33 2.1
stratification plane, moved along slope towards about Parallel 180 0.34 1.65
fifteen meters, the slope body structure dismembered
to fractures. It’s a typical dip bedded rock slope.
This segment lithology contains sandy mud rock
and argillaceous sand conglomerated rock, layer thick-
ness is 0.5∼2.0 meter, the rock layer inclines to
speedway line, dip angle is 15◦ ∼20◦ , the max altitude
of slope is 30 meter, slope designed four stages, 1st
step ratio of slope is 1:0.75, stage height is 8 meter,
ratios of 2nd and 3rd stages are 1:1.0, stage height
are 6 meter, per stage have 2 meters platform. It was
TBS planting grass operation in 1st and 2nd stage pre-
designed, 3rd stage was arch framework planting grass
operation, 4th stage was triaxiality net planting grass
operation. Figure 5. Stress-strain curve about vertical compression on
bedding plane.

5.2 Numerical simulation analyses

Procured rock swatch from the 2nd stage top made
into 3 samples, 50 mm ∗ 50 mm ∗ 50 mm, to do com-
pression test; from the 3rd stage toe taken disturbed
soil sample, did soil quick shear strength, Es, without
lateral confinement resist compression test, obtained
mechanics parameter (mean) in table 1.
Based on these physical mechanics parameter
obtained in lab, utilized RFPA finite element means
to do the same direction as layer rock slope proceed
resist compression, shear resistant numerical experi-
mentation, obtained this slope typical equivalent unit
transversely isotropy. Figure 6. Stress-strain curve about parallel compression on
The mechanical property parameters saw in table1 bedding plane.
of the compression test of horizontal and vertical com-
pression test. As per try, drawn the load-displacement Table 3. Parameters of shear numerical of slope rock and
curve, stress-strained curve, obtained cell equiva- structural plane.
lent body compression resistance parameter saw in
Table 2. Mud-sandstone
Based on the structure characteristic of rock slope (strongly weathering) Weak structure plane
and familiar destroy form, this model had be done two
types shear test: one was shear surface along rock bed- G (Mpa) τ (Mpa) G (Mpa) τ (Mpa)
ding plane, this method could erect sub-macro model
5.1 1.2 2.7 0.023
of bedding structure independently. The other was
erect interlamination structural plane model, erected

Table 1. Parameters of slope rock and structural plane.

sub-macro model for bedded rock body, shorn the rock
E (Mpa) υ φ σ bedding.
As per numerical test result, obtained typical
Mud-sandstone equivalent unit shear resistant parameter (saw watch
(strongly weathering) 240 0.31 45 8.3
table 3), then drawn homologous shear stress-strain
Weak structure plane 3.2 0.35 10.2 0.055
curve (saw chart Figure 7 and Figure 8).

785
 Figure 9. Slope model of finite element method.

Figure 7. Stress-strain curve about shear compression on Table 4. Jinyin mountain slide numerical calculate
bedding plane. achievements.

Maximum
shear Maximum tensile Total plastic
stress (Mpa) stress (Mpa) strain

0.420 0.041 0.0002

0.979 0.129 0.0016

Figure 8. Stress-strain curve about shear vertical compres-

sion on bedding plane.

Through the numerical experiment of this bed-

ded rock body sub-macro model above, obtained
this equivalent cell body transversely isotropy phys- Figure 10. Plastic zone of the 2nd stage excavation.
ical mechanic property parameter, then used ADINA
finite element software, proceeding numerical sim-
ulation to the same direction as layer rock slope. stage more augmentation, and extended approach
Analyses its instability mode, in turn raises reinforce stage top following bedding angle, showed the fact
countermeasure. that there was damage trend of following soft stratum
The primary geological complexion of this slope bedding (Figure 10).
is 0∼32m dip weathering Fractional sand-mudstone, From tension stress distributive zone figure, after
then the under layer is weak sand-mudstone, the inter- the 4th stage’s excavation, stage top appeared small
face is weak spot, the obliquity is 20◦ . Based on the scope tension stress region, magnitude was small
above-mentioned data, and combined pre-designed too; toe region appeared tension stress region in the
instance, we establish numerical stimulant model, the small too, furthermore top stress region and toe stress
figure 9 is discrete cells gridding, mechanical parame- region had run-trough trend; after the 3rd stage exca-
ters are showed in table 4. We will analysis this bedded vation, top tension stress region grew bigger, toe
rock slope through calculating, undo the changes of tension turned into stress concentration point, the max-
slope stress scene, plastic region and other parameters imum tension stress value was 0.129 Mpa, furthermore
in digging. developed from top to toe as plastic zone (Figure 11).
Numerical calculation parameters saw table 4. Before slope excavation, maximum shear followed
Based on the calculation result, from plastic zone linear distribution as depth, shear isolines followed
figure upper can be saw, after the fourth stage excava- depth adequate distribution from shear stress regular-
tion, this stage toe appeared plastic zone, extended to ities of distribution. During the 4th stage excavation,
slope body middle following inclination 20◦ bedding shear stress field occurred defluxion, the characteris-
angle, formed clearly plane plastic zone strip; after the tic approximately was run parallel with line of slope,
third stage excavation, this stage toe appeared plas- toe shear stress became overt concentration of about
tic zone, plastic zone width compared with the fourth 0.420 Mpa, behaved as toe fail in shear. After the 3rd

786
 rock bolt as fence reinforcement, the simulation result
clearly showed that the application ameliorated state
of stress, far and away reduced plastic zone develop-
ment (figure 13), thereby leaded the slope could be
keeping on excavation.
Kept on excavating the 3rd and 2nd stages, the exca-
vation feet brought distinct distortion and structural
plane slack characteristic, plastic zone had followed
inclination 20◦ upswing towards this stage middle,
immediately adopted systematic rock bolt as fence
Figure 11. Tensile stress zone of the 2nd stage’s excavation. reinforcement (if to be necessary, executed excavation
and anchor measure simultaneously, body stress state
got ameliorate, led this stage toe’s plastic zone wane
to pimping scope (figure 14).
Proceeding the 1st stage excavation, this stage
toe came forth plastic zone, distribution ranges
approached its level top; furthermore the interface of
the 4th stage bottom and the 3rd stage was on the plas-
tic zone which was augmenting, considering the result
of stress adjustment and interface influencing synthe-
ses; for the sake of stabilization of wholly slope, the
1st stage exerted retaining wall as shallow layer fence
reinforce, more reduced clearly plastic zone, brought
Figure 12. Shear stress zone of the 2nd stage’s excavation.
wholly slope body into stable state (Figure 15).
According to correlative slope design specifica-
tions, the safety factor responded for 1.2, could we
stage excavation, shear stress field kept on deflux- through the means of reduced intensity ascertain for
ion, toe maximum shear concentration turned more the stability coefficient of side slope. Through the
apparent, added to about 0.979 Mpa, the characteris- means of repeat calculation, increased and adjusted
tic of damage became more apparent too (Figure 12). the 2nd and 3rd stages prestressing force magnitude,
The bedding rock anisotropy induced large shear, the
more larger of aeolotropism was, slope body stress
redistributed after excavation, the more bigger of high
intensity bedding than low intensity bedding.
Summed up all of upper discuss, after 2nd stage
excavation of the same direction as layer rock slope,
the distribution and development of plastic zone,
tension stress and maximum shear showed that, in con-
ditions without reinforcement work, only two stages
excavation could be done. Keeping on excavation sim-
ulation, figure had not convergence, showed that the
slope had damaged, so approved that this was a lim-
iting state of equilibrium. the simulation conclusion Figure 13. Plastic zone of the 4th stage suspension roof
was very anastomotic with realism. It was, aforesaid support.
analog computation parameter equal to project site, by
in-house experiments achieved fundamental mechan-
ical parameter, not by amending, adjusting, delet-
ing or filling, fully accounted that its engineering
applicability, result veracity and conclusion reliability.

5.3 Numerical simulation

Both simulation conclusion and engineering practice
showed that, this slope stability was very difference,
needed for fence reinforcement in time.
After the 4th stage excavation, toe site appeared
distinct distortion, immediately adopted systematic Figure 14. Plastic zone of the 3rd stage excavation.

787
 It was important for process distortion of the same
direction bedding rock slope construction of that initial
stress pressured anchor measure, advanced slope short
term and long term stability. As engineering execu-
tive physical circumstances, responded in appropriate
reinforced scheme.
It was very important of that adopting excavation
and reinforcement step by step to prevent stress condi-
tion exacerbation of slope body, control slope cutting
distortion, long-term stability.
Figure 15. Plastic zone of the 4 stage suspension roof With ‘‘Representative cell equivalent’’ model inves-
support. tigating bedded rock mechanical property, the ratio-
nale was: First of all, diverged representative cell
equivalent as finite element’ require, brought rock cell
made slope’s stability factor satisfy standard require- and stratification plane cell; And then ‘‘Representa-
ment, then, resumed practicable strength indexing, by tive cell equivalent’’ proceeded numerical simulation,
calculation body stress mode gradually ameliorated, looked into distortion and intension property of ‘‘rep-
plastic zone reduced obviously, brought slope get off resentative cell equivalent’’, obtained ‘‘equivalent’’
to stable state chronically. materials parameters; Be the last treat macro layers
After the sliding mass was purged in this project, rock mass transverse isotropy media, then analyzed
according to aforesaid fence reinforce measure was engineering stability.
proceeded construction, hitherto the slope is stable,
so the conclusion was brought of that it was feasible,
secure, applied and economical. REFERENCES

Chen Zu-yu, Wang Xiao-gang, Yang Jian, et al. 2005. Rock

6 CONCLUSIONS slope stability analysis—theory, method and programs.
Beijing: China Water Power Press (in Chinese).
Through the analysis of above we drew the several Liang Z.Z., Tang C.A., LI H.X. & Zhang Y.B. 2004. Numer-
ical simulation of the 3D failure process in heterogeneous
conclusions:
rocks. Int J Rock Mech Min Sci, 41(3):419.
Representative cell equivalent was from macro- Long J.C.S. & Witherspoon P.A. 1985. The relationship of
scopic bedding slope be chosen macroscopic rock the degree of interconnection to permeability in frac-
mass. Its size was adequately small than this slope tured. Journal of Geophysical Research. 90(B4):
project, but was big enough compared joint plane and 3087–3097.
stratification plane, furthermore it could be used to He Man-chao, et al. 2001. A new way of deter-
accurately simulate each stratification plane and per mining mechanical parameters of engineering rock
strip joint, and could drew precise conclusion. In this masses. Chinese J of Rock Mechanics and Engineering.
way it was reflected of strength of rock mass and 20(2):225–229 (in Chinese).
Ling Zheng-zhao. 2005. Three-dimensional numerical mod-
deformation behavior enough.
eling of rock failure process. Shenyang: Northeastern
Through RFPA numerical simulation test of repre- University (in Chinese).
sentative cell equivalent, made structural plane lie in Xie Wei-hong, Xie He-ping & Zhao-Peng. 1996. Photo-
wholly rock mass, made discontinuous bedding rock elastic simulation of shear strength of rock fractal joints.
slope into macroscopical isotropic continuous body, Mechanics and Practice. 18(5):39–43 (in Chinese).
in this way simplified this kind questions, at the same Xu Qiang & Huang Run-qiu. 1996. Discussion on self-
time again could reflect rock mass discontinuity and organized critical characters in the course of rock fail-
anisotropy characteristic. This way can be used in ure. Journal of Geological Hazards and Environment
engineering practice widely. Preservation. 7(1):25–30 (in Chinese).
Yang Qiang, Chen Xin & Zhou Wei-yuan. 2005. Anisotropic
Used finite element means carried through numeri-
yield criterion for jointed rock masses based on a tow-
cal simulation analyses to flat-lying bedding rock slope order damage tensor. Chinese Journal of Rock Mechanics
stability, put forward homologous work countermea- and Engineering. 24(8):1275–1282.
sure, used reduced intensity method or increased bulk Zhang Gui-ke. 2006. Study on equivalent orthotropic
density method, ascertained fence reinforced safety mechanical parameters and yield criterion of jointed rock
factor, founded fence work numerical simulation mass and its engineering application. Nanjing; Hohai
technology of flat-lying bedding rock slope. University (in Chinese).

788
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

A GIS-supported logistic regression model applied in regional

slope stability evaluation

Xueping Li & Huiming Tang

Engineering Faculty, China University of Geosciences, Wuhan, China

Shi Chen
ChangCheng Special Steel Stock Company, Jiangyou, China

ABSTRACT: The logistic regression model is a statistic analytical method, in which there usually are
dichotomous variables to be used (two values are given, for example, to the variable ‘‘y’’ ). It can provide
the results in the form of event probability. Supported by GIS Platform, a logistic regression model for the
regional slope stability evaluation was used in evaluation of regional slope stability in new Wushan city site.
When the criterion of probability is 0.163, the logistic regression model gives the prediction accuracies as fol-
lows: 72.08% for landslide occurrence and 81.44% for no landslide occurrence. As a comparison, when the
criterion of probability is 0.274, the Bayesian integrating model gives the prediction accuracies as follows:
79.24% for landslide occurrence and 80.82% for no landslide occurrence and when the criterion of probability is
0.2171, the grouped data-based logistic regression model gives the prediction accuracies as follows: 71.85% for
landslide occurrence and 71.70% for no landslide occurrence predicts actual landslide occurrence at accuracy
and landslide non-occurrence at 80.82%. Therefore, the logistic regression model is a better election in regional
slope stability evaluation.

1 INTRODUCTION theories so that different kinds of modern theories

such as information theory, systematic theory, fuzzy
1.1 Study progress mathematics, expert system, neural network. Com-
puter technology has been widely applied to landslide
Landslide has characteristics of wide distribution, fre-
spatial analysis (Ohlmacher & Davis 2003).
quent happening and quick causing, which is one of
the main disasters in the world. It has been more than
20 years since the study of regional landslide disaster
1.2 Character of Logistic regression model
evaluation and mapping in the world and it has made
remarkable progress during the recent years. Logistic regression model is a type of statistic ana-
The methods of landslide disaster spatial evalua- lytical method, which is usually used by dichotomous
tion can be summarized statistics, sensitivity mapping, variables (the variable ‘‘y’’ is given two values). Dif-
information and all kinds of experience etc (Tang & ferent from linear regression, Logistic regression is a
Jorg 1998). non-linear model. Parameter estimation is Maximum
Although landslide forecast theories have differ- likelihood estimate (MLE) method. It is proven that the
ent evaluation methods, they are based on mechanism MLE of Logistic model possess consistency, asymp-
analyses, favorableness, and their similarities are to totically efficiency and asymptotically normal in the
be taken fully consideration of overlapping for vari- random sample situation. Logistic regression method
ous geological environments to evaluate the possibility can set up models for categorical independent vari-
of causing landslide in certain geological environment able, categorical dependent variable (or continuous
and to circle the relatively ‘‘dangerous section’’ in the dependent variable or mix variable), and there is a
hazardous area center. Different ways of overlapping whole set of criterion to test regression models and
are adapted in evaluation process, and the prediction the parameters. It provides the result as the form of
result is expressed by section graphs. event probability, and it is undoubtedly a better solu-
The characteristics of landslide prediction study tion to apply to the regional slope stability evaluation
are intersected and penetrated by various kinds of (Wang & Guo 2001). According to the comparison

789
of regression coefficients, the slope stability effect then (Mao 1999, Zhang 1991)
factor can be quantitatively analyzed. With the regres-
sion models contrast, the best model on slope stability P(Ai )P(B | Ai )
evaluation can be obtained. For the event probability p(Ai | B) = k , i = 1, 2, . . . , k
result is offered, it will be convenient and assured to j=1 P(Aj )P(B | Aj )
analyze or predict the result. (6)

2.2.2 Bayesian random variable form

2 BASIC RESEARCH METHOD Suppose united distribution density of random variable
ξ , η is p(x, y) = pξ (x)fη|ξ ( y | x), pξ (x) is brink density
2.1 Probability formula of Logistic regression of ξ , fη | ξ (y | x) is conditional density of η to ξ when
Suppose variable Y is used to represent a sample in ξ = x, conditional density of η to ξ , gξ | η (x | y) can be
a group of independent variable’s effect, Y’s evaluate expressed as (when η = y)
rule is
 pξ (x)fη|ξ ( y|x)
gξ |η (x|y) =  ∞ (7)
Y = 1 (right)
(1) −∞ pξ (x)fη|ξ ( y|x)dx
Y = 0 ( false)

The probability of right is P and false is Q, 2.3 Grouped data logistic regression
x1 , x2 , . . . , xm is to represent m influencing factors Suppose independent variable X = (x1 , x2 , . . ., xk ),
to result Y. Logistic regression formula is used to which is effect factor of P, is observed L groups results.
represent the probability of right, which is The jth group is observed nj and A had occurred mj .
It is possible to calculate the frequency when samples
eβ0 +β1 x1 +···+βm xm are large. The equation is fj = mj /nj . We could use the
P= (2)
1 + eβ0 +β1 x1 +···+βm xm frequency as group’s probability estimate. That is

For the relational expression P + Q = 1. The p̂j fˆj

probability of false will be got according to (2), ln = ln = β0 + β1 xj1
1 − p̂j 1 − fˆj
1 + β2 xj2 + · · · + βk xjk + εj (8)
Q= (3)
1 + eβ0 +β1 x1 +···+βm xm
where xjk is the value of the kth variable in the jth group.
Deduced from above formulas, a sample’s result
εj is stochastic error.
and it’s relational factors are nonlinear. Define the two
The regression model may be calculated with Ordi-
probabilities’ rate is odds,
narily Least Square (OLS). At first, individual data
is grouped according to independent variable. Sec-
odds = P/Q = eβ0 +β1 x1 +···+βm xm ondly, every group event probability is estimated.
= eβ0 × eβ1 x1 × · · · eβm xm (4) Then change every group event probability into loga-
rithm odds which see as dependent variable of linear
regression model. All of the dependent variables are
The natural logarithm is categorical variable. Equation (8) has heteroskedastic.
There are some methods to eliminate heteroskedas-
ln(P/Q) = β0 + β1 x1 + · · · + βm xm (5) tic such as weighted least squares (WLS), Box-Cox
transformed method and squared error stability trans-
where β0 , β1 , . . . , βm are called logistics’ regression formed method. All original data including constant
coefficients. Formula (5) is a nonlinear equation, are treated through weighted transformed method in
so the Newton-Raphson Method is applied to solve WLS model (Dang 1995). After data treated, the model
Logistics’ regression coefficient. become (9) equation.
   
1 fj 1 1
2.2 Bayesian statistics inference technology ln = β0∗ + β1∗ xj1
Sj 1 − fj Sj Sj
2.2.1 Events of bayesian formula 
Suppose A1 , . . . , Ak are incompatible events and their 1
6 6 + · · · + βk∗ xjk + uj (9)
sum ki=1 Ai includes event B, that is B ⊂ ki=1 Ai , Sj

790
 
εj 1 (T2 b3 ). The featured landslide of new west area
Sj = V̂ar = (10) is Xiufeng Temple landslide and Sidao Bridge to
pj (1 − pj ) nj fj (1 − fj )
Dengjiawuchang.
where Sj is weight. uj is transformed residual which has
homoscedasticity. The model is also named minimum 3.2 Slope stability influencing factor
chi-square estimation. On one hand, slope stability is influenced by the intrin-
sic factors, such as lithological character, gradient and
slope elevation etc. On the other hand, it is influenced
3 GENERAL INTRODUCTION by the extrinsic factors, such as rain, human activities
OF THE APPLIED SITUATION etc. Rainfall has not been included in the influencing
factor because lack the related data between rainfall
3.1 Geologic Characteristics of the and landslide occurrence in the experimental area. The
experimental Area rock type includes T1 j4 , T2 b1 , T2 b2 , T2 b3 . T1 j4 has
New city site, Wushan county, Three Gorges is the highest intensity so it belongs to firm lithological
selected as the experimental area. Wushan County lies group. T2 b1 and T2 b3 are limestone, marlite which
in the east of Sichuan basin, the north of Daba Moun- has the medium intensity, so they belong to half-firm
tains, the hinterland of Wushan mountain. The county lithological group. T2 b2 has low intensity so it belongs
lies in the center of Three Gorges Reservoir. The to soft lithological group. In the experimental area,
water level is 175.4 meter after reservoir impounding. T2 b1 occupies 17.17% and T2 b3 occupies 44.85%. If
Except some of the inhabiting district of Beimenpo, they are put into the same group, it will occupy 2/3
the majority of the city will be submerged after the and maybe influence regression analysis precise, so
reservoir has impounded. New area lies in the joint they are treated as different group. The closest dis-
part of three structures, that is Huaiyang mountainous- tance to effected tectonic line is regarded as geological
type structure west wring reflect arc—Daba mountain structure factor. The calculation method is as follows.
arc structure, east of Sichuan belt of folded strata and Each tectonic line above level 4 has been treated single
Sichuan, Hubei, Hunan, Guizhou belt of rise folded factor logistic regression, and chosen the tectonic line
structure. The geographical structure is complicated by relevant coefficient R. In the paper, R’s limitation
fold and fault. The bedrock is mainly coastal carbonate is 0.1. The calculation method of the closest distance
of Trias, and then the continental sandy argillaceous to effected tectonic line is to search closest distance
rock. They include light grey medium thick layer to the effected tectonic line of each unit. The result
limestone, dolomite and argillaceous dolomite of Jial- of classified group is ≤61 m, 61–122 m, 122–183 m,
ing River’s fourth segment under the Trias system and >183 m by the singular factor logistic regression
(T1 j4 ); grey argillaceous limestone, dark grey argilla- coefficient. The detailed selection and classification
ceous limestone, dolomite limestone of Badong’s first is as follows.
segment, under Trias system (T2 b1 ); purple argilla-
ceous limestone contaminating silty argillaceous lime-
Table 1. Influencing factors and classification.
stone, silty limestone of the second segment (T2 b2 );
grey dolomite limestone, argillaceous dolomite, dark Landslide
grey medium thick limestone of the third segment factor Expression Classification

Gradient (0 ) X1 ≤10, 10–15, 15–20, 20–25,

25–30, 30–35, 35–40, >40
Elevation/(m) X2 ≤ 150, 150–200, 200–250,
250–300, 300–350, 350–400,
400–450, 450–500, > 500
Slope X3 315–45, 45–90, 90–135,
direction/(0 ) 135–180, 180–225, 225–270,
270–315
Lithologic X4 T1 j4 , T2 b1 , T2 b2 , T2 b3
character
Slope shape X5 Consequent slope 1(β < α),
consequent slope 2(β > α),
reverse slope, tangential slope
The distance to X6 < 61, 61–122, 122–183,>183
the closest
Figure 1. Landslide occurrence section in experimental tectonic
area. line/(m)
Note: The darkest color is the landslide occurrence section.

791
3.3 Bayesian statistics inference research method 4.1.2 Prediction ability
The landslide occurrence accuracy is 72.08%, and
The study is supported by influence factor database
non-occurrence accuracy is 81.44%. The criterion of
and comprehensively applied two multivariate regres-
landslide is Y0 = 0.163. If grid probability P > Y0 ,
sion models, trend surface model and Logistic factor
the grid is judged as the slide will happen, otherwise
model. Combined with GIS technology and statistics
it will not happen.
technology, regional slope stability has been evaluated.
The probability of the first model is regarded as the
Bayesian statistic inference’s priori information. Fac-
tor Logistic model is used to modify trend surface 4.2 The result of bayesian integrating model
model and to form Bayesian integrating model, and
evaluate regional stability. 4.2.1 Trend surface model
Trend surface model is a multivariate Logistic
regression model which applies coordinate u, v of
3.4 Result of the model applied in the grid to explain to variable’s n times multinomial.
experimental area The effect of the second multinomial trend sur-
Supported by MAPGIS, topography data, lithological face equation is better than triple multinomial. The
type data and tectonic line data, etc are input to result is,

−2 −2 −4 2 −4 −4 2
e−2.5682+2.3054×10 μ−1.8320×10 ν−1.4877×10 μ +4.175×10 μν−3.0117×10 ν
pT = (12)
1 + e−2.5682+2.3054×10−2 μ−1.8320×10−2 ν−1.4877×10−4 μ2 +4.175×10−4 μν−3.0117×10−4 ν 2

establish the relevant files. Supported by MAPGIS 4.2.2 Integrating model

Secondary Development Database and Visual C++
6.0, slope stability evaluation system based on Logistic 1
regression model has been established. The experi- PB = 1−PL PT
(13)
ln −ln
mental area is divided into 14, 450 cells with 10 m × 1+e PL 1−PT

10 m size.
Where PL is equation (11).
4 RESULT OF THE MODEL APPLIED IN THE
EXPERIMENTAL AREA 4.2.3 Prediction ability
The landslide occurrence accuracy is 79.24%, and
4.1 The result of six factor Logistic non-occurrence accuracy is 80.82%. The criterion of
regression model landslide is YB0 = 0.274.
4.1.1 Regression coefficient
The regression equation is 4.3 The result of grouped data Logistic model

e−4.0767+0.2278[x1 =1]+0.5328[x1 =2]+···+0.2550[x6 =3] 4.3.1 Choice and coding of categorical variable
p= Three factors have been chosen. They are lithological
1 + e−4.0767+0.2278[x1 =1]+0.5328[x1 =2]+···+0.2550[x6 =3]
(11) character factor, elevation factor and distance to the
closest tectonic line factor. The categorical variable
coding is as follows.
Table 2. Regression coefficient β.

Expression X1 = 1 X1 = 2 X1 = 3 X1 = 4 X1 = 5 X1 = 6 X1 = 7
β 0.2278 0.5328 0.4490 0.5884 0.3516 0.4342 0.4096
Expression X2 = 1 X2 = 2 X2 = 3 X2 = 4 X2 = 5 X2 = 6 X2 = 7 X2 = 8
β 0.1207 −0.0270 0.0087 −0.1734 −0.9061 −0.9127 −0.6961 −0.0067
Expression X3 = 1 X3 = 2 X3 = 3 X3 = 4 X3 = 5 X3 = 6
β −0.0654 0.2507 0.2698 0.1175 0.0625 0.1394
Expression X4 = 1 X4 = 2 X4 = 3
β 1.5995 0.5668 0.8146
Expression X5 = 1 X5 = 2 X5 = 3
β 0.6710 0.2294 0.4236
Expression X6 = 1 X6 = 2 X6 = 3
β 0.7725 0.5752 0.2550

β0 is −4.0767.

792
4.3.2 Grouped data Logistic regression equation 4.3.3 Prediction ability
The landslide occurrence accuracy is 71.85%, and
exp(−4.3976 + 0.8261x1 + 0.4595x2 + 0.1286x3 ) non-occurrence accuracy is 71.70%. The criterion of
P̂ = landslide is Y0 = 0.2171.
1 + exp(−4.3976 + 0.8261x1 + 0.4595x2 + 0.1286x3 )

(14) 5 CONCLUSIONS

1. Bayesian Integrating Model yields the highest land-

slide occurrence accuracy in experimental area.
2. With support of the system, select 100 m ×
100 m, 40 m × 40 m, 4 m × 4 m comparative cal-
culation to test the influence of cell’s scale to the
calculation result. The result shows that three types
scale have the same calculation result.
3. The lithological character is the most sensitivity
factor, the elevation is the second sensitivity fac-
tor and the distance to effected tectonic line is
the third sensitivity factor in six factor Logistic
regression model. But the coefficients of these fac-
tors are 0.8261, 0.1286 and 0.4595 respectively in
grouped data Logistic regression model. The pos-
sible cause is qualitative variable coding. The code
Figure 2. Landslide occurrence section by six factor of 0, 1, 2 and so on imposes amount relative on
Logistic regression model. code in grouped data Logistic regression model.
Note: The darkest color is the landslide occurrence section. This type coding method maybe effect regression

Figure 3. Landslide occurrence section by Bayesian inte-

grating model.
Note: The darkest color is the landslide occurrence section. Figure 4. Landslide occurrence section by grouped data
Logistic regression model.
Note: The darkest color is the landslide occurrence section.

Table 3. Categorical variable coding.

Lithological character Classification T1 j4 T2 b2 T 2 b3 T2 b1

X1 0 1 2 3
The distance to the closest Classification >183 122–183 61–122 ≤61
tectonic line/ (m) X2 0 1 2 3
Elevation/(m) Classification 400–450 350–400 450–500 300–350 200–250 >500 <150 250–300 150–200
X3 0 1 2 3 4 5 6 7 8

793
 precision, but it is not inevitable in grouped data Tang, Ch. & Jorg G. 1998. The Principles and Methodology
Logistic regression model. of Landslide Hazard Assessment. ACTA GEOGRAPHICA
4. GIS plays an important role in the decision-making. SINICA (53) S:157–159.
On the base of GIS, the paper sets up the slope Wang, J.Ch. & Guo, Zh.G. 2001. Logistic regression
stability evaluation Logistic model, Bayesian inte- model—method and application. Beijing: higher educa-
tion press.
grating model by employing the Bayesian statistic Wu, X.C. etc. 2002. The Principles and Methodology of
technique and grouped Logistic model. The results Geographical Information System. Beijing: electronic
are satisfactory. industry press.
Zhang, H.L. & Li Zh.X. & Wang R.Ch. Etc. 2000. Applica-
tion of Bayesian Statistics Inference Techniques Based on
REFERENCES GIS to the
Evaluation of Habitat Probabilities of Bos Gaurus Readei.
Dang, J.B. 1995. Logistic linear regression model is applied Journal of remote sensing (4) 1:66–70.
by dealing with qualitative index. Weaving college foun- Zhang, Q.R. 1990. Geological Trend Surface Analysis.
dation course transaction 8(4): 322–325. Beijing: science press.
He, X.Q. & Liu, W.Q. 2001. Application of regression Zhang, Y.T. & Chen, H.F. 1991. Bayesian Statistics Inference.
analysis. Beijing: Renmin University of China press. Beijing: science press.
Mao, Sh.S. 1999. Bayesian Statistics. Beijing: China Statistic
Press.
Ohlmacher, G.C. & Davis, J.C. 2003. Using multiple logistic
regression and GIS technology to predict landslide hazard
in Northeast Kansas, USA. Engineering Geology (69)
3–4: 331–343.

794
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

The stability analysis for FaNai landslide in Lubuge hydropower station

Kaide Li, Jin Zhang, Sihe Zhang & Shuming He

Kunming Hydroelectric Investigation Design & Research Institute, State Power Corporation of China

ABSTRACT: The stability of large scale landslide. Fanai in Lubuge hydropower station, with complexed
properties, would be the key issue to affect downstream area and villages. So, it is necessary to research, calculate
and estimate the stability by any investigation, including drilling holes and test etc. Fanai landslide, located in
the reservoir area of Lubuge hydropower station, had been slipped partly which affected by construction and
reservoir water. We had lots of detail investigation to estimate the effects on downstream area. Keeping resident
safely on the landslide and getting any significant treatments are our purposes.

1 BASIC GEOLOGICAL CONDITION on both river banks and as the formation of part
landslide.
1.1 Topography 2. Sandstone, shale and mudstone deposited in Tri-
assic period (T1f ), mainly distributed middle-
Fanai landslide located on the reservoir left bank with
southern of landslide. Part of this formation out-
the distance about 1.4 km to 2.2 km from damsite.
cropped in back slope.
Upstream valley is open with 15◦ to 25◦ of both slopes
3. Another deposited layer of Triassic period (T1y )
but downstream is ‘‘V’’ type with slope more than 45◦ .
composed by limestone with sandstone and shale,
The landslide on left bank where valley changed
distributed on south area and downstream of land-
from wide to narrow and flowing direction is S34 W.
slide. Part pf this layers outcropped in back slope
Riverbed elevation is between 1090 m to 1096 m and
and downstream of mountain ridge.
the slope height is about 500 m to 600 m. Reservior
water surface altitude is 1130 m. Quaternary covers normally distributed in the gen-
This landslide locates at 1065 m to 1375 m area. tle slopes and river beds.
According to the topography, its appearance is just a
convex slope or asperity with bluff along trailing edge.
Generally, its geomorphic land surface is showing as 1.3 Geological structure
‘‘arc-chair’’ shape landslide.
Stratum strike is N◦ 35∼65◦ W and the dip is
SW∠30◦ ∼40◦ .Three faults developed in the landslide
1.2 Stratum area showing in Figure 2.
Fault No.1 (F1 ) developed in trailing edge of
The rock types in this region are: landslide with attitude of N30◦ E, NW∠85◦ ∼90◦ .
1. Sandstone, shale intercalated carbon layer or coal Fault No.2 (F2 ) extended from downstream to trail-
shed deposited in Permian age (P2l ), outcroped ing edge of landslide. Fault No.3 (F3 ), with attitude
of N20◦ E, SE∠85◦ ∼90◦ , distributed in the landslide
trailing edge too but it approach downstream. Rock
Joints in the slide region developed normally.

1.4 Hydrogeology
Groundwaters in this region are pore-water, crack-
water & Karst-water etc. Pore-water stored in Qua-
ternary cover layers, crack-water in rock mass and
Karst-water in limestone which distributed on back-
slope. Because Quaternary colluvium is a lossen layer
with strong perviousness, rainfall often discharges
Figure 1. Geological profile of landslide. quickly.

795
 Figure 3. Three-dimensional shape of slip bed.

The leading zone on landslide surface is open and

changed narrow backward (Figure 3). Mid surface is
showing as gentle and rolling with of depth ranging
from 30 m to 100 m. Trailing edge is steep and deep
with the depth of 140 m to 170 m. Total sliding direc-
tion is N61.5◦ W. Total volume of Fanai slip mass is
about 4.3 × 107 m3 .

2.2 Ancient slide formation mechanism

It is an ancient rock landslide. Sliding plane is con-
trolled by rock bedding surface and structure. Based on
physical classification, it is dragging and driving one.
Because of sedimentary layers, such as limestone,
fine sandstone and marl dips along the slope; sliding
Figure 2. Geological map of landslide. region is shown as uniclinal structure.
According to geological history, this ancient slide
formation has been caused by stream corroding and
two faults distributed in trailing edge. Whit the sub-
A few of springs spilled over in this region. Just one sequent river cutting process, the slope had been
spring located on the slopeslide with water discharge deformed backward and created patulous fractures
about 1 L/min to 2 L/min.The groundwater level is along trailing edge. Moreover, unload and loosed
very deep in the back of landslide and the normal depth rockmass happen to collapse and slide.
is 80 m to 110 m. But it gradually become shallower The present landform and surface feature are related
closed to the bank of reservoir. The gradient of water to slided movement.
table is too gentle. With the reservoir level increased, water softens
sedimentary layers and other formation, broken the
leading end and lead to slide finally.

2 LANDSLIDE CHARACTERS
2.3 Developing process of slip mass
2.1 Landslide configuration
This ancient slide was still stabile at the time of pre-
This landslide covered about 0.58 km2 area, 750 m construction of Lubuge hydropower project.
width in leading edge, 690 m to 820 m in middle band In 1979, at the beginning of construction, man-
and 380 m to 520 m in trailing edge, with longitudinal ual activities, such as living waste water drainage and
length of 850 m. The landslide center is max thick with working load, destroy the primal balance. In 1883,
100 m to 160 m. tension cracks were found in No.1 zone (showed as
Generally, sliding thickness is controlled by the bro- Figure 2), but the slope deformation was not lead to
ken types. The mid zone is thicker than the surrounding slide.
belt but average thickness is 14.2 m. Because of the In November and December 1988, after Lubuge
soft layer broken by sliding in the leading belt, slip- reservoir put into use, it induced to 2.9 and 3.1earth-
mass is forced forward and accumulative layer about quake intensity when the storage level arised to 35 m,
15 m to 40 m. Moreover, rock layers on both sides and leading to multi-slides in this region. In the rainy sea-
back belt are thicker than the mid area. But average son, from June to September, 1989, the slope had
thickness is less than 10 m. slided again.

796
Table 1. Accumulated landslide deformation between 2002 to 2003.

Horizontal Perpendicular Horizontal Perpendicular Horizontal Perpendicular

displace- displace- displace- displace- displace- displace-
Observation ment ment Observation ment ment Observation ment ment
point (mm) (mm) point (mm) (mm) point (mm) (mm)

C1 43.1 −4.5 N2 20.7 −9.8 P5 8.8 1.9

D1 24.5 0.2 P1 70.7 −0.8 P6 16.5 −1.0
F 22.9 0.6 P3 28.8 −2.3 P7 14.1 −0.1
G 25.4 −5.1 P4 22.0 0.3 P8 12.1 −0.6
H 20.4 −2.9 P4 16.2 −3.2

At end of 2001, observation network had been

rebuilt in active zone for this landslide. We have got a
lot of typical information at that time of 2001 to 2003
(Figure 5 & Table 1). These data indicate that land-
slide deformation is related to rainfall and mutation of
reservoir water level.

2.4 Material of slip mass

The slip mass is mainly composed of rock block,
granule, silty soil and clay, part rock block are still
keeping the original rock structure. Rock block are
limestone, sandstone and mudstone. Generally, the
frame of landslide is a farraginous accumulation with
rocks and soil.
Figure 5. Curve, histogram reflecting to horizontal dis-
placement, rainfall and water level change (2002∼2003). (a)
Precipitation (b) Displacement vs. time (c) Pool water level 2.5 Formation of sliding belt
vs. time. Generally, the sliding belt is mainly composed of soft
layers or materials, such as carbon clay and silty soil.
Strength of the formation would be decrease while
Researches for this ancient landslide begin in April, meeting water.
1989, when there were 13 observation points built on
sliding surface. But all of these points had been broken
by the sliding. 3 FACTORS AFFECTED ON SLIDING
According to observation datum, deformation rate
in leading edge is related to the first storage time Geological structure, earthquake, groundwater and
and to the strengthen rainfall. During the period process of surface water are main factors acted on the
from May to June, 1989 and July to August, 1990, slide. Moreover, reservoir water mutation and human
this deformation was obvious (Figure 4). In Decem- activity would other issues induced sliding.
ber, 1990, the accumulative horizontal displacement Fanai landslide is an ancient slide. Compositions of
was 32000.4 mm and Perpendicular displacement was slip mass and slid belt, shear strength are key issues
4300.3 mm. These deformations were decreased and for the slide stability. Slide mass is revealing which
tend to stable after September, 1990. maxed with silt and stone. It would be happened to
No.1 zone happened to slide on July 13th, 1997, slipe when affected by external processes.
when reservoir emptied. Most fell under the nor- According to Chinese seismic zoning and seismo-
mal pool level (1130 m) and flowed into riverbed. logical parameter standards, Landslide area intensity
No.2 zone occurred small-scale surface sliding and is VI and the matching horizontal seismic coefficient
collapse. is 0.05 g. It would occur to sliding and collapse when
Additionally, land sliding leaded to orbicular cracks earthquake happened. Based on the observation datum
in the trailing edge and transverse fissures in mid belt. of reservoir storage, induced earthquake is the key
The land sliding had broken farmer’s houses. issue for the slide stability.

797
 Groundwater processing on the slipbody includes results suppose the slip mass in the state of stabile
hydrostatic pressure, hydrodynamic pressure, load, limitation.
sofen in slidebelt and contact brushing. While, sur-
face water process includes brush and cut formed air
4.2 Estimate of stability
face or imperfect topography leading to collapse.
Additionally, external factors induced sliding Based on the calculation of five geological profiles,
includes mutation of storage level and manual exca- analysis and comparing results each other, gener-
vation. If storage level changes frequently, it would ally, we can get a conclusion that Fanai landslide is
not lead to groundwater drain off in short time. Steep steady in natural state, but part of mass, especially the
hydraulic gradient would increase hydrodynamic pres- downstream area, is unstable.
sure in slidbody. On the other hand, large scale change Profile C1 shows as stable in all kinds of condition
of storage would induce earthquake, slope cutting and and its safety coefficient is more than 1.2. Safety coef-
brushing. It would be disadvantaged for the stability ficients of Profile C2, C4 and C5 are more than 1.0
of slope. Figure 4 reflected relationships between stor- except earthquake happened.
age level change, rainfall and horizontal displacement The safety coefficients of Profile C3 close to 1.0.
based on observation datum. Road construction had However, considering the following three conditions
changed pre-stress state of slope, broken the stability (natural state, earthquake and water sudden draw-
of landslide. down), stability coefficient by searching and the slide
risky area are less than 1.0 except Zone ① in nor-
mal state. For the searching process in Profile C3,
4 STABILITY ESTIMATE with deep sliding bed, strong lateral friction drag in
Zone ② and Zone ③, the calculation result would be
4.1 Calculation of stability lead to higher than in actual condition. But Zone ②
and Zone ③ belong to risky area. Whereas, sliding
4.1.1 Calculation methods bed in Zone ① is shallower than in Zone ② and Zone
For analysis the slope stability, EMU (Energy Method ③, with lower lateral friction drag, so the calculation
Upper bound limit analysis) has been used, comparing result would be closed actual condition which belong
with Sarma and other methods. to unstable area.
Different mediums have various anti-seisms in slip
4.1.2 Physical and mechanical parameter body. Slip bed consist rock with lower anti-seism
According to test datum and geological analysis, we but materials of sliding body are loose whit high
ascertained physical and mechanical parameter for cal- earthquake resistance. Calculation result proved that
culating slide stability (Table 2). Considering basic earthquake would cause holistic land sliding, but
intensity as VI and the matching horizontal seismic not affect on part area. Generally, this dead land-
coefficient should be 0.05 g. slide would be re-active in part area when happen
seismic events up to a maximum recorded intensity
of VIII.
4.1.3 Calculation result Sudden drawdown of reservoir level would be the
Considering any kinds of condition, normal state, main issues for sliding. Especially, this mutation of
earthquake and water sudden drawdown, we have three water level would affect to the lower belt in down-
kinds calculation based on five geological profiles. stream area. But middle and higher area would keep
Calculated results are shown as Table 3, Table 4, stable. Lower belt of C3, C4 and C5 with deep slid-
Table 5 and Figure 6, Figure 7. ing bed, closed to or under the reservoir level, would
Datum in Table 3 is the calculated results which happen to slide while meeting sudden drawdown of
suppose the unitary body sliding along the bottom. water. According to the observation on 16th to 29th,
Datum in table 4 and Table 5 are other calculation May 2002, when storage level fell to 16.46 m, sliding
horizontal displacement was obvious. But the hori-
Table 2. Physical and mechanical parameter. zontal displacement would not develop more at that
time on 5th to 19th, May 2003, when storage level
Friction Natural Saturated fell to 12.08 m. According to calculation and obser-
angle Cohesion density unit weight vation result, the influenced extent by water sudden
◦
( ) (kPa) γ (kN/m ) γsat (kN/m3 )
3
drawdown is so limited.
Because it has various stability in different area, cal-
Slip mass 25 35 20 22 culation results and geological condition proved that
Sliding belt 18 28 19 21
Sliding
Fanai slide can be classify into four zones (A, B, C and
plane 35 500 25 25 D in Figure 2). Zone A is stable but would happen to a
little slide. Zone B is a stable area would not happen to

798
Table 3. Stability calculation of unitary landslide.

Computing method Calculated profile C1 C2 C3 C4 C5

Normal state 1.829 1.630 1.177 1.355 1.221

EMU VI seismic intensity 1.471 1.408 1.025 1.121 1.032
20 m Sudden drawdown 1.794 1.585 1.124 1.336 1.203
Normal state 1.324 1.250 1.014 1.282 1.154
Sarma VI seismic intensity 1.104 1.115 0.892 1.084 0.996
20 m Sudden drawdown 1.302 1.232 0.984 1.252 1.128
Normal state 1.492 1.425 1.110 1.358 1.226
Non-equilibrium traction VI seismic intensity 1.228 1.255 0.970 1.138 1.050
20 m Sudden drawdown 1.462 1.393 1.066 1.322 1.192

Note: 1. we supposed slide moved along the sliding plane; 2. Considing the storage level as 1130 m and coefficient of seismic
force is 0.05.

Table 4. Unitary landslide searching calculation of limiting stability.

Computing method Calculated profile C1 C2 C3 (③) C4 C5

Normal state 1.593 1.154 0.970 1.355 1.221

EMU VI seismic intensity 1.317 0.983 0.871 1.121 1.032
20 m Sudden drawdown 1.573 1.121 0.911 1.336 1.203
Normal state 1.486 1.000 0.898 1.282 1.154
Sarma VI seismic intensity 1.248 0.874 0.846 1.084 0.996
20 m Sudden drawdown 1.464 0.980 0.896 1.252 1.128
Normal state 1.540 1.066 0.954 1.358 1.226
Non-equilibrium traction VI seismic intensity 1.280 0.926 0.836 1.138 1.050
20 m Sudden drawdown 1.508 1.038 0.914 1.322 1.192

Note: 1. We supposed both ends of slip are fixed when automatic searching; 2. Considing the storage level as 1130 m and
coefficient of seismic force is 0.05.

Table 5. Upper landslide searching calculation of limiting stability.

Computing method Calculated profile C1 C2 C3 (②) C4 C5 C3 (①)

Normal state 1.278 1.230 0.966 1.283 1.185 1.012

EMU VI seismic intensity 1.113 1.093 0.885 1.093 1.027 0.972
20 m Sudden drawdown 1.268 1.189 0.891 1.240 1.153
Normal state 1.186 1.050 0.874 1.054 0.988 0.972
Sarma VI seismic intensity 1.032 0.932 0.794 0.924 0.870 0.906
20 m Sudden drawdown 1.176 1.042 0.832 1.036 0.968
Normal state 1.218 1.116 0.916 1.110 1.054 1.014
Non-equilibrium traction VI seismic intensity 1.058 1.022 0.816 0.968 0.920 0.912
20 m Sudden drawdown 1.208 1.116 0.874 1.084 1.020

Note: 1. Rearching sliding plane form inner to exterior. 2. Considing the storage level as 1130 m and coefficient of seismic
force is 0.05.

large scale slide but would be moved when caused by 5 AFFECTION OF SLIDING
external effects. Zone C is an unstable area too where it
5.1 Affection for reservoir
would be happen to large scale sliding if it was caused
by external effects. Zone D is another unstable body According to analysis for the stability of landslide,
and would be sliding under other proper conditions. it would be affect on the running of the reservoir.
All of these zones are tight each other. One of them Its movement is small and would not cause to large
slides would be lead to next zone unstably. scale collapse or sudden falling. Sliding activity would

799
Figure 6. Profiles for unitary landslide searching calculation of limiting stability.

Figure 7. Searching calculation profiles of limiting stability (searched from inner to exterior).

800
not destroy the reservoir and power station. However, slidings but would not influence the running of
sliding caused by earthquake would lead to reservoir hydropower station.
sedimentation. 2. The groundwater movement would influence the
landslide stability to some extent. It would be
5.2 Affection for residents and houses dangerous if the storage level fell down quickly.
Emptying the reservoir quickly for flushing sed-
More than 100 resident families living on Fanai slide imentation is risky behavior. In order to keep
surface. Sliding had been broken resident houses. slope stably, controlling water falling speed and
Especially Zone D, in 1997, surface sliding leaded to drawdown are more necessary.
cracks in the farmer’s soil wall. Sliding in some surface 3. Surface sliding is dangerous for the resident build-
areas is the issue for house and building stability. ing foundation. Human activities are one of another
reasons leading further sliding. So it is neces-
sary to emigrate the inhabitant who living on this
6 CONCLUSION AND ADVICE unstable area.
4. Keeping on observation and inspection for the slid-
1. Generally, the landslide would not happen to large ing, building new drainage establishment would
scale sliding. According to limitation of stability, be in favor of slope stability and protecting the
Fanai landslide can be classified into 4 zones. permanent inhabitant.
Zone C and Zone D may be induced some small

801
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Numerical analysis of slope stability influenced by varying water conditions

in the reservoir area of the Three Gorges, China

Shaojun Li & Xiating Feng

Institute of Rock and Soil Mechanics, Chinese Academy of Sciences, Wuhan, China

J.A. Knappett
Division of Civil Engineering, University of Dundee, Dundee, Scotland

ABSTRACT: Landslide disasters have become one of the main problems after the Three Gorges impounded.
Scholars from all over the world have paid much more attention to the stability of slopes in the reservoir area.
Water has been regarded as the main factor which may directly trigger landslides, while the change of water
conditions can’t be avoided in the slopes in the Three Gorges. The water conditions mentioned in this paper
refer to reservoir water level and water table in slope body. They will inevitably deteriorate the slope stability in
some degree. The paper aims to investigate their influence on slope stability. A promising numerical analysis
technique by shear strength reduction based on finite differential method is introduced here to deal with such
a problem. According to the results from a real landslide in the Three Gorges, the curve between the factor of
safety and the change of water level is parabolic: the slope stability will be reduced at the first stage of raising
reservoir water level and then the factor of safety will be increased after that. However, the landslide will always
greatly suffer from the instability when the reservoir water level drops down and the water table rises. It will
provide a direct proof for the landslide prevention and remedial designs in this case.

1 INTRODUCTION limited equilibrium method using Geo-Slope software.

Ding et al. (2004) studied process of the seepage-
The Three Gorges, located on the Yangzi river in stress and obtained the deformation trend and failure
Yichang city of Hubei province, China, is the cur- process in view of water fluctuation for a real land-
rent giantest hydraulic and hydroelectric project in slide based on Flac4.0 . Zhu et al. (2002) & Liu et al.
the world. After the impoundment in 2003, about (2005) discussed the influence by the permeability
2490 potential landslides in the reservoir area will coefficient of soil under water drawdown based on
be induced according to some materials. All the limited equilibrium method. Tang et al. (2005) & Shi
landslides will be inevitably influenced by varying et al. (2004) also focused on the process of water
water conditions, the maximum variable magnitude level drawdown and evaluated the stability toward
is up to 30 m. In Japan, about 60 percent of landslides a real landslide respectively. In addition, Liu et al.
occurred during the sudden drawdown of reservoir (2005) analyzed the deformation mechanism when
water level, while 40 percent are induced by water water level raised and calculated the corresponding
level rise (Liao et al. 2005). The water condition men- factor of safety.
tioned in this paper can be divided into two cases, the In this paper, in view of varying water conditions
first one is originated from the impoundment which in the Three Gorges mentioned above, a different and
will change the water level from the toe to the top of promising numerical method for slope stability anal-
the landslides, the other is from the season’s rainfall ysis called shear strength reduction is to be adopted.
which will alter the water table in the landslide body. Based on a real landslide in the reservoir area, the rela-
In recent years, many Chinese scholars have done a tionship between factors of safety and varying water
lot of works related to the slopes stability in the reser- condition is obtained with shear strength reduction
voir area of the Three Gorges. Liao et al. (2005) got the technique, results will contribute a lot to the stability
relationship between the landslide stability, the per- evaluation and remedial works of slopes in the Three
meability coefficient and drawdown speed based on Gorges.

803
2 SHEAR STRENGTH REDUCTION BASED 3.2 Numerical model
ON FINITE DIFFERENTIAL METHOD
According to the impounding scheme of the Three
Gorges, the simulation of varying water level is
Factor of safety F is widely adopted for slopes sta-
divided into these following steps: 125 m, 130 m,
bility analysis. Shear strength reduction technique,
135 m, 140 m, 145 m, 150 m, 155 m, 160 m, 165 m,
integrated with finite element and finite difference,
170 m, 175 m, 180 m, 185 m.
was put forward as early as 1975 by Zienkiewicz et al.
In order to simulate the change of water table, a
(1975), and has been applied by Matsui (1982), Daw-
coordinate is created as Figure 2, the assumed water
son (1999), Zheng et al. (2004) and others. Compared
tables are almost the same shape and parallel with the
to traditional limit analysis solution, the shear strength
initial water table, there is 2 m space between each of
reduction technique has a lot of advantages. For
the water tables. Final scheme for change of water table
example, the critical failure surface can be found auto-
is listed below: −6 m, −4 m, −2 m, 0 m, 2 m, 4 m, 6 m.
matically, the plastic zone and displacement can also
In the numerical model, the Mohr-Coulomb elasto-
be obtained at the same time. Therefore, it has become
plastic model is adopted, properties of soil for comput-
a promising method applied in geotechnical slopes.
ing analysis are selected from the Table 1. Finally a two
In order to perform slope stability analysis with
dimensional numerical computing model is created
shear strength reduction technique, the soil/rock
as shown in Figure 3. However, during the comput-
strength should be reduced continually until collapse
ing procedure, the dynamic seepage and the strength
occurs. In this case a series of trial factors of safety
reduction of soil due to water immersion are not taken
F trial with cohesion C and friction φ are to be adjusted
into account in this paper.
according to the following equation:

1 3.3 Computing process

C trial = C (1)
F trial The computing starts from a value of F trial = 0.8.
1 According to the Equation 1 and Equation 2, the
φ trial
= φ (2)
F trial decreased C trial and φ trial will be obtained respectively.
Then F inc is set to 0.2, in the next steps, the value of
Factor of safety is to be computed using the explicit- F trial should be increased according to the following
finite difference method in this paper, FLAC. In the formula:
computing principal of FLAC, for given element shape
functions, the set of algebraic equations solved by Fntrial = Fn−1
trial
+ F inc (3)
FLAC is identical to that solved with the finite ele-
ment method. The convergence criterion for FLAC is During the trial computing, if the slope fails, the
the nodal unbalanced force, the sum of forces acting F inc should be reduced to 1/5F inc , then repeat the com-
on a node from its neighbouring elements. Obviously, puting process until the F inc is less than a preset value
it is easier for convergence and will be better for the ε = 0.001. At this time, the slope is in the limit equi-
problems with big deformation. librium state and the corresponding F trial is the final
factor of safety, and the failure surface can be also
obtained finally.
3 CASE STUDY AND NUMERICAL MODEL

3.1 Muzishu landslide in the Three Gorges

4 RESULTS AND ANALYSIS
Muzishu landslide, locating on the south bank of
Yangzi river, is in Zigui county, Chongqing city, China. With a view to the landslide field condition and com-
The landslide has total potential slipping volume of puting schemes mentioned above, the technique of
4.0 × 105 m3 with the length of 130 m∼190 m and shear strength reduction based on finite differential
the with of 80 m∼120 m. According to the mate- is performed by FLAC software, a series of factors of
rial of geological condition, the bedrock, originated safety are calculated as shown in Figure 4 and Figure 5.
from Jurassic system, is mainly made up of sand- Obviously, the relationship curve between reser-
stone. The soil in landslide body, with the maximum voir water level and factor of safety F is similar to
thickness of 25.0 m, consists artificial macadam soil parabola. Firstly F is equal to 1.04 for the moisture
(thickness: 0.5 m∼8.0 m), gravelly soil (thickness: slope with no water table and the reservoir water level
2.0 m∼15.0 m), block stone and macadam soil (thick- is 0 m. However, as the water level rises from 125 m to
ness: 0.5 m∼13.0 m). The parameters of soil and 135 m∼145 m, F decreases from 0.95 to 0.90. Then
bedrock are listed in Table 1 and the typical geological F will increase gradually until to 0.98 with the reser-
section is shown in Figure 1. voir water level 175 m. After that, F will increase

804
Table 1. Properties of gravelly soil in Muzishu landslide.

Physical properties

Bulk density
Moisture (kN/m3 ) Liquid limit Plastic limit Plasticity index
content ω(%) WI(%) Wp(%) Ip(%)
Wet Dry

18.8∼21.9 19.6∼20.9 16.3∼17.6 25.6∼29.5 14.8∼18.0 10.2∼11.5

∗ 20.1 ∗ 20.3 ∗ 17.0 ∗ 17.8 ∗ 16.9 ∗ 10.8

Mechanical properties

Consolidated fast shear Residual shear

Cohesive Internal friction Cohesive Internal friction

strength friction angle strength friction
C(kPa) ϕ(degree) C (kPa) ϕ  (degree)

6.0∼18.0 12.0∼25.0 0∼6.0 10.0∼25.0

∗ 13.0 ∗ 19.3 # 4.0 # 18.0

∗
Mean value of five samples, # mean value of three samples.

198˚
Crack
195 195
19.08
Z6 22.55
Lf

185 J 1X
185
Normal water level after final impoundment
175.5
175 rQ 175
175 Crack
Water level(m)

col-dlQ 165
165 19.25
Crack Z8 22.35

155 Z9
19.5
31.5
J 1X
155
Highway Lf 150.86

145 145
delQ
rQ:Artificial macadam soil
135 134.81
135
col-dlQ:Gravelly soil
125 125
J 1X
Bedrock delQ:Block stone and macadam soil
115 115
0 25 50 75 100 125 150 175 (m)

Figure 1. Geological section of Muzishu landside.

195

Y 185

175

X 165
Space

Highway
tabl
e
155
Clay with gravels
er
ial wat 145
Init 135
Bedrock
J
125

115
0 25 50 75 100 125 150 175 (m)

Figure 2. Scheme for change of water table. Figure 3. Computing model of Muzishu landslide.

805
 1.2 1.13
1.1
1.04 1.05
Factor of safety

1.1
1.0 0.98
0.95 0.95 0.96 0.96
1.0 0.93 0.92
0.90 0.90 0.90 0.91
0.9

0.9
0.8
0 125 130 135 140 145 150 155 160 165 170 175 180 185

Reservoir water level/m

Figure 4. Relationship between the reservoir water level (a) Moisture slope (No reservoir water level and water table)
and factor of safety, Muzishu landslide.

1.1 1.04 1.02

0.99
Factor of safety

1.0 0.95
0.88
0.9
0.81
0.8 0.73
0.7
0.6
-6 -4 -2 0 2 4 6 (b) Reservoir water level 145m
Space based on initial water table/m

Figure 5. Relationship between the water table in landslide

body and factor of safety, Muzishu landslide.

rapidly. The result indicates the natural slope is almost

in the sate of limit equilibrium, at the first stage of
raising reservoir water level, the slope is in unstable
condition, this is why many landslides have occurred
during this time. Then after the water level rises over
145 m, the slope stability will be improved. On the (c) Reservoir water level 155m
contrary, if the reservoir water drops down, the slope
stability will be greatly reduced, and it will possibly
induce the landslide.
On the other hand, as the water table in landslide
body increases as shown in Figure 5, the factor of
safety will be greatly decreased from 1.04 to 0.73, the
landslide will definitely be triggered. This condition is
more dangerous and probably takes place due to heavy
rains.
At the same time, the slope deformation (plastic
zone and displacement vector) of each condition is
(d) Reservoir water level 175m
also obtained as shown in Figure 6 and Figure 7. For
the moisture slope, it is presumed that there are no
Figure 6. Distribution of plastic zone (potential failure
water level in front of slope toe and no water table in surface) under different reservoir water levels.
landslide body. The potential failure surface doesn’t
run through as shown in Figure 6(a). On this moment
FOS is 1.04, the slope is almost in critical condition. move up toward the slope top gradually, see Figure
The results of plastic zone indicate that the potential 6(c) and 6(d).
failure surface varies with the change of water condi- Great attention should be paid to the condition of
tions. When the reservoir water level is lower than water level 175 m. In this case, although it seems the
145 m, the potential failure surface almost runs from whole slope is stable, the failure surface still locates on
the slope top to toe, Figure 6(b). However, as the reser- the top of the slope. This part will definitely slip due to
voir water level rises, the potential failure surface will some deteriorating external factors, such as heavy rain

806
 ACKNOWLEDGEMENT

Financial support from the Pilot Project of Knowl-

edge Innovation Program of the Chinese Academy
of Sciences under Grant no. KJCX2-YW-L01 and
the Special Funds for Major State Basic Research
Project under Grant no. 2002CB412708 are gratefully
acknowledged.

REFERENCES
Figure 7. Distribution of plastic zone (potential failure
surface) with water table +6 m as shown in Figure 2. Liu Caihua, Chen Congxin & Feng Xia-ting. 2005. Study on
mechanism of slope instability due to reservoir water level
and sudden drawdown of reservoir water level. There- rise, Rock and Soil Mechanics, Vol.26(5), 769–773.
fore, some reinforcing measures such as anti-sliding Dawson, E.M., Roth, W.H. & Drescher, A. 1999. Slope sta-
bility analysis by strength reduction, Geotechnique, 49,
pile and anchor cables should be taken to control the
No.6, 835–840.
potential failure. Zhu Donglin, Ren Guangming & Nie Dexin. 2002. Effect-
ing and forecasting of landslide stability with the change
of reservoir water level, Hydrogeology and Engineering
5 CONCLUSION Geology, No.3, 6–9.
Liao Hongjian, Sheng Qian & Gao Shihang. 2005. Influence
In this study, the analysis of slope stability influenced of drawdown of reservoir water level on landslide stabil-
by varying water condition is presented based on shear ity, Chinese Journal of Rock Mechanics and Engineering,
strength reduction technique, from the case study of Vol.24(19), 3454–3458.
Muzishu landslide, some conclusions can be drawn as Tang Huiming & Zhang Guangcheng. 2005. Study on slope
follows: stability during reservoir water level falling. Rock and Soil
Mechanics, Vol.26 (SUP.2), 11–15.
1. Shear strength reduction technique based on finite Matsui, T. & San, K.C. 1992. Finite element slope stability
differential is an effective method for slope numeri- analysis by shear strength reduction technique. Soils and
cal analysis, through which both the factor of safety Foundations, 32, No.1, 59–70.
and failure surface can be obtained. Shi Weiming & Zheng Yinren. 2004. Stability evaluation of
2. During the rise of reservoir water level, the factor landslide under reservoir water level drawdown condition.
Geotechnical Investigation & Surveying, No.1, 27–35.
of safety will decrease at the first stage, the slope
Liu Xinxi, Xia Yuanyou & Lian Cao. 2005. Research on
maybe fail in this condition. While F will increase method of landslide stability valuation during sudden
when the reservoir water reaches a certain level drawdown of reservoir level, Rock and Soil Mechanics,
which is 145 m in Muzishu landslide. On the other Vol.26(9), 1427–1431.
hand, the slope will always suffer from the seri- Ding Xiuli, Fu Jing & Zhang Qihua. 2004. Stability analysis
ous instability when decreasing the reservoir water of landslide in the south end of Fengjie highway bridge
level and raising the water table in the slope body. with fluctuation of water level of Three Gorges reservoir,
3. After the impoundment in the Three Gorges, plas- Chinese Journal of Rock Mechanics and Engineering,
tic zone and potential failure surfaces will move Vol.23(17), 2913–2919.
Zheng Yinren & Zhao Shangyi. 2004. Application of strength
up toward the top of slope. Much more attention
reduction FEM in soil and rock slope, Chinese Jour-
should be paid to this phenomenon. Particularly nal of Rock Mechanics and Engineering, Vol.23(19),
after the final impounding level of 175 m, some 3381–3388.
effective reinforcing measures, such as anchor Zienkiewicz, O.C., Humpheson, C. & Lewis, R.W. 1975.
cables and anti-sliding piles, are essential in those Associated and non-associated visco-plasticity and plas-
potential dangerous parts. ticity in oil mechanics. Geotechnique 25, No. 4, 671–689.

807
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

A numerical study of interaction between rock bolt and rock mass

X.P. Li & S.M. He

Key Laboratory of Mountain Hazards and Surface Process, Chinese Academy of Science,
Institute of Mountain Hazards and Environment, CAS, Chengdu, China

ABSTRACT: In order to improve the design of rock bolts, it is necessary to have a good understanding
of the behavior of rock bolts in rock masses. Several analytical models of describing the interaction among
of the rock bolts, the grout medium and the rock mass have been proposed in literatures. In this paper, a
numerical simulation procedure was developed for a rock bolt subjected to a concentrated load in pullout tests.
The study was concentrated on the failure at the interface between the bolt and the grout. The commercial
code FLAC3D was used to analyze the coupling and decoupling phenomenon at the bolt-grout interface. The
interface provided by FLAC3D which is characterized by Coulomb sliding and/or tensile separation is proved
the premise for simulating the problem. On the basis of a numerical parametric study, the effects of dilation
angle and confining pressure on the behavior of the interface were clarified and the failure mechanism on the
interface was identified.

1 INSTRUCTIONS 2 OVERVIEW OF PREVIOUS WORK

Rock bolts have been widely used to reinforce rock When a fully grouted bolt is subjected to a tension
slopes, hydro dams and underground works such as load, the shear stress along the bolt before decoupling
tunnels and mine workings for a long time. Several was expressed by Hawkers & Evansas (1951) as:
analytical methods are proposed for bolting design
(Farmer, 1975; Freeman, 1978; Li & Stillborg 1999;
τx = τ0 e− d
Ax
Cai et al. 2004). However, the interaction mecha- (1)
nism of the rock bolt and the rock mass is not well
understood, and the bolting design is still empirical where τ0 is the shear stress at the loading end; d is the
so far. diameter of the bolt; A is a material parameter which
Pullout tests are currently used to examine the describes the interaction properties of the rock bolt and
anchoring capacity of rock bolts. Unlike bolts in situ, grout medium.
bolts in a pullout test only have an anchor length. The Equation (1) can describe the exponentially
Therefore the anchorage mechanism of bolts and rock attenuation of the shear stress at the bolt interface,
masses can be simply investigated through the study on but it is too simple to represent the effect of the grout
the pullout test. When a fully grouted bolt is subjected and surrounding rock mass on the stress distribution.
to a tension load, failure may occur at the bolt-grout In Farmer’s work (1975), the attenuation of the shear
interface, in the grout medium or at the grout-rock stress was expressed as:
interface, depending on which one is the weakest (Li &
Stillborg 1999). In this study we concentrate on the α −2α dx
failure at the interface between the bolt and the grout. τb = σb0 e b (2)
2
To investigate the coupling and decoupling behavior
of the rock bolt in pullout tests, numerical analysis has
where
been carried out using the explicit finite difference
code FLAC3D for better understanding the interface 2Gr Gg
bonding failure phenomena (FLAC3D, 2003). From a α2 =     
dg
series of numerical experiments, it is found that the Eb Gr ln db + Gg ln ddg0
dilation angle of the interface and confining pressure
on the bolt have significant effects on the coupling and Er Eg
Gr = , Gg =

decoupling of the interface. 2 (1 + νr ) 2 1 + νg

809
where σb0 is the axial stress of the bolt at the loading this process. The shear strength after interface decou-
point, Eb is Young’s modulus of the bolt steel, Er is pling is called the residual shear strength in this paper.
Young’s modulus of the rock mass, Eg is Young’s mod- Based on experimental results, Li & Stillborg (1999)
ulus of the grout, νr is Poisson’s ratio of the rock mass, proposed a model for the shear stress along a fully
νg is Poisson’s ratio of the grout, dg is the diameter of grouted bolt illustrated in Figure 2. This model con-
the borehole, and d0 is the diameter of a circle in the cludes the decoupling along the interface. The stressed
rock outside which the influence of the bolt disappears. in different sections of the bolt is described as follows:
Cai et al. (2004) proposed a model based on the
improved shear-lag theory, in which the shear stress τb (x) = 0 when 0 ≤ x < x0 (4)
distribution along the bolt before decoupling was
τb (x) = sr when x0 ≤ x < x1
expressed as:
x − x1
τb (x) = ωsp + (1 − ω)sp when x1 ≤ x< x2
τb (x) = τ0 cosh[α(L − x)]/ cosh(αL) (3)
(x−x2 )
−2α
where τb (x) = sp e db
when x > x2

α = H (1/(Ab Eb ) + 1/(Eb Ab )) where = x2 /x1 , and ω = sr /sp , the ratio of the
residual shear strength to the peak shear strength; x1
where τ0 is the shear stress at the loading end; L is and x2 are shown in Figure 2.
the length of the bolt. H is a material parameter which Numerical experimental results revealed that a con-
describes the interaction properties of the rock bolt, fining pressure influences the strength of the interface
grout and the surrounding rock mass. Eb and Ab are dramatically (Moosavi et al. 2005). So Cai et al. (2004)
the Young’s modulus and cross-sectional area of the proposed a Mohr-Coulomb law to describe the decou-
bolt. Fig. 1 demonstrates the constitutive law of the pling behavior of the rock bolt and the rock mass. The
interface media used by Cai et al. shear strength of the interface is expressed as
In general, the shear strength of an interface in
coupling stage comprises three components: adhe- τm = c + σnb tan ϕ (5)
sion/cohesion, mechanical interlock and friction. At
this stage, the deformation of the bolt and the sur- where ϕ and c are the friction angle and cohesion of
rounding rock mass is compatible along the interface. the interface, σnb is the normal stress perpendicular
If the interface medium is ruptured, slippage may to the rock bolt. The constitutive law of the interface
take place, which is termed the decoupling behavior. media is demonstrated in Fig. 2, where umax is the
The shear strength of the interface decreases during ultimate coupling shear displacement of the interface.
The parameters umax can be obtained by experiment
or by analytical calculation from τm and kini .
Shear strength
3 NUMERICAL MODELING PROCEDURE
friction bolt
m 3.1 Case study
This paper deals with the numerical study of interac-
tion mechanism of rock bolts and the grout medium
at the interface. A pullout test conducted by Stillborg
Shear strength

grouted bolt = Sp
mo b

kini b
= Sr (x)
b

=0
b

umax displacement x0 x1 x2 x

Figure 1. Distribution of shear stress along a fully grouted Figure 2. Relationship of the shear stress and displacement
rock bolt subjected to an axial load (Cai et al. 2004). at rock bolt interface (Li & Stillborg, 1999).

810
(1994) is considered in this study. In his test, a 3 m long E
rebar with a diameter of 20 mm was grouted within K= (6)
3(1 − 2ν)
two identical concrete blocks. The length of the bolt
in each block was 1.5 m. One block was fixed to the E
G= (7)
ground and the other was pulled. The bolt was pulled 2(1 + ν)
out without rupture, indicating that decoupling failure
of the interface occurred along the entire length of the 3.2 Interface modeling
bolt. The maximum pullout load was 180 kN. Other
necessary parameters are indicated in Table 1. In the case of a rough bolt, modeling the interface
This problem has many parameters: bolt and bore- between the grout and the bolt is invariably an inte-
hole geometry, material properties of bolt, grout and gral part of the analysis. In the case of bolt—grout
rock mass, confining pressure and interface proper- interaction, the interface is considered stiff compared
ties etc. The objective of this work is not to consider to the surrounding grout medium, but it can slip and
the influence of all these parameters but to check if a may be open in response to the loading. Joints with
numerical analysis using the finite difference approach zero thickness are more suitable for simulating the
can describe correctly the coupling and decoupling frictional behavior at the interface between the bolt
mechanisms observed in pullout tests. Therefore the and the grout medium.
problem is focused on the shear stress and axial load The interface model shown in Figure 3 has been
distribution along the bolt in coupling and decoupling used to simulate the grout/bolt contact described by
stages. In addition, the effect of confining pressure and Coulomb law. The logic contact for either side of the
interface dilation on the shear strength of the interface interface is similar in nature to the interface used in
is studied. the distinct element method.
The analysis is carried out using the com- The spring in the tangential direction and the slider
puter code FLAC3D (Fast Lagrangian Analysis of (Figure 3) represent the Coulomb shear-strength crite-
Continua) which is a commercially available finite rion. The spring in the normal direction and the limit
difference explicit program. strength represent the normal contact. The interface
The bolt, grout and surrounding rock mass (which has a cohesion c, a friction angle ϕ, a dilation angle δ,
is replaced by concrete block in pullout test) are all a normal stiffness Kn , a shear stiffness Ks and a tensile
modeled by the elastic model encoded in FLAC3D strength T . The value of interface properties is given
code. The elastic bulk modulus K and shear modulus in Table 1.
G which are input material properties in numerical
analysis can be calculated from Young’s modulus E
3.3 Mesh modeling and boundary conditions
and Poisson’s ratio ν by the following equations:
In general, fine meshes give more precise results in
finite element or finite difference analysis than coarse
meshes. But too fine meshes will cause dramatically
prolonged cycling time without evident improvement
Table 1. Material properties and geometries adopted in on the results. Figure 4 shows the mesh and bound-
numerical analysis. ary conditions retained for this analysis. The model
Material Properties
is constructed based on the pullout test referred to in

Rock Elastic modulus (GPa) 45

Poisson’s ratio 0.25 target face
Grout medium Elastic modulus (GPa) 30
Poisson’s ratio 0.25 SS S kS
Thickness (mm) 7.5
Steel bolt Elastic modulus (GPa) 210
Poisson’s ratio 0.3 TS
Length (m) 0.5
Diameter (mm) 20 S = slider
Ts = tensile strength
Interface Shear stiffness (GPa) 100 Ss = shear strength D
Normal stiffness (GPa) 100 D = dilation
Poisson’s ratio 0.25 ks = shear stiffness
Cohesion (MPa) 13 kn = normal stiffness kn
Friction angle (◦ ) 30 P = interface node
Dilation angle (◦ ) 0, 6, 10 P
Decoupled cohesion (MPa) 0
Confining pressure (MPa) 3, 6
Figure 3. Interface element used (FLAC3D, 2003).

811
 on the bolt end. Then, the system is cycled to
equilibrium again.

0.175 m
To simulate the decoupling between the bolt and the
grout medium, the cohesion of the interface is set to
pull 0.5 m
zero when the shear displacement exceed umax which
0.6 m can be calculated by relation shown in Fig. 1.

4 RESULTS AND DISCUSSION

Numerical studies are performed for different dila-

tion angles and confining pressures and the results
of fully coupling and partially decoupling for dila-
0.175 m
tion angle = 6.0◦ , confining pressure = 6.0 MPa
are presented. Figure 5(a) shows the axial displace-
Figure 4. Mesh used and boundary conditions. ment contour of the system in fully coupling state, and
it indicates clearly that the displacement of the bolt
and the grout medium is consistent at the interface.
3.1, but the length of the bolt is shortened to 0.5 m. Figure 5(b) shows a clear decoupling section in blue.
This can save cycling time without effects on the anal- Figure 6 present the shear strain increment contour of
ysis results. The mesh size is fine near the bolt where the system and the decoupling mechanism can also be
deformations and stress gradients are concentrated. In shown from the discontinuity. Figure 7 shows the shear
order to minimize boundary effects, the widths of rock
mass mesh are set to five bolt diameter in radius and
at the far end of the bolt. As a general rule for the
boundary conditions, the top, back and right surfaces
are assumed to be fixed, the front and bottom sur-
faces are assumed to be fixed in horizontal and vertical
directions respectively and the left surface is free. (a) σ0=100MPa

3.4 Simulation procedure with FISH routine

The internal routine of FLAC3D code (FISH) was
developed to control the calculation in cycling pro- (b) σ0=160MPa
cess and to record the results for plotting and charting.
The shear stress, normal stress and shear displacement Figure 5. The displacement contour for (a) fully coupling
on the interface can all be read out from the memory state and (b) partially decoupling state.
by some FISH routines. In addition, FISH functions
may be called from several places in the FLAC3D pro-
gram while it is executing. By a subroutine developed
with FISH language, the properties of the interface
can be automatically changed when the shear displace-
ment reach a special value. Therefore, the constitutive
law of the interface media shown in Fig. 1 can be
achieved in numerical analysis. To approximate the (a) σ0=100MPa
actual load conditions on the interface, the following
two simulation procedure steps are adopted:

1. The boundary conditions of the two longitudinal

sections of the bolt are set to be free. Both of the
section surfaces are subjected to a homogeneous
pressure to simulate the effect of confining pres-
sure on the bolt. Then, the system is cycled to (b) σ0=160Mpa
equilibrium.
2. The two longitudinal sections of the bolt are fixed Figure 6. Shear strain increment contours of (a) fully
in radial direction and a pull pressure σ0 is applied coupling state and (b) partially decoupling state.

812
 Shear stress (MPa) Shear stress (MPa)
12
8 Dilation angel = 10°
10

6
8
(a)
(b) 6
4
Dilation angel = 6°
4
2
2

0 0
0.0 0.1 0.2 0.3 0.4 0.5 0.0 0.1 0.2 0.3 0.4 0.5
Distance from borehole collar (m) Distance from borehole collar (m)

Figure 7. Shear stress distribution along the bolt for (a) Figure 8. Distribution of shear stress along the bolt interface
σ0 = 100 MPa, (b) σ0 = 160 MPa. for dilation angle = 66◦ and 10◦ .

Table 2. Pull pressures σ0 (MPa) for the initial

decoupling at the interface. 5 CONCLUSIONS
Confining pressure (MPa)
A numerical simulation procedure is proposed to
describe the interaction mechanism between bolt and
Dilation angle (◦ ) 3 6
grout medium for fully grouted rock bolt. In spite
0 53 102 of the simplicity of the procedure, the coupling and
6 53 102 partially decoupling states at the interface of the bolt
10 53 102 have been obtained. Specially, different dilation angles
and confining pressures have been checked in the
paper. The results of these simulations have shown
the following:
stress distribution along the bolt. It can be seen that the
curves for two different pull pressures have different 1. The shear stress decrease exponentially from the
forms because decoupling has occurred on the inter- point of loading or decoupling front to the far end
face of bolt-rock in curve (b). It is clearly indicated of the bolt in coupling section.
that the shear stress decrease exponentially from the 2. Dilation angle has no effect on the shear stress
point of loading or decoupling front to the far end of distribution before decoupling and will not affect
the bolt in coupling section. the bonding capability, but it has significant effect
Table 2 shows the pull pressures of the initial decou- on the anchorage in decoupling section. More-
pling occur at the interface for different dilation angles over, greater dilation angle can mobilize greater
and confining pressures. It can be shown from Table 2 shear strength of the interface and lead to a shorter
that (1) the initial decoupling load of the bolt is inde- decoupling length at same pull load.
pendent of the dilation angle; and (2) increasing of 3. An increased confining pressure can effectively
confining pressure can add to the coupling capability improve the bonding capability and the load bearing
of the interface. The explanation for this phenomenon capability of the bolt.
may lies in that the dilation only takes effect in the
decoupling section in which the shear dilation occurs Finally, the obtained results validate the simula-
with the shear displacement. Fig. 8 shows the shear tion procedure using FLAC3D code to simulate the
stress distributions of the bolt-grout interface for dila- interaction between bolt and grout of fully grouted
tion angle = 6◦ and 10◦ at the same pull pressure rock bolts. Such approaches can thus be used to
σ0 = 160 MPa. It indicate that greater dilation angle analyse the more complex failure mechanism of
can mobilize greater shear strength of the interface and grouted bolt, such as failure in grout medium and
lead to a shorter decoupling length at same pull load. rock mass.

813
ACKNOWLEDGEMENTS Freeman, T.J. 1978. The behaviour of fully-bonded rock
bolts in the Kielder experimental tunnel. Tunnels and
The study presented is supported by the open fund Tunnelling June: 37–40.
of Key laboratory of Mountain Hazards and Surface Li, C. & Stillborg, B. 1999. Analytical models for rock bolts.
Process, Chinese Academy of Science and the China Int J Rock Mech Min Sci Geomech Abstr 36: 1013–1029.
Stillborg, B. 1994. Professional users handbook for rock
National Natural Science No. 40572158. bolting, 2nded. Trans. Germany: Trans Tech Publications.
FLAC3D, 2003. Fast Lagrangian analysis of continua,
version 2.1. Itasca Consulting Group.
REFERENCES Moosavi, M. et al. 2005. Bond of cement grouted reinforcing
bars under constant radial pressure. Cement & concrete
Cai, Y. et al. 2004. A rock bolt and rock mass interaction composites 27: 103–109.
model. Int J Rock Mech Min Sci Geomech Abstr 41: Hawkes, J.M. & Evans, R.H. 1951. Bond stresses in rein-
1055–1067. forced concrete columns and beams. Journal of the
Farmer, I.W. 1975. Stress distribution along a resin grouted Institute of Structural Engineers 24(10): 323–327.
rock anchor. Int J Rock Mech Min Sci Geomech Abstr 12:
347–351.

814
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Macroscopic effects of rock slopes before and after grouting of joint planes

Hang Lin, Ping Cao, Jiang-Teng Li & Xue-liang Jiang

School of Resources & Safety Engineering, Central South University, Changsha, China

ABSTRACT: Studies on slope reinforcement methods are important for geotechnical engineering. All of
the reinforcement methods can be divided into two groups, the passive reinforcement methods and the active
reinforcement method. Grouting is an active reinforcement method. It is to inject a cement grout mixture into
the rock mass to reduce the permeability of the rock mass and improve the mechanical properties of the rock
mass itself. In order to find differences of parameters between the rock mass before and after grouting, the
comparative experiments of direct shear were done that indicates that the shear strength and shear stiffness of
rock samples are improved greatly after grouting. However, no the whole slope must be grouted. The effective
grouting areas (EGA) can be defined by relative calculations. The three dimensional explicit finite difference
code, FLAC3D , was adopted to study the differences of mechanical properties of the rock mass before and after
grouting. Comparative analysis shows that stress field and displacement field of the rock mass become more
symmetrical and continuous after grouting. The tensile stress area reduces and the displacement bifurcation near
joint planes vanishes, both are good to the stability of slope.

1 INTRODUCTION analysis for the geotechnical engineering (Grgic et al.

2003, Exadaktylos et al. 2002, Guo et al. 2002, Boidy
Cement grouting constitutes one of the major tech- et al. 2002, Waltham et al. 2004).
niques used for stabilization in civil engineering. It
is widely applied in tunnel engineering (Hoek, 2000,
Eberhardt, 2002, Varol et al. 2006, Yesilnacar, 2003,
Turkmen et al. 2003). It consists in injecting into the 2 TEST
rock mass a cement grout mixture under controlled
pressures and volumes. The main results expected Joint plane, if loaded in two directions with normal
from this process are a reduction in permeability and an stress and shear stress, will arise one or several dif-
improvement in terms of mechanical properties. But it ferent directions of deformation, which exhibits as
is seldom applied in slope engineering, with the reason the closure and slippage deformation. In order to
that the mechanism of grouting is still not well known. describe the discipline of deformation, we can use
Xu (2006) has studied the effects of grouting to rock stress-displacement relationship curves, whose slope
mass experimentally. In his studies, grouting is done can reflect the shear stiffness Kss and shear stiffness
in all the joint planes in samples, but slope is large in of rock mass. The stress-displacement relationship is
size, it is not realistic or practical to grout in all the described with their increment,
areas. So it is necessary to decide which joint planes
should be grouted and which should not. In present dσij = K · dδij (1)
paper, firstly we try to find out the effective grouting
area (EGA), then study the macroscopic effect such where K is the stiffness matrix.
as stress and deformation of rock jointed slope before Substituting σij with τs and σn , δij with δs and δn ,
and after grouting by the numerical calculation soft- we can obtain,
ware FLAC3D . Numerical calculation results are inte-     
grals of reflections in small units, whose behavior is dτs Kss Ksn dδs
= (2)
exactly based on the classical mechanical theory, so we dσn Kns Knn dδn
think the numerical analysis is a good way, at least
better than the theory deduction based on the simple where τs is shear stress; σn is normal stress; δs is shear
models, to study the effect of huge body, just as long displacement; δn is normal displacement; Kss is shear
as the simulation models are close to the reality situa- stiffness coefficient, which indicates the influence of
tion. And these methods are recently widely applied in δs to τs ; Ksn indicates the influence of σn to τs ; Kns is

815
dilation stiffness coefficient, which indicates the influ-
ence of δs to σn ; Knn is the normal stiffness coefficient, 1.0 = 0.8Mpa
which indicates the influence of δn to σn . = 0.6Mpa
0.8
For macroscopic body, effect of normal displace-

s (MPa)
ment to shear stress can be ignored (Xia, 2002), which
means Ksn = 0, then eq. (1) can be evolved to, 0.6

Shear stress
dτs = Kss dδs (3) 0.4

= 0.2Mpa
0.2
and average shear stiffness is,
= 0.4Mpa
 7 0.0

Kss = dτs dδs (4) 0 5 10 15 20 25

δ δ
Shear displacement s ( mm)

2.1 Materials
Figure 2. Shear stress and shear displacement after
Rock samples are uniform in the size of 25 cm × grouting.
20 cm × 10 cm. In the tests, M20 cement mortar is
chosen to simulate rock, while M7.5 for joint plane, peak value. Under combined loads of compressive
whose strength is lower than rock block. Thickness of stress and shear stress, joint plane becomes smooth
grouting is 20 mm, weight proportioning of cement during slipping, the fluctuation value of peak strength
and sand is 1:4.74. reduces, and the strength of joint plane comes to the
residual strength phase. With the increasing of normal
stress, the peak shear resistance stress increases. Shear
2.2 Results and discussion resisting strengths of the four samples are 0.22 MPa,
Normal stress changes from 0.2 MPa to 0.8 MPa in the 0.33 MPa, 0.52 MPa and 0.64 MPa for the normal
test. The curves of relationship between shear stress stress of 0.2 MPa, 0.4 MPa, 0.6 MPa, and 0.8 MPa,
and shear displacement for rock sample with and with- respectively.
out grouting in joint plane are showed in Figure 1 and Relationship of shear stress and shear displace-
Figure 2. ment for samples after grouting is shown in Figure 2.
Figure 1 shows that, each curve slope changes in The linear elastic states are more obvious and curves
the phase prior to peak stress, which indicates that become steeper when compared with the curves of
the stiffness Kss changes at the same time. When rock samples without grouting in Figure 1. After curves
sample begins to slip, value of Kss in is small. During reach peak of stress, they decrease quickly which
the procedure of slippage, the protuberant fractions is very different from that of Figure 1, and exhibits
meet closer, and the phenomenon of shear dilation the stiffness characteristic. Shear resistance stress for
occurs, which in turn causes shear stress reaching different normal stresses are 0.54 MPa, 0.66 MPa,
0.89 MPa and 0.99 MPa, respectively, which are larger
in magnitude than that before grouting. After grout-
0.7 ing, average stiffness increases and joint plane changes
0.6
from frictional joint plane to stiffness joint plane.
= 0.8Mpa
0.5
s (MPa)

3 EFFECTIVE GROUTING AREA

0.4
= 0.6Mpa
It is not practical and realistic to grout in all the joint
Shear stress

0.3
= 0.4Mpa planes of slope, so it is necessary to find the effective
0.2 grouting areas (EGA). We make the study by build-
= 0.2Mpa
ing the calculation model, showed in Figure 3, with
0.1 assumption that the slope face strikes parallel to the
underlying exfoliation surface and hence the slope can
0.0
be analyzed by means of a two dimensional model.
0 5 10 15 20 25 Natural angle is β2 , thickness of slope is d, dip angle
Shear displacement s (mm) of excavated slope is α, joint plane inclination is β1 ,
distance between Bi and C is Li . The stability of slope
Figure 1. Shear stress and shear displacement before is controlled by joint plane. Studies are done for the
grouting. situation of α > β1 , and makes the assumption that

816
 A Bi 2cj
2
Lcr = (11)
E C γ sin(β1 − β2 )(sin β1 − cos β1 tan φj )
F Joint plane
Supposing that |AC| = L0 , the following statements
can be obtained,
Di
a. if Lcr > L0 , slope is in stable state;
O b. if Lcr = L0 , slope is in critical stability state;
Rock mass 1 c. if Lcr < L0 , slope is in unstable state.

Then the EGA can be obtained from Lcr to L0 . After

grouted, c and φ of rock mass increase, and Lcr also
increase according to eq. (11). If Lcr reaches the con-
Figure 3. Theory deduction model.
dition of (a), slippage along joint plane is controlled,
which means that the slope is in stable state.

failure mode of slope is shear strength failure with

unstable block slipping along joint plane. 4 NUMERICAL SIMULATION
For plane Bi Di , the safety factor of slope is calcu-
lated as follow, 4.1 Simulation technique

S BD Fi Macroscopic effect of rock slope is simulated by three

Fs = = ki (5) dimensions explicit finite difference code, FLAC3D .
τ BD Fx For given element shape functions, the set of alge-
braic equations solved by FLAC3D is identical to that
where Fki is resisting force of block Bi CDi , Fxi is solved with the finite element method. However, in
driving force of block Bi CDi . FLAC3D , equations are solved using dynamic relax-
Relationships of parameters are, ation (Dawson et al.1999), an explicit, time-marching
procedure in which the full dynamic equations of
Li sin(β1 − β2 ) motion are integrated step by step. Static solutions
|Di C| = (6) are obtained by including damping terms that grad-
sin(α − β1 )
ually remove kinetic energy from the system. The
1 convergence criterion of calculation is the unbal-
GBi CDi = γ · d|Bi C| · |Di C| sin(α − β2 ) (7)
2 ance force ratio which is defined by the ratio of
nodes’ average force to the maximum unbalance force
where GBi CDi is the gravity of Bi CDi , γ is the average reaches 10−7 .
unit weight of Bi CDi .
Then we can obtain,
4.2 Model
cj Li sin(α − β2 ) · d
Fki = GBi CDi cos β1 tan φj + (8) Models for jointed slope are divided into two groups:
sin(α − β1 )
(1) before grouting; (2) after grouting. Three joint
Fxi = GBi CDi sin β1 (9) planes are preset in the slope. The elasto-platic block
element with lower strength is chosen to simulate joint
where φj is friction angle of joint plane, cj is cohesion plane, and the inclination of joint plane is 40◦ with
of joint plane. 0.1 meters thickness. Plane strain model is built with
Substituting Fki and Fxi in eqs. (5) with eqs. (8) and 9 795 triangular elements and 3 293 grids, shown in
(9), yields, Figure 4. Size of slope are shown in Figure 5, and
rock parameters are shown in Table 1. Natural angle
GBi CDi cos β1 tan φj +
cj Li sin(α−β2 ) is 10◦ , slope angle after excavation is 75◦ . Gravity
sin(α−β1 ) load is performed as an initial free mechanical state.
Fs = (10)
GBi CDi sin β1 Critical slippage length can be obtained by eq. (11).
Then we can obtain, Lcr = 12.71 m, L0 = 54.26 m,
If Fs = 1, block Bi CDi is in the critical state L1 = 9.51 m and L2 = 29.09 m. It is obvious that
of stability, then the critical slippage length can be Lcr < L2 < L0 , so the second and third joint planes
calculated as follow, according to eq. (10), are in EGA.

817
 19.3m 19.3m 9.3m
32.1m 70m

K04
First joint plane
K03
K02 K01
P01 Nature slope direction

20m
P02
P03 K05
P04 Second
P05
Third joint P06 K06 joint plane

32.2m
P07
plane

104.8m
P08
P09 K07
K08 P10
50m

45m
Rock mass

150m

Figure 4. Numerical model.

Figure 5. Monitoring points.

4.3 Monitoring points

Failure of rock mass is close related to its inner struc- displacements at the same step reduce gradually
ture plane and stress field. Perturbation induced by from surface to the inner place in sequence of
excavation can cause deformation of rock mass, which K01 → K02 → K03 → K04. At their con-
is nonlinear mechanical procedure. In order to reflect vergence phases, the peak values of displacement
the nonlinear procedure of displacement in different as well as curve’s slope reduce in the same sequ-
time steps, monitoring points from K01 to K08 are ence. For points on the surface of slope, values of
located in slope. These displacements change with displacement as well as curve’s slope reduce in the
time steps, and we call them dynamic displacement. sequence of K05 → K06 → K07 → K08. Among
After excavation, the displacements of particles will all the monitoring points, K01 has the largest magni-
converge to certain values, then we call them static tude of displacement, so it is suitable to choose K01
displacements. Monitoring points from P01 to P10 are as the characteristic point for deformation monitoring.
located in the slope body to record the static displace- Besides, displacement of K08 stays at the same value
ments of rock particles. All the monitoring points are in most part of calculation procedure, with the rea-
shown in Figure 5. son that it is under the joint plane and little affected
by slippage of block along joint plane. After grouting,
magnitudes of displacements reduce greatly, which is
4.4 Results and discussion
because grouting strengthens the joint plane. For the
Static total displacements of slope before and after monitoring point on the top of slope, displacements
grouting are shown in Figure 6. Before grouting, dis- reduce in the sequence K01 → K02 → K03 → K04.
placements of monitoring points are divided into three For monitoring points on the surface of slope, val-
groups (three bifurcation groups along monitoring ues of displacement increase in sequence of K06 →
lines), but the magnitudes of bifurcation are small. K07 → K08, which is different from the slope before
The bifurcation occurs at the places of joint planes. grouting. This is because rock mass become uniform
Values of displacement reduce gradually from surface and continuous after grouting. But the displacement
of slope to inner place. After grouting, monitoring dis- of K08 remains the same after grouting when com-
placements in each monitoring line are different from pared to that before grouting, which results from that
that before grouting. In the same monitoring line, dis- K08 is at the toe of slope, and the toe of slope remain
placements of each monitoring line (except line P01 stable before grouting, so grouting done in joint plane
and line P02) change fluently without mutation phe- affect little the displacement of K08. After grouting
nomenon. The displacements reduce gradually from joint planes change from weakness plane to stiffness
surface of slope to inner place, which is the same plane with the increasing of vertical stress. And we can
as that before grouting. But displacements of some also see that, the slope of each curve becomes steeper
monitoring lines (line P01 and line P02) still remain than that before grouting, which indicates the stiffness
bifurcation, which is because the first joint plane has improvement.
not been grouted. Figure 8 shows the maximum principal stress of
Figure 7 shows the relationships of dynamic total slope before and after grouting. Maximum principal
displacements and calculation steps. For monitoring stress is an important factor to determine the state of
points located on the top of slope, the magnitudes of element. According to Mohr-Coulomb criterion, when

818
 Table 1. Calculation parameters.

Unit Elastic Friction Tensile

weight modulus Poissons’ Cohesion angle strength
Material (kN/m3 ) (GPa) ratio (kPa) (◦ ) (MPa)

Rock 25 3.0 0.20 500 35 1.00

mass
Joint 17 0.01 0.30 120 20 0.05
plane
Grouting 20 0.1 0.20 240 23 1.00

60 60

50 Before grouting 50

Total displacement/mm
Total displacement/mm

p01 40 K01
40
p02 K02
p03 K03
30 p04 30 Before grouting K04
p05 K05
p06 20 K06
20
p07 K07
p08 K08
10 p09 10
p10
0 0

0 10 20 30 40 50 0 20 40 60 80 100 120
Along nature slope direction/m Step/10 4
(a) before grouting (a) before grouting
8 8

7 7
p01
6 p02 6
Total displacement/mm

After grouting
Total displacement/mm

p03 K01
p04 5
5 K02
p05 K03
4
4 p06 K04
p07 K05
3
3 p08 K06
p09 K07
2
2 p10 K08
1
1
0
0 0 5 10 15 20 25 30
0 10 20 30 40 50
Step/10 3
Along natural slope direction/m
(b) after grouting
(b) after grouting

Figure 7. Dynamic total displacement of slope.

Figure 6. Static total displacement of slope.

places near joint plane, the stress is tensile and concen-

certain principal stress states of a particle have reached trated, in turn, it causes hidden factor for the instability
the failure envelope, the unbalance force of this par- of slope because rock mass can not resist great ten-
ticle can then be transferred into adjacent particles, sile stress. After grouting, physical parameters of joint
which contribute to the increase of its unbalance force, plane are improved, and the shear resisting strength and
in turn, the sliding mechanism may be initiated. If shear resisting stiffness increase, which result in stress
the failure points connect to form a sliding plane, the fields in slope becoming more continuous. The mag-
unstable rock mass will slip. After excavation, most nitudes of stress near joint plane become smaller, with
parts of slope are in stable state, the maximum prin- the phenomenon of bifurcation in joint plane reduced,
cipal stress in these places is compressive. But at the and stability, uniformity of slope improved.

819
 after grouting, displacement field and stress field
become smooth, large magnitudes of deformations
near joint plane are controlled. At the same time,
the tensile crannies on the top of slope reduce,
which is good to the stability of slope.

REFERENCES

A.C. Waltham & G.M. Swift. 2004. Bearing capacity of rock

over mined cavities in Nottingham. Engineering Geology
(75):15–31.
Ahmet Varol & Sueyman Dalg. 2006. Grouting appli-
cations in the Istanbul metro, Turkey. Tunnelling and
Underground Space Technology (21):602–612.
Dragan Grgic, Francoise Homand & Dashnor Hoxha. 2003.
A short- and long-term rheological model to understand
the collapses of iron mines in Lorraine. France.Computer
and Geotechnics (30):557–570.
E. Boidy, A. Bouvard & F. Pellet. 2002. Back analysis
of time-dependent behaviour of a test gallery in clay-
stone. Tunnelling and Underground Space Technology
(17):415–424.
E.M. Dawson, W.H. Roth & A. Drescher. 1999. Slope Sta-
bility analysis by strength reduction. Geotechnique 49
(6):835–840.
Erik Eberhardt. 2002. Rock Slope Stability Analysis Uti-
Figure 8. The maximum principal stress. lization of Advanced Numerical Techniques. ETH zurich,
Switzerland: Swiss Federal Institute of Technology.
Evert Hoek. 2000. Rock Engineering. North Vancouver:
5 CONCLUSION Evert Hoek Consulting Engineering Inc.
G.E. Exadaktylos & M.C. Stavropoulou. 2002. Aclosed-form
elastic solution for stresses and displacements around tun-
1. Direct shear tests for rock samples before and after nels. International Journal of Rock Mechanics & Mining
grouting are done, results of tests show that, before Sciences (39):905–916.
grouting, joint plane belongs to frictional joint M.I. Yesilnacar. 2003. Grouting applications in the Sanli-
plane, its shear mechanical characteristic is influ- urfa tunnels of GAP, Turkey. Tunnelling and Underground
enced by the waviness degree and occlusive degree, Space Technology (18):321–330.
and the residual stress descends little compared to R. Guo & P. Thompson. 2002. Influences of changes in
peak stress; After grouting, strength and stiffness mechanical properties of an overcored sample on the far-
of rock sample increase. field stress calculation. International Journal of Rock
2. Theory models are founded for the slopes with its Mechanics & Mining Sciences (39):1153–1166.
Sedat Turkmen & Nuri Ozguzel. 2003. Grouting a tunnel
angle α greater than joint plane angleβ, critical slip- cave-in from the surface a case study on Kurtkula irriga-
page length Lcr for joint slope is deducted, then the tion tunnel, Turkey. Tunnelling and Underground Space
effective grouting areas are conformed. Technology (18):365–375.
3. Numerical simulation models are founded for exca- Xia Cai-chu & Sun Zong-qi. 2002. The Mechanics of Engi-
vated slope before and after grouting by FLAC3D , neering Joint Rock Mass. Shanghai: Tongji university
results of calculations show that, for slope before Press (in Chinese).
grouting, displacements fields of rock mass redis- Xu Wan-zhong, Cao Ping, Peng Zhen-bin, et al. 2006. Anal-
tribute during the excavation procedure, and there ysis of Mechanical Mechanism of Shear Characteristic
are many tensile areas near joint planes which lead of Bolted Layered Slope Considering Grouting Factor of
Weak Plane. Chinese Journal of Rock Mechanics and
to the tensile crannies in slope. The magnitude of Engineering 25 (7):1475–1480 (in Chinese).
displacement and deformation are large near joint Xu Wan-zhong, Peng Zhen-bing, Hu Yi-fu, et al. 2006. Simu-
plane. The bifurcations of displacements happen lation Experiment Study on The Treatment of Bolting and
when the monitoring lines passing joint planes, Grouting in Rock Slope Reinforcement. China Railway
stress distributions are asymmetry in rock mass, Science 27 (4):11–16 (in Chinese).
which influence the stability of slope; For slope

820
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Two- and three-dimensional analysis of a fossil landslide with FLAC

X.L. Liu
College of Environmental Science and Engineering, Ocean University of China, Qingdao, China

J.H. Deng
College of Water Resources and Hydropower, Sichuan University, Chengdu, China

ABSTRACT: Xietan landslide, a fossil landslide located in Zigui County, Hubei Province, China, is one of the
important geo-hazards at the northern bank of Yangtze River in the Three Gorges Reservoir areas. The landslide
is mainly composed of sliding mass, slip soil (the weak layer) and bedrock. Slip soil of the landslide, containing
coarse particles up to 40% (by weight), is very typical in the Three Gorges Reservoir areas. In this paper, first,
the shear strength parameters of the slip soils of Xietan landslide were analyzed by numerical computing. Then
with the FLAC program, the stress and deformation characteristics of the fossil landslide at the reservoir water
level of 139 m caused by impoundment were simulated using two- and three-dimensional model respectively.
According to comparisons of deformation between the monitoring data and the numerical computing data of
the two- and three-dimensional simulations, the followings can be concluded. (1) Shear strength properties of
slip soils have important influence on landslide stability analysis. The shear strength parameters of the slip soils
containing coarse particles of Xietan landslide used for numerical simulation are acceptable. (2) Results of the
numerical analysis of the three-dimensional model agree better with the actual condition of the landslide than
that of the two-dimensional model. (3) It is feasible to analyze landslides by numerical method, for example by
the FLAC program, provided that the reasonable parameters of the landslide for computing can be gained.

1 INTRODUCTION investigation of the shear strength properties of this

kind of slip soils is important for numerical analysis
Xietan landslide, a fossil landslide located in Zigui of the fossil landslide.
County, Hubei Province, China, may reactivate to In this paper, shear strength properties of slip
become one of the geo-hazards at the northern bank soils with coarse particles of Xietan landslide were
of Yangtze River due to impoundment of the Three investigated first. Then the stress and deformation
Gorges Reservoir. The main longitudinal section is characteristics of the landslide were analyzed accord-
about 800 meters long, and the averaged width of ing to the numerical computing results and the in situ
the transverse section is about 260 meters. The fos- monitoring data.
sil landslide is mainly composed of sliding mass, slip
soil (the weak layer) and bedrock. The sliding mass
is about 10 to 40 meters thick. Thickness of the slip
2 SHEAR STRENGTH PROPERTIES OF SLIP
soil ranges from about 0.5 to 3.0 meters. The whole
SOILS OF XIETAN LANDSLIDE
landslide bulks about 624 ten thousand cubic meters.
Permeability parameter of the sliding mass is 4 ×
2.1 Index properties of the slip soils
10−2 cm/s and that is 3 × 10−7 cm/s for the slip soil.
The slip soil can be regarded as the aquifuge in the The slip soil of Xietan landslide is mainly composed of
landslide. There is no consistent groundwater table in sandy/silty clay containing gravels of sub-angular and
the landslide. Perched water exists on the localized sub-rounded fragments of sandstone, mudstone and
aquifuge of the slip soils. shale, etc. Gravel content of the slip soil ranges from
Shear strength properties of slip soils have great 20% to 40% by weight, with 2–60 mm size. A typi-
influence on stability of landslides. Slip soil of Xietan cal particle size distribution (PSD) curve is plotted in
landslide is very typical in the Three Gorges Reser- Fig. 1 as curve 3.
voir areas, containing gravels up to 40% (by weight). The average specific gravity of the slip soil is 2.78.
In view of little knowledge about the shear strength Average density of the undisturbed slip soil is about
characteristics of slip soils containing coarse particles, 2237 kg/m3 . Natural water content of the slip soil

821
 Curve 2 Curve 1 once under the unconsolidated undrained condition.
100
90
The block sample was consolidated first under nor-
Percent finer by weight / %

80 mal stress, and then was sheared at the speed of

70 Curve 3 0.7 mm/min to remain undrained. The block sample
60 was sheared only one time corresponding to one differ-
50 ent normal stress. In the whole tests, the same block
40 sample was sheared for 4 times corresponding to 4
30 different consolidation normal stresses. From the test
20
program, we knew that before consolidation each time,
10
0
there had been a pre-existing sheared surface in the
100 10 1 0.1 0.01 0.001 block sample.
Grain size/mm Results of the total shear strength (the consolidated
undrained shear strength) tests of series one to three are
Figure 1. Particle size distribution curves of different series presented in Fig. 2, from which it was found that the
of tests. total shear strength parameters of d < 60 mm (d rep-
resents the gravel diameter) series were very similar
Table 1. Shear strength tests on slip soils of Xietan to that of the d < 20 mm series. Taking the exist-
landslide. ing sheared surface into account for the d < 60 mm
series of tests, it could be inferred that shear strength
Test series Maximum Coarse of the large undisturbed sample was greater than that
(gravel Type particle content
size) of shear test size (by weight)
of the small ones in the second series of tests, and
the shear strength of the second series of tests could
1 DST-slow (60 mm be regarded as the lower bound of the shear strength
(2–5 mm) box) 5 mm 8% of the in situ slip soils. Comparing the PSD curves of
Triaxial tests CU these three series, it can be expected that the increasing
(38 mm
diameter) 5 mm
2 Triaxial tests
450
(2–20 mm) CU/CD (101 mm
400
diameter) 20 mm 23%
350
Shear stress / kPa

3 Multi-reversal 300

(2–60 mm) LDST (500 mm 250

(d<60mm)= tan(16.67)+27.39
200 (d<5mm)= tan(14.88)+12.20
box) 60 mm 38%
150
100 d<5mm
d<20mm
50 d<60mm
0
0 200 400 600 800 1000 1200 1400
ranges from about 12% to 16%, resulting in saturation Normal stress / kPa

of the in situ slip soil ranges from 85% to 95%.

Figure 2. Total shear strength results of series tests of one
to three.
2.2 Shear strength properties of the slip soils
In order to investigate the shear strength properties of
the slip soils containing gravels of Xietan landslide, Table 2. Parameters of the landslide used for numerical
three series of tests were carried out, as indicated simulations.
in Table 1. Direct shear tests (DST), consolidated
Components of Slip Sliding
undrained/drained (CU/CD) triaxial tests, and the the landslide Bedrock soils mass
multi-reversal large direct shear tests (LDST) were
performed with specimens of different gravel con- Young’s modulus/
tent and size. Remolded specimens were used for GPa 25 0.03 0.08
series tests 1 and the undisturbed samples were used Poisson’s ratio 0.25 0.38 0.35
for series tests 2 to 3. The PSD curves of the first Drained cohesion/ — 29.29 Natural 10
to third series of tests have been plotted in Fig. 1, kPa Saturated 2
corresponding to curve 1 to 3 respectively. Drained frictional
angle/degree — 21.34 40
The third series of tests were non-standard multi-
Density/kg/m3
reversal direct shear tests on one block sample using a Natural 2670 2237 2070
large direct shear box (500 mm × 500 mm × 400 mm). Saturated 2670 2275 2250
Prior to tests, the block sample had been sheared

822
proportion of coarse particles and their sizes result in a
significant increase in shear strength of slip soils (Liu
et al. 2006).
According to the analysis above, it was acceptable
to consider the second set of test results as the shear
strength parameters of the slip soils used for numerical
simulations of Xietan landslide. The Young’s modulus
and the Poisson’s ratio were also estimated from the
CD triaxial tests in the second series, as presented in
Table 2.

3 NUMERICAL ANALYSIS OF XIETAN

Figure 4. 3D numerical model of Xietan landslide.
LANDSLIDE

3.1 Numerical models

The original Yangtze River water level is 76 m which Reservoir water Level
has little effect on analysis of the Xietan landslide.
Considering availability of the monitoring data, the Free surface
landslide at long term reservoir water level of 139 m Linearly increasing hw
due to impoundment of the Three Gorges Reservoir normal stress from
Landslide zero to h w
was simulated by numerical method.
The relative position of the water level line of
139 m and the location of the bore ZK1 for inclination
Constant normal stress h w
monitor can be referred to Fig. 3.
The landslide was simulated with FLAC (Fast
Lagrange Analysis of Continua), using small-strain Figure 5. Sketch for considering influence of the reservoir
water level.
mode. The numerical model can be referred to Fig. 3
for 2-dimensional (2D) and Fig. 4 for 3-dimensional
(3D) analysis. The main longitudinal profile of the
3D model was used to represent the 2D landslide, as material, together with the Mohr-Coulomb yield crite-
indicated in Fig. 3. rion and the non-associated flow rule with the dilation
Horizontal displacements were fixed for nodes angle of zero. Rock-base of the landslide was mod-
along the left and right boundaries (in 3D model, hori- eled as a linear elastic material. All the parameters
zontal displacements of nodes along the front and back used in the numerical model have been presented in
boundaries were also fixed) while both horizontal and Table 2.
vertical displacements were fixed along the bottom Influence of the reservoir water was considered by
boundary of the model. The sliding mass and the slip the way of Griffiths and Lane (1999). As shown in
soils were modeled as a linear elastic-perfectly plastic Fig. 5, with water pressures acting on the landslide
surface in normal directions, the pore pressure at a
point was estimated as the product of the unit weight of
water γw and the vertical distance of the point beneath
the free surface.

3.2 Numerical result analysis

A bore for inclination
monitor, ZK1 Stress state of Xietan landslide in the situation of 139 m
139 m reservoir water level for 2D and 3D model has been
shown in Fig. 6. Results of the main profile of the 3D
model were given to represent the 3D analysis ones.
Only elements near the toe of the landslide were given
for comparison. Most of the slip soil elements yielded
due to the shear stress. The stress state distribution
Figure 3. 2D numerical model of Xietan landslide (Main showed that the main anti-sliding zone was the toe of
longitudinal section). the landslide.

823
 Elastic elements 2006. Comparison of the two series of monitoring data
Plastic elements due to shear stress showed that deformation of ZK1 in the landslide had
been unchangeable after Nov., 5, 2005 at the reservoir
Plastic elements due to tensile stress water level of 139 m. So the monitoring data indicated
in Fig. 7 can be regarded as the long term deformation
of ZK1 at the reservoir water level of 139 m. In view
of the long term deformation of the landslide com-
puted with the FLAC program, the monitoring and
computing data can be compared.
According to horizontal displacements shown in
Fig. 7, we found that all the horizontal displacements
(a) Stress state distribution of the 2D model of the sliding mass were similar to that of the slip
soils, which implied that deformation of the slip soils
dominated the whole sliding trend of the landslide.
Therefore, determination of the shear strength prop-
erties of the slip soils is important for simulating the
landslide properly.
Figure 7 has obviously shown that computing
results of the 3D model agreed better with the monitor-
ing data than that of the 2D model, which implied that
the 3D analysis could reflect the actual characteristics
of the landslide. At the same time, good agreement of
the 3D computing data with the monitoring data also
(b) Stress state distribution on the main longitudinal
indicated that the used parameters of the slip soils with
profile of the 3D model gravels were reasonable for Xietan landslide numerical
simulation.
Figure 6. Stress state distribution of Xietan landslide at the
reservoir water level of 139 m.

40
4 CONCLUSIONS
Distance from the rockbed / m

35
30
Shear strength parameters of the slip soils containing
25 coarse particles of Xietan landslide were investigated
3D model 20 by a series of tests. Two- and three-dimensional
2D model 15 simulations of the landslide were performed by the
Monitoring data 1 10 FLAC program. Some conclusions can be inferred as
Monitoring data 2 5 follows.
0
9 8 7 6 5 4 3 2 1 0 1. Shear strength properties of the slip soils have
Horizontal displacement / cm important influence on stability analysis of land-
slides. The acceptable shear strength parameters
Figure 7. Comparison of the computing and monitor- of slip soils containing coarse particles of Xietan
ing data. landslide have been concluded by tests, which can
provide some experience for the similar landslides
A bore for inclination monitor, ZK1 in the land- analysis.
slide, has been indicated in Fig. 3. According to the 2. For numerical simulations of landslide, results
in situ monitoring data, the actual displacements of of 3D model agree better with the actual char-
the landslide at the location of ZK1 can be gained. acteristics of the landslide than that of the 2D
Monitoring and the 2D and 3D numerical computing model.
horizontal displacements of the landslide at location of 3. It is feasible to analyze landslides by numeri-
ZK1 for the reservoir water level of 139 m are indicated cal method, for example by the FLAC program,
in Fig. 7. Here, the monitoring data 1 was collected provided that the reasonable parameters of the
on Nov., 5, 2005 and the monitoring data 2 on Oct., 7, landslide can be gained.

824
ACKNOWLEDGEMENTS REFERENCES

Financial support from National Science Foundation Griffiths, D.V. & Lane, P.A. 1999. Slope stability analysis by
of China under grant 40702044 and the permission of finite elements. Geotechnique 49(3): 387–403.
using the FLAC program from Institute of Rock and Liu X.L., Loo, H. & Min, H. et al. 2006. Shear Strength of
Soil Mechanics, the Chinese Academy of Science are Slip Soils Containing Coarse Particles of Xietan Land-
slide. Geotechnical Special Publication 151: 142–149.
greatly appreciated.

825
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Application of the coupled thin-layer element in forecasting the behaviors

of landslide with weak intercalated layers

Y.L. Luo & H. Peng

School of Civil and Architectural Engineering, Wuhan University, Wuhan, China

ABSTRACT: The distribution and seepage-stress coupled interaction of Weak Intercalated Layers (WIL) in
landslide are key factors resulting in the landslide disaster. Forecasting the behaviors of landslide with the
WIL and mitigating the disaster maximumly are very important. But the studies on the coupled interaction of
WIL are seldom reported. So this paper proposed a Coupled Thin-Layer Element (CTE) based on the original
Goodman’s joint to model both the stress-deformation behaviour and water pressure response of the WIL. An
elastic viscoplastic constitutive law was adopted to model the creeping behavior and then a FE program of CTE
was developed to analyze the Dayeping landslide. The calculated groundwater free surface is consistent well
with the in-situ testing data. The distributions of displacements and stresses suggest that the landslide may slide
along the WIL at the elevation of 230 m to 370 m. In the end, the strength reduction method was used to calculate
the stabilities for CTE, Solid Element (SE) and Desai Element (DE) used to model the WIL. The results indicate
that SE and DE can’t simulate the tension fracture damage or pore pressure, although the minimum safety factors
(Fmin ) obtained, 1.22 and 1.27, are acceptable. CTE makes up the above faults. It can comprehensively model
the true status of stresses in the landslide and the Fmin obtained is only 1.19, which is lower than those for SE
and De, but larger than the allowable safety factor 1.15. Thus CTE is superior to the other elements and can be
used in forecasting the behaviors of landslide with WIL correctly.

1 INTRODUCTION finally the coupled analysis efficiency will slow down

seriously. (2) Goodman element or Desai element (DE)
Since the 20th century, with the increasing of (Goodman, R.E., Taylor, R.L. & Brekke, T.L., 1968,
population, the extensions of human activities and Desai, C.S., Zaman, M.M. & Lightner, J.G., et al.
the influences of engineering activities on geological 1984). The conception of Goodman element is clear,
environment disturbances, the landslide disaster has and it can model the sliding and opening displacement
become the most frequent and serious geological haz- of contact interface well. But its main problems are
ards. It is well known that the distribution and seepage- the randomicity adopting normal stiffness and the dif-
stress coupled interaction of the weak intercalated ficulty deciding reasonable stiffness; though the DE
layers (WIL) in landslide are the key factors resulting introduces the interpenetration control method, which
in the landslide disaster (Oda, Masanobu 1986). Thus overcomes the faults of the original Goodman joint
modelling the coupled interaction of WIL correctly is element, it is short of theoretical bases on determining
the key forecasting the behaviours of landslide, but the calculation parameters (Luan Maotian & Wu Yajun
coupled interaction of WIL is seldom considered. So 2004). In addition, the Goodman and Desai element
this research is very important and meaningful. can’t consider the pore pressure, so they can’t directly
At present, there are two methods modelling the apply to model the seepage-stress coupled interac-
WIL: (1) solid element (SE) (Ding Xiuli, Fu Jing & tion of WIL, they can only take the seepage force
Zhang Qihua, 2004, Chai Junrui & Li Shouyi 2004). or pore pressure as the external load (Desai, C.S.,
Because the mechanical behaviours of materials on Samtani, N.C., et al. 1995, Samtani, N.C., Desai, C.S.,
the top and bottom of WIL are very different, and et al. 1996), then a single stress analysis is carried out
the relative sliding, opening, even the tension fracture for the WIL. So it can be seen that the two methods will
damage, may occur between the top and bottom of occur large errors and affect the validity forecasting the
WIL. Thus the solid element can’t model the WIL well; behaviours of landslide, and some improvements need
in addition, the solid elements nearby the WIL must to do for the above elements.
be refined to ensure the computational precisions, the This paper proposed a new type of 2D four nodes
corresponding quantity of nodes will increase rapidly, seepage-stress coupled thin-layer element (CTE)

827
   
based on the original Goodman’s joint and Biot’s σi δui dV + pδun dV + Ti δui dS = 0 (6)
consolidation differential equations. An elastic vis- V V S
coplastic constitutive model was adopted, and a FE  
kx h ∂p ∂δp ∂un
program was developed to verify the new model. In − dV − δpdV = 0 (7)
the end, it was applied to analyze the coupled effect of V rw ∂x ∂x V ∂t
Dayeping landslide.
where δui and δui represent any arbitrarily change in
2 FE MODEL relative and absolute displacements, respectively; δun
is the arbitrarily change in the relative normal displace-
2.1 Biot consolidation theories ment along the coupled element; Ti are the components
of traction along the boundary S.
The proposed coupled thin-layer element is shown in To obtain an approximate solution of Eqns (6) and
figure 1. Every node has three freedoms: u, v and p. (7), it is necessary to first relate the displacements
Two key assumptions are made here: (1) there are and water pressures to the nodal values. According
only two components of stress σn , τs in the coupled to the Goodman’s joint formulation, the displace-
element; (2) the velocity of seepage flow can only ments and pore pressures adopt the linear interpolation
occur in the tangential direction of the element, that is, function. The variables u, p, un , σ , ∂p/∂x in the global
p1 = p3 , p2 = p4 . The well known Biot’s consolida- coordinate system can be expressed
tion equations for the coupled thin-layer element are as
follows (Small, J.C., Booker, J.R., Davis, E.H., 1976):
u = BRδu (8)
σn = σn − p (1)
p = aT δp (9)
τs = τs (2)
Pk = −Hkl∗ δl (3) un = vtop − vbottom = V T Rδu (10)

∂vx ∂θ 1 ∂un ∂p ∂ T
=− =− = (a δp ) = Ex δp (11)
∂x ∂t h ∂t
(4) ∂x ∂x

kx ∂p σ  = −Du (12)
vx = − (5)  
rw ∂x −N1 0 −N2 0 N1 0 N2 0
B=
1 2 0 −N1 0 −N2 0 N1 0 N2
 T
where P = τs σn ; δl = {u v} , u, v represent the
T

relative shear and normal displacement components (13)

along the coupled element, respectively. 
1 2x
N1 = N3 = 1− (14)
2 L
2.2 FE approximation 
1 2x
Introduce the stress and seepage boundary conditions, N2 = N4 = 1+ (15)
into the stress equilibrium equations and water con- 2 L
tinuity equation, the finite element equations of CTE
can be derived by using the Virtual Work principle as δu = [u1 v1 u2 v2 u3 v3 u4 v4 ]T (16)
follows: 1
aT = [N1 N2 N1 N2 ] (17)
2
δp = [p1 p2 p3 p4 ]T (18)
n y
3 s 4 V T = [0 −N1 0 −N2 0 N1 0 N2 ] (19)
x  
h 1 ∂N1 ∂N2 ∂N1 ∂N2
Ex = (20)
1
s
2 2 ∂x ∂x ∂x ∂x
n

D is the constitutive matrix; De is the elastic matrix,

Figure 1. Coupled thin-layer element model (local coordi- the elastic viscoplastic matrix Dvp will be discussed
nate system) later.

828
  
λs 0 2.3 Constitutive law
De = (21)
0 λn
Rock slope always behaves slow and continuous
sliding under the load of gravity and seepage force,
where λs , λn are the unit shear and normal stiffnesses this characteristic is called creep (Desai, C.S.,
(Desai, C.S., Zaman, M.M., Lightner, J.G., et al. Samtani, N.C., et al., 1995). So this paper adopted
1984). the elastic viscoplastic constitutive law based on the
R is the transform matrix; r is the direction cosine Perzyna theory (Perzyna, P. 1966, Liu Baoguo & Qiao
matrix, θ is the angle between the global X axis and Chunsheng 2004). Suppose the relative displacements
the local x axis (Zhu Bofang 1998) consist of two parts: elastic and viscoplastic displace-
ments, the elastic displacements ue are independent of
R = diag [r r r r] (22) time and they can be decided by the elastic constitu-
tive equations; and yet the viscoplastic displacements
  uvp are dependent of time, they can be decided by
cos θ sin θ
r= (23) the Perzyna viscoplastic constitutive equations. The
− sin θ cos θ
increments of the total relative displacements in t n

If Eqns (8∼10) and (12) are substituted into Eqn (6) un = uen + uvp
n
(32)
then
The increments of effective stresses in t n
Kδu − L δp − m = 0T
(24)
σn = De uen = De ( un − uvp
n
) (33)

K = RT BT DBRdx (25) The increments of viscoplastic displacements based
L
 on the Euler time integration method
LT = RT VaT dx (26)  
L uvp
n
= t n (1 − )u̇vp
n
+ u̇vp
n+1
(34)

m= RT N T TdS (27) n+1
Expand the u̇vp using the Taylor series
S
   n
N1 0 N2 0 N1 0 N2 0 n+1
= u̇vp
n
+ ∂ u̇vp
n
/∂σ  σn = u̇vp
n
+ H n σn
N = (28) u̇vp
0 N1 0 N2 0 N1 0 N2
(35)

If Eqns (9∼11) are substituted into Eqn (7) then If Eqns (34∼35) are substituted into Eqn (31) then

∂δu σn = Dvp

n
(Bn δun − t n u̇vp
n
) (36)
−L − δp = 0 (29)
∂t
 De  −1
kx h
n
Dvp = = (De )−1 +  t n H n
= E T Ex dx (30) 1 + De t n H n
rw A x (37)

Eqns (24) and (29) can be solved by the increment of where  = 0.0, 0.5, 1.0. If the H n is decided, the elas-
displacements δu and pore pressures δp . Therefore tic viscoplastic matrix Dvp will be decided, and the H n
we are solving a set of coupled equations, can be decided by the Perzyna viscoplastic theory as
follows. The relative viscoplastic displacements can
     be expressed by the displacements rate
K −LT δu m
= (31) 
−L −α t δp tpt F ∂Q
u̇vp = γ   (38)
F0 ∂σ 
where pt are the pore pressures at time t, denotes the 3  4 
increments from time t to t + t; α is the integrating F (F/F0 ) F/F0 > 0
 = (39)
factor, its value is related to the integration scheme F0 0 F/F0 ≤ 0
and ranges from 0 to 1, a value of 0.5 was used here
to ensure unconditional stability of the analysis with γ represents the flow parameter, it can be decided by
respect to time (Booker, J.R. & Small, J.C. 1975). rheological test of interface;  represents the function

829
of viscoplastic displacements rate.

(F/F0 ) = (F/F0 )n (40)

thin-layer 1

where n represents the material parameter; F repre- thin-layer 1

landslide 1

sents the material yield criterion; F0 is the reference thin-layer 1

bedrock
value of F (Perzyna, P. 1966); this paper adopted landslide 2 294m
the Mohr-Coulomb yield criterion and associated flow landslide 3
235m
rule
thin-layer 2

F = Q = τs + σn tan ϕ − c (41) Figure 2. Material zone of landslide section 1.

F0 = c cos ϕ (42)

where c represents the cohesion of interface; ϕ repre-

sents the frictional angle (Swoboda, G., Mertz and W.,
Beer, G., 1987). If Eqns (38∼42) are substituted into
Eqn (35), the Dvp in local coordinate can be obtained.

3 EXAMPLES

A relevant FE program based on the above derivation

was developed on the secondary development plat-
form of Abaqus (Hibbitte, Karlsson, Sorenson INC.,
2002), then it was applied to the coupled analysis of Figure 3. Finite element mesh of landslide section 1.
the Dayeping landslide.
The Dayeping landslide lies on the Loushui river,
its main material compositions are nubby dolostone
and few clays. The sliding zone is the layered com- the corresponding flow rule is associated. Boundary
pression crush zone, whose original rockmass is the conditions: the x and y displacements at the bottom are
dolostone fragment, because of the integrated effect restricted, and it is impermeable; the x displacement
of groundwater and special tectonic zone, the dolo- on the left vertical plane is restricted; the upriver and
stone fragments weather seriously, dissolute and scour downriver water level are 500 m and 235 m, respec-
to form the WIL, and then the WIL results in the slid- tively. Because it is short of seepage-stress coupled
ing. The main sliding direction is S70◦ W, the plane test data, this paper doesn’t consider the relation-
shape of the landslide is round-backed armchair, its ship between permeability coefficient, void ratio and
elevation is from 230 m to 480 m, and the dip angle of stresses, strains in analysis, suppose the permeability
slide surface is between 18◦ and 26◦ . coefficient is constant. The groundwater free surface
The upriver groundwater level in the landslide is comparisons between the coupled analysis and in-situ
constant, the elevation is 500 m, the downriver normal testing data of section 1 and 2 are shown in figure 4,
water level is 294 m, and the dead water level is 235 m. the displacement vector and contour of section 1 are
The material zone and finite element mesh of the land- shown in figure 5, and the stresses contours of section 1
slide section 1 are shown in figure 1 and 2, respectively, are shown in figure 6.
and the thicknesses of WIL 1 and 2 are both 20 cm. It is seen in Figure 4 that the groundwater free sur-
The indexes n of the WIL and general rock are 3 and 2, face often passes through the WIL, whether or not
respectively, other material parameters are shown in simulating the coupled interaction of WIL correctly
table 1. The WIL adopted the coupled element pro- matters the location of critical slip surface. Through
posed in this paper to model, and the general rock comparisons between the coupled analysis and in-situ
adopted the 2D four nodes coupled element Cpe4p testing data, we found that the results are consistent
to model (Hibbitte, Karlsson, Sorenson, INC. 2002). well on the drill holes, and some discrepancies are also
The quantity of elements and nodes is 3615 and 3765, observed at the top bedrock in figure (a) and stepwise
respectively. The constitutive law of all materials is WIL in figure (b), the reasons may be few drill holes
elastic viscoplastic model based on the Perzyna theory, and the errors in the course of fitting the in-situ testing
and the yield criterion is the Mohr-Coulomb criterion, data, we think that these discrepancies are reasonable,

830
Table 1. Material parameters.

Material ρ/(kN/m3 ) E/(MPa) μ k/(m/s) e c/(kPa) ϕ/(◦ ) γ /(min−1 )

Bedrock 26.5 8000.0 0.26 4.98 × 10−6 0.012 300.0 50.0 0.00015
Thin-layer 1 19.5 λs (MPa/m) λn (MPa/m) 0.6(shear) 0.220 30.0 19.8 0.053
7.28 133.0 0.0(normal)
Landslide 1 22.0 10.0 0.35 2.3 × 10−7 0.201 55.0 25.0 0.00089
Landslide 2 21.0 9.8 0.33 2.0 × 10−3 0.195 58.0 27.0 0.00095
Landslide 3 20.5 9.5 0.31 3.2 × 10−4 0.180 60.0 31.0 0.0015
Thin-layer 2 19.8 λs (MPa/m) λn (MPa/m) 2.0(shear) 0.257 28.0 20.3 0.063
5.34 105.0 0.0(normal)

(m)

600

500
0
HZK6
421.0

400
100 0

HZK5
300 271.7
200 100 0

200

0 100 200 300 400 500 600 700 800 (m)

(a) Section 1

(m)
600

500
HZK3
437.9
0 HZK2
392.5
400
0
100
300 HZK1
267.0
100 0 Figure 5. Displacement of landslide section 1 (unit:m).
200

0 100 200 300 400 500 600 700 800 (m)

(b) Section 2
It is seen in figure 6(a) that the σx is the tensile
Figure 4. Groundwater free surface comparisons. stress nearby the rear end of landslide, the maximum
of the tensile stress is 0.27 Mpa, this may result in the
decrease of the rear end’s stability. Figure 6(b) shows
that the σy is the compressive stress in the whole land-
and they are in the permission range of engineering. slide, the maximum is 1.82 Mpa, it lies on the first
Therefore adopting the proposed CTE to model the platform, whose elevation is 421 m. Figure 6(c) shows
seepage-stress coupled interaction of WIL is feasible that the maximumof the τxy nearby the WIL is 0.3 Mpa,
and effective. it lies in the front of the landslide, its elevation is 230 m,
The displacement contour shown in figure 5(a) and its direction is upward along the WIL, the reason
indicates that the displacement on the surface of land- is that the landslide has the trend of sliding along the
slide is very large, and it decreases gradually with the WIL, and yet the bedrock resists sliding, especially in
increase of depth. In particular, the maximum occurs the front of the landslide. The τxy far from the WIL
on the first and second platform, the elevation is 421 m makes the landslide tend to slide downwards, and the
and 340 m, respectively. Figure 5(b) shows that the dis- maximum is −0.25 Mpa (‘‘−’’ represents the direction
placement vector has the trend of sliding along with is opposite to the stated positive direction), it lies on
the WIL, especially at the elevation from 230 m to the surface of landslide, its elevation is about 380 m.
370 m, Thus it can be primarily concluded that the Finally, this paper adopted the strength reduction
landslide will occur partial broken-line sliding along method to analyze the stability under three cases of
with the WIL at the elevation from 230 m to 370 m. CTE, SE and DE modeling the WIL. The cases of

831
 y (m)
659

600

500

400

300

200 (m)
0 100 200 300 400 500 600 700 800 x

Figure 7. Critical slip surface.

Table 2. Minimum safety factors.

y Coupled thin-layer element Solid element Desai element

1.19 1.22 1.27

4 CONCLUSIONS

1. A new type of coupled thin-layer element based on

xy
Goodman’s joint has been presented, which is based
on Biot’s well known consolidation theory, it can
Figure 6. Stress contours of landslide section 1 (unit: MPa). be modeled both the stress-deformation behaviours
and water pressure response of WIL, and it also
considered the time-dependent behaviors of WIL.
CTE and SE considered the coupled effect, and yet the So the proposed element is fully coupled.
case of DE only carried out stress analysis, the pore 2. A FE program of the CTE was developed to analyze
pressure was considered to be the external load. The the Dayeping landslide, the calculated groundwater
critical slide surface under the case of CTE is shown free surface is consistent well with the in-situ pore
in figure 7, and the minimum safety factors are shown pressure testing data. It indicates that the proposed
in table 2. coupled element is right.
The results indicate that the critical slide surfaces 3. The strength reduction method was adopted to
of three cases are consistent basically, the elevation appraise the stability of landslide. The results indi-
ranges from 230 m to 370 m, but the minimum safety cate that the SE and DE can’t model the tension
factors change evidently, the minimum safety factor of fracture damage or pore pressure of WIL, the cor-
CTE is the smallest, and yet that of DE is the largest. responding results are inclined to safe. And yet the
The case of CTE considers the coupled interaction of coupled element makes up the above faults, it can
landslide and WIL comprehensively, so it is close to the model the relative slide, tension fracture damage
true status of stresses in the landslide; The SE only con- and pore pressure, therefore its advantages are very
siders the coupled interaction of landslide partly, but evident. The corresponding calculation is close to
the most important affecting factor, that is, the coupled the true status of stresses in the landslide, the Fmin
effect of WIL is ignored, so it results in the increase is 1.19, which is evidently smaller than those of
of safety factor; And yet the DE doesn’t consider the SE and DE (1.22, 1.27), but it is larger than the
above coupled interactions at all, it only adopts the allowable safety factor 1.15, the landslide is still
discoupling style to solve the coupled problem, thus stable. So the proposed coupled element is superior
the corresponding error is large, and the safety factor to the other elements; it can forecast the stability of
also increases. landslide with WIL correctly, and the results also

832
 provide a theoretical basis for the landslide disaster reservoir [J]. Chinese Journal of Rock Mechanics and
prevention. Engineering, 23(17): 2913–2919.
4. The proposed CTE model is a completely coupled Goodman, R.E., Taylor, R.L. & Brekke, T.L. 1968.
model, it can consider the relationship between A model for the mechanics of jointed rock [J]. Journal
permeability coefficient, void ratio and stresses, of the Soil Mechanics and Foundations Division Proceed-
ings of the American Society of Civil Engineers, 94(SM3):
strains, but due to be short of seepage-stress 637–659.
coupled testing data, the coupled analysis in the Hibbitte, Karlsson, Sorenson, INC. 2002. ABAQUS/Standard
Dayeping landslide only supposed the permeabil- user’s manual[S].
ity coefficient is constant, this affects the results to Liu Baoguo & Qiao Chunsheng. 2004. Back analysis of
some extent. So developing seepage-stress coupled visco-plastic model parameters of rock mass [J]. Engi-
test equipments and researching the relationship neering Mechanics, 21(4): 118–122.
between the permeability coefficient, void ratio Luan Maotian & Wu Yajun. 2004. A nonlinear elasto-
and the stresses, strains are the basically but urgent perfectly plastic model of interface element for soil-
problems in rock seepage-stress coupled research. structure interaction and its applications [J]. Rock and Soil
Mechanics, 25(4): 507–513.
Oda Masanobu. 1986. Equivalent continuum model for cou-
pled stress and fluid flow analysis in jointed rock masses
REFERENCES [J]. Water Resources Research, 22(13): 1845–1856.
Perzyna, P. 1966. Fundamental problems in viscoplasticity
Booker, J.R. & Small, J.C. 1975. An investigation of the [J]. Advances in Applied Mechanics, 9: 243–377.
stability of numerical solutions of Biot’s equations of Samtani, N.C., Desai, C.S., et al. 1996. An interface model to
consolidation [J]. International Journal of Solids and describe viscoplastic behavior [J]. International Journal
Structures, 11(7–8): 907–917. for Numerical and Analytical Methods in Geomechanics,
Chai Junrui, Li Shouyi. 2004. Coupling analysis of seep- 20: 231–252.
age and stress fields in Xietan Landslide in Three Gorges Small, J.C., Booker, J.R. & Davis, E.H. 1976. Elasto-plastic
region [J]. Chinese Journal of Rock Mechanics and consolidation of soil [J]. International Journal of Solids
Engineering, 23(8): 1280–1284. and Structures, 12: 431–448.
Desai, C.S., Samtani, N.C., et al. 1995. Constitutive mod- Swoboda, G., Mertz, W. & Beer, G. 1987. Rheologi-
eling and analysis of creeping slopes [J]. Journal of cal analysis of tunnel excavations by means of coupled
Geotechnical Engineering, 121(1): 43–56. finite element-boundary element analysis [J]. Interna-
Desai, C.S., Zaman, M.M., Lightner, J.G., et al. 1984. tional Journal for Numerical and Analytical Methods in
Thin-layer element for interfaces and joints [J]. Interna- Geomechanics, 11(2): 115–129.
tional Journal for Numerical and Analytical Methods in Zhu Bofang. 1998. The finite element method theory and
Geomechanics, 8: 19–43. applications (2nd edition) [M]. Beijing: China Water-
Ding Xiuli, Fu Jing & Zhang Qihua. 2004. Stability anal- power Press, 597–599.
ysis of landslide in the south end of Fengjie highway
bridge with fluctuation of water level of three Gorges

833
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Numerical modelling of a rock avalanche laboratory experiment

in the framework of the ‘‘Rockslidetec’’ alpine project

I. Manzella
Rock Mechanics Laboratory, Swiss Federal Institute of Technology, Lausanne, Switzerland

M. Pirulli
Politecnico di Torino, Turin, Italy

M. Naaim
Cemagref, Grenoble, France

J. F. Serratrice
CETE Méditerranée, Aix en Provence, France

V. Labiouse
Rock Mechanics Laboratory, Swiss Federal Institute of Technology, Lausanne, Switzerland

ABSTRACT: This paper illustrates the results of a collaboration born among the authors, in the framework
of the INTERREG IIIA ‘‘Rockslidetec’’ project, with the aim of studying rock avalanche propagation by means
of physical and numerical modelling. The EPFL rock mechanics laboratory has carried out an experimental
campaign. Tests consist in releasing unconstrained gravel flows on an inclined panel. One of the experiments has
been simulated by the codes developed by the partners. The Cemagref code and RASH3D (Politecnico di Torino)
are based on a continuum mechanics approach; EPAN3D (CETE) on a discrete approach. Cemagref results,
although rather far from observations made on deposit morphology, reproduce quite well the travel distance,
without any parameter calibration. RASH3D reproduces well the velocity of propagation of the rear and front of
the mass, but less the final deposit shape. EPAN3D reproduces very well the final deposit but the calculations
are based on the calibration of 11 parameters.

1 INTRODUCTION The INTERREG IIIA ‘‘Rockslidetec’’ project

(2003–2006), among Italian, French and Swiss
Large rock avalanches or sturzstorms, as called by partners, has born to partly fill up these lacu-
Heim (1932), involve volumes of more than one mil- nas: its main goal is to develop methodological
lion of cubic metres of material and are rare but devices to define areas that could be affected by rock
catastrophic events. They engage a great amount of avalanche phenomena. The project is organised in
energy and exhibit a much greater mobility than could three sections:
be predicted using frictional models (Hungr et al.
– Action A: realization of an itemized list of historical
2001). For these reasons the most effective way to pre-
rockslide events;
vent them to cause damages and victims would be to
– Action B: rockslide geometrical characterization
forecast their runout and consequently to define areas
and determination of instable volumes;
that could be affected by their occurrence. On the other
– Action C: study and modelling of rockslide runout
hand, there are still a large number of unknown fac-
with comparison of different codes.
tors and numerical codes are still too sensible to some
geomechanical parameters like friction angle. More- In the action C framework the EPFL rock mechan-
over, since rock avalanches are luckily not so frequent, ics laboratory has carried out a research by means of
there is a lack of well known real cases on which to a physical model to determine the main parameters
base proper back analysis studies or code validations. that influence propagation. Even if the quantitative

835
interpretation of the results is not straightforward, change its inclination. A wooden cuboidal container
owing to the difficulty in matching the scaling laws, measuring 0.20 m height × 0.40 m width × 0.65 m
laboratory experiments are very helpful for the phe- length is filled with different amount of material and
nomenological study and the assessment of relevant placed on the tilting panel (Manzella and Labiouse
physical parameters as well as for numerical model 2007a, b). The box is opened in an almost instanta-
validation (Manzella & Labiouse, 2007a). neous way using a spring-loaded bottom gate and the
For this reason test results constitute a base of exper- material is released directly onto the slip surface.
imental comparison for the validation of the numerical Each test is filmed by a digital high speed camera
codes developed by the partners of the project. In placed at a height of about 5 m over the horizontal
particular in the action C framework one specific panel.
experiment has been used as a benchmark by Cemagref The parameters varied during the whole experimen-
(Grenoble, France), Politecnico di Torino (Turin, Italy) tal campaign are:
and CETE Méditerranée (Aix en Provence, France).
– material volume (10, 20, 30, 40 litres);
The results of the simulations have been useful to
– releasing height (1, 1.5, 2 m);
compare the three codes and underline their specific
– slope angle (30◦ , 37, 5◦ , 45◦ );
characteristics.
– nature of released material: Aquarium gravel of two
In this paper the experimental set-up and the three
different grain size distribution, Hostun sand and
codes are described in details, then results of the exer-
small bricks;
cise on the benchmark are shown and the comparison
– the number of releases (40 litres in one or in two
among the codes is illustrated.
consecutive releases of 20 litres)
– base friction coefficient (high roughness using
wood and low roughness using forex, a light PVC
2 LABORATORY TESTS sheet, to cover the model boards).
In this paper only one set of parameters is considered.
Tests mainly consist in simulating a rock avalanche
releasing an unconstrained flow of granular material
on an inclined panel. The experimental set-up (Fig. 1) 2.1 Measurements
mainly consists of two rectangular panels (3 m × 2 m)
The displacement and the velocity of the mass front
joined by a hinge. The first panel is fixed horizon-
during sliding have been evaluated through image
tally on a concrete floor slab and the second one can
analysis using the software WINanalyze. Run-out,
length, height and width of the final deposit have been
manually measured after each test.
Volume, deposit geometry and morphology can be
evaluated by means of the fringes projection method.
This is an innovative method which consists in pro-
jecting alternate lines of dark and light (fringes) on
the deposit surface. When fringes are projected on a
planar surface, they are straight and equally spaced,
whereas on a rough surface they are distorted and this
distortion is related to the shape of the object (Desman-
gles, 2003). Consequently, it is possible to retrieve the
information on the deposit thickness deriving it from
the departure from straightness of the fringes. The pre-
cision obtained using this method with the mentioned
test set-up is approximately 4 mm, which is in the
order of the maximum grain diameter of the gravel
used (D = 0.5 − 3 mm).

2.2 Benchmark test

The test on which the partners worked has the follow-
ing characteristics: a release of 20 litres of aquarium
gravel from 1 m height on a 45◦ inclined panel.
Aquarium gravel is a material used for aquarium
decoration; this material is suitable for laboratory
Figure 1. The experimental setup. experiments since it is treated during production in

836
order to reduce to minimum the quantity of dust. It topography, h the fluid depth, ρ the mass density, gx , gy
has a quite homogeneous grain size distribution with the projection of the gravity vector along the x and
a diameter D = 0.5 − 3 mm. Its unit weight is of y directions, σxx , σyy , τzx , τzy the normal and shear
14.3 kN/m3 . Its internal static friction angle is φi = stresses components and t the time.
34◦ and the one at the base between the gravel and the Adopting the above described approach, changes of
panel surface is φb = 32◦ . the mechanical behaviour within the flow are ignored
Thanks to the fringes projection method and the and the complex rheology of the flowing material
film registered with the high speed camera it has been is incorporated in a single term, describing the fric-
possible to furnish to the partners the digital elevation tional stress that develops at the interface between the
model of the final deposit and images of the flow each flowing material and the rough surface.
1/60 second. On the other hand a model such as EPAN3D (CETE
Méditerranée) using a discrete element method (DEM)
assumes that the mass consists of separate, discrete
3 NUMERICAL MODELS particles and that the moving mass is the result of the
interaction of the particles. The forces acting on each
Three numerical codes are considered in the present particle are computed taking into account the initial
paper. Two of them are based on a continuum mechan- configuration and the relevant physical laws. Accord-
ics approach while the third, EPAN3D, on a discrete ing to Cundall (1988) any particle may interact with
element method. any other particle and there are no limits on particle
RASH3D (Politecnico di Torino) and the code displacements and rotations.
developed by Cemagref treat the moving mass as a In the case of rock avalanche modelling, particles
homogeneous continuum, assuming that both depth interact between them through friction and the equa-
and length of the flowing mass are usually large if com- tions of the conservation of mass, of energy and of
pared with the characteristic dimension of the particles quantity of momentum are used. The velocity distri-
involved in the movement. Under this assumption, bution field and the thickness of the mass in movement
it becomes possible to replace the real moving mix- are obtained by the computation of the dynamics of a
ture of solid and fluid phases by an ‘‘equivalent’’ discretized deformable mass on a grid.
fluid, whose rheological properties have to approxi-
mate the behaviour of the real mixture (Hungr 1995).
3.1 Cemagref
By this way, the dynamic behaviour of the flowing
mass can be described by the mass and momentum In Cemagref model the domain is restricted to rock
conservation laws. Further, assuming that the ver- avalanches made of dry and cohesion-less grains. The
tical structure of the flow is much smaller than its Bagnold’s profile is adopted to represent the variation
characteristic length, the codes integrate the balance of velocity within the depth of the flow. A friction
equations in depth, obtaining the so-called depth- model issued from the recent progress on granular
averaged continuum flow models (Savage and Hutter material flows (Pouliquen, 1999) is chosen to rep-
1989). The general system of equation to be solved is resent the momentum loss at the substratum level.
as follows: The pressure distribution inside the flowing mate-
rial is considered as isotropic, which means that the
⎧ ∂h ∂(v h) ∂(v h) earth pressure coefficient is taken equal to 1. After-
⎪
⎪ +
x
+
y
=0 wards, the shallow water equations were extended
⎪
⎪ ∂t ∂x ∂y
⎪
⎪ to large rock avalanche flows over a complex topog-
⎪
⎪ 8 9
⎪
⎪ ∂(vx h) ∂(vx2 h) ∂(vx vy h) raphy. The variables of the equations are the depth
⎪
⎪ ρ + +
⎪
⎪ and the velocities. The system of equations is written
⎪
⎪ ∂t ∂x ∂y
⎪
⎪ in curvilinear coordinates. The mathematical con-
⎪
⎪
⎨ ∂(σxx h)
servative equations system is solved by a Godunov
=− + τzx(z=b) + ρgx h (1) numerical scheme using a finite element unstructured
⎪
⎪ ∂x mesh. This kind of mesh allows taking into account
⎪
⎪ 8 9
⎪
⎪ the complexity of the mountain terrain especially near
⎪ ρ ∂(vy h) + ∂(vy vx h) + ∂(vy h)
⎪ 2
⎪
⎪ singularities. To increase the accuracy of the solver
⎪
⎪ ∂t ∂x ∂y
⎪
⎪ a linear variation of the height and mean velocities
⎪
⎪
⎪
⎪ inside each cell is considered. In order to avoid numer-
⎪
⎪ ∂(σyy h)
⎩ =− + τzy(z=b) + ρgy h ical instabilities, the gradients of these variables were
∂y limited according to the criterion defined by Van Leer
(1979). Each component of this model was severely
where v = (vx̂, vŷ) denotes the depth-averaged flow tested using both analytical solutions and laboratory
velocity in a reference frame (x, y, z) linked to the experiments.

837
3.2 RASH3D dispersive shear stress components, k is the flow
resistance parameter.
The RASH3D code (Pirulli, 2005) originates from
a pre-existing model (SHWCIN) based on the clas- The first and the second terms on the right hand side
sical finite volume approach for solving hyperbolic of equation [4] are, respectively, the yield term and
systems using the concept of cell centred conserva- the viscous term as defined in the Bingham equation.
tive quantities, developed by Audusse et al. (2000) The last term represents the turbulence contribution
and Bristeau et al. (2001) to compute Saint-Venant (O’Brien et al. 1993).
equation in hydraulic problems. The RASH3D code has been widely validated sim-
An extension of SHWCIN for simulating dry gran- ulating laboratory tests (e.g. Mangeney et al., 2003;
ular flows using a kinetic scheme was initially intro- Pirulli, 2005) and back analysing real cases (e.g.
duced by Mangeney-Castelnau et al. (2003). Pirulli Pirulli et al., 2007; Pirulli and Mangeney, 2007).
(2005) proposed further modifications to SHWCIN Time for calculation of each numerical analysis can-
to reduce observed mesh-dependency problems, per- not be univocally defined but it is mainly a function of
mit simulation of motion across irregular 3D terrain, the digital elevation model discretization.
incorporate the influence of internal strength and allow Further developments of RASH3D include the
the selection of more than one possible basal resistance numerical implementation of entrainment of material
relationship. along the path of propagation and the arrangement of
As input data the code requires the digital elevation a more user friendly interface.
model of the studied area, the identification of the
boundary of the source area and the geometry of the
initial volume. 3.3 EPAN3D
As for the rheological characteristics of the flowing The objective of EPAN3D (CETE Méditerranée) is
mass, three different rheologies are implemented in to simulate a rock mass of large volume, its propa-
RASH3D at the present time: gation along a mountain slope then the accumulation
and the spreading out in the valley. The unstable vol-
1. the simple frictional rheology, based on a constant
ume of rock is discretized in small elements of volume.
friction angle ϕ, which implies a constant ratio of
After failure, these elements slide on the digital terrain
the shear stress to the normal stress. Shear forces,
model (DTM), until they find equilibrium at the bot-
τ , are independent of velocity.
tom of the valley where they accumulate and spread
out. The moving rock cluster is permanently associated
τ = γ h tan ϕ (2) with morphology of the slope by updating the DTM
at each calculation step. The energy is dissipated pri-
where γ = unit weight, ϕ = basal friction angle marily through friction between the sliding elements
and h = flow depth. and at the base.
2. the Voellmy flow relation, which consists of a Two local evolution laws are used for modelling.
turbulent term, v2 /g, accounting for velocity- A first law fixes the ratio of the friction μ = tgφ
dependent friction losses, and a Coulomb or basal against the instantaneous velocity v of a sliding
friction term for describing the stopping mecha- element. It takes the form:
nism. The resulting basal shear stress is given by
the following equation: v ≤ v0 μ = μ0
γ v2 v > v0 μ = μ0 + (μ1 + μ0 )
τ = γ h tan ϕ + (3)   
ξ (v − v0 )
× 1 − exp − (5)
v1
where v = the mean flow velocity, ξ =
turbulence coefficient; the others symbols are where μ0 = tgφ0 , μ1 = tgφ1 , v0 et v1 are
similar as in equation [2]. parameters (μ0 ≤ μ1 and v0 < v1 ).
3. the quadratic rheology, where the total friction is Friction increases asymptotically with velocity
provided by the following expression: towards a limit value μ1 . On the other hand it remains
equal to a minimal value μ0 when v is lower than v0 .
kηv ntd v2 This formulation is inspired by the laws suggested by
τ = γ hτy + + β (4)
8h h Pouliquen (1999).
Another law describes the evolution of the shape
where τ is the shear stress, τy is the Bingham and the volume of each sliding element along its path.
yield stress, η is the Bingham viscosity, ntd is the Each elementary volume is supposed to take the shape
equivalent Manning’s coefficient for turbulent and of a paraboloid with elliptic section whose shape can

838
be parameterized by means of an elongation modu- – The RASH3D simulation was also based on a model
lus and a spreading coefficient. The volume of the of the type ‘‘frictional’’. The basal friction angle was
sliding elements also evolves according to a bulking fixed at 42◦ , after calibration on the results of the
law that depends on the distance covered from failure. laboratory test.
Finally, the initial unit weight of the rock is given. On – For the EPAN3D code the released mass was divided
the whole, calculations are based on ten mechanical in 1911 elementary volumes arranged in seven lay-
parameters. ers according to a cubic network. A back-analysis
After having discretized the unstable rock volume of the laboratory test allowed a calibration of the
in small elements, a calculation ‘‘by spreading’’ (par parameters of the constitutive model, such as γ =
épandage) is carried out by successive time steps: at a 16.5 kN/m3 , φ0 = 11◦ , φ1 = 44.5◦ , v0 = 0.1 m/s
given time, the trajectory of a sliding element within and v1 = 0.5 m/s.
the rock cluster is determined by linear and angular
momentum balance conservation equations applied to Digitalization of the geometry of the model (an
the local tangent plan covered by the element. Thus, inclined panel at 45◦ degrees and a horizontal panel)
calculations by spreading allow following the position as well as the definition of the zone of departure
of all the sliding elements at every time and making (0.25 m × 0.20 m × 0.4 m box filled with 20 litres of
the form of the topographic surface evolve during the granular material) was in charge of the teams, but there
movement by update of the DTM. Therefore this pro- should not be any differences.
cess provides a method of modelling the propagation
and the spreading of the rock cluster. Furthermore it
gives at every moment the position of all its elements,
the extent and shape of the cluster, from the time of
release until its complete stabilization.
EPAN3D is designed to study various scenarios of
rupture or ruptures by successive stages (step-by-step
scenarios).

4 SIMULATION RESULTS

To simulate the benchmark test the partners have used

the following data depending on their codes:
Figure 3. 3D view of the observed deposit and the codes
– For his code, based on a ‘‘frictional’’ model, Cema- simulations. Data are presented at the same scale.
gref used a friction coefficient equal to 0.62 corre-
sponding to the tangent of the friction angle at the
base (φb = 32◦ ). No calibration has been made.

Figure 4. Comparison between the mass propagation and

Figure 2. Longitudinal section along the symmetry axis of spreading observed in laboratory (images in background) and
the observed deposit (light grey) and codes simulations (black those modelled by the three codes. The outline represented
points). in white concerns the 1 mm thickness contour line.

839
 Table 1. Qualitative comparison among the codes.

Cemagref RASH3D EPAN3D

Model frictional frictional pouliquen
Analysis without back- back-
calibration analysis analysis

Number of parameters 1 1 11
Propagation Runout +/− +/− +
Velocity − + −
Deposit Longitudinal ext. − +/− +
Dimensions Lateral extension +/− − +
Height +/− +/− +
Deposit General shape − − +
Morphology Front − +/− +/−
Rear − − +

In Figures 2, 3 and 4 the simulation results are test allowed a calibration of respectively the friction
shown. angle for RASH3D and the eleven parameters of the
Although relatively far from observations made developed constitutive model for EPAN3D.
on the extension and the morphology of the gravel The simulated test consisted in an unconstrained
deposit, the results obtained by Cemagref are rele- flow of 20 litres of aquarium gravel, released from
vant in terms of runout distance, if it is taken into 1 m height on a slope inclined at 45◦ .
account that no parameter calibration has been made Cemagref reproduces quite well the runout dis-
(class A prediction). As can be seen from Figure 4, tance, without any parameter calibration (class A
the calculation made by RASH3D reproduces well the prediction). RASH3D reproduces well the velocity
velocity of propagation of the rear and front of the of propagation of the rear and front of the mass, but
mass, but less the final deposit shape, especially in less the final deposit, in particular in the transversal
the transversal direction. Besides, the chosen angle of direction. Besides, the chosen angle of friction dif-
friction is far from the experimental value. Observing fers from the experimental value. EPAN3D is better in
Figures 2 and 3 it is found that the deposit mor- reproducing deposit morphology and dimensions, but
phology and dimensions are better simulated by the calculations are based on a back-analysis and on the
model EPAN3D, but it has to be taken into account calibration of 11 parameters.
that they result from a back-analysis and from the On the other hand, since only one test has been
calibration of 11 parameters, not all with physical used for this work of comparison, it is important to
meaning. put the above-mentioned statements into perspective.
In Table 1 the predictions of the three codes are qual- Only a comparison of the codes among several test
itatively compared with the observations made during configurations would allow a proper analysis.
the laboratory test. The symbols +, +/− and − mean This study underlines the importance of the valida-
respectively good, average, imperfect adequacy, for tion of numerical models on well-defined laboratory
the characteristic indicated on the corresponding line. tests since this contributes to understand advantages
and limits of the codes, improving their development
and their use and consequently enhancing forecast of
5 CONCLUSIONS rock avalanche propagation.

In the INTERREG IIIA ‘‘Rockslidetec’’ framework a

test carried out at the LMR-EPFL has been used as a ACKNOWLEDGEMENTS
benchmark for three different codes developed by the
project partners. Cemagref method, commonly used The authors thank the Canton of Valais, the OFEG, the
for snow avalanches, is based on continuum mechan- SECO, the Regione Valle d’Aosta, the Région Rhône-
ics approach and a ‘‘frictional’’ model; RASH3D, is Alpes and the departments of Isère and Savoie for
also based on a continuum mechanics approach and, funding. Further acknowledgements go to the Insti-
in the case considered, a model of the type ‘‘frictional’’ tut de Physique du Globe de Paris for having put the
is used as well; EPAN3D is based on the discrete SHWCIN code at disposition and to Professor Claudio
elements method. A back-analysis of the laboratory Scavia for fruitful discussions.

840
REFERENCES modelling of non-constrained gravel flows. Rock Mech.
and Roch Engineering Journal (DO10.1007/s00603-007
Audusse, E., Bristeau, M.O. & Perthame, B.T. 2000. Kinetic -0134-y).
schemes for Saint-Venant equations with source terms on Manzella, I. & Labiouse, V. 2007b. Rock avalanches: exper-
unstructured grids. INRIA Report 3989, National Institute imental study of the main parameters influencing prop-
for Research and Computational Sciences and Control, agation. Proc. 11th ISRM, 9–13 July, 2007, Lisbon, 1:
LeChesnay, France. 657–660.
Bristeau, M.O., Coussin, B. & Perthame, B. 2001. Bound- O’Brien, J.S., Julien, P.Y. & Fullerton, W.T. 1993. Two-
ary conditions for the shallow water equations solved by dimensional water flood and mudflow simulation. Journal
kinetic schemes. INRIA Rep. 4282, National Institute of Hydrological Engineering, 119(2): 244–261.
for Research and Computational Sciences and Control, Pirulli, M. 2005. Numerical modelling of landslide runout, a
LeChesnay, France. continuum mechanics approach. PhD Thesis, Department
Cundall, P.A. 1988. Formulation of a three dimensional of Structural and Geotechnical Engineering, Politecnico
distinct element model-Part I. A scheme to detect and rep- di Torino, Italy.
resent contacts in a system composed of many polyhedral Pirulli, M. & Mangeney, A. 2007. Results of back-analysis
blocks. Int. J. Rock Mech. Min. Sci., 25: 107–116. of the propagation of rock avalanches as a function of
Desmangles, A.I. 2003. Extension of the fringe projection the assumed rheology. Rock Mech. and Roch Engineering
method to large object for shape and deformation mea- Journal (DO10.1007/s00603-007-0143-x).
surement. Ph.D. thesis no 2734, Ecole Polytechnique Pirulli, M., Bristeau, M.O., Mangeney, A. & Scavia, C.
Fédérale de Lausanne, CH. 2007. The effect of the earth pressure coefficients on the
Heim, A. 1932. Bergsturz und Menschenleben. Frets und runout of granular material. Environmental modelling &
Wasmuth software, 22: 1437–1454.
Hungr, O., Evans, S.G., Bovis, M. & Hutchinson, J.N. 2001. Pouliquen, O. 1999. Scaling laws in granular flows down
Review of the classification of landslides of the flow rough inclined planes. Physics of Fluids, 11(3): 542–548.
type. Environmental and Engineering Geoscience, VII: Savage, S.B. & Hutter, K. 1989. The motion of a finite mass
221–238. of granular material down a rough incline. Journal of Fluid
Hungr, O. 1995. A model for the runout analysis of rapid Mechanics, 199: 177–215.
flow slides, debris flows, and avalanches. Canadian Serratrice, J.F. 2006. Modélisation des grands éboulements
Geotechnical Journal, 32(4): 610–623. rocheux par épandage. Application aux sites de la Clapière
Mangeney-Castelnau, A., Vilotte, J.P., Bristeau, O., (Alpes Maritimes) et de Séchilienne (Isère). Bulletin des
Perthame, B., Bouchut, F., Simeoni, C. & Yerneni, S. laboratoires des ponts et chaussées, no 263–264, juillet-
2003. Numerical modeling of avalanches based on Saint août 2006: 53–69.
Venant equations using a kinetic scheme. Journal of Van Leer, B. 1979. Towards the ultimate conservative
Geophysical Research, 108(B11): 2527. difference scheme, V, J. Comp. Phys., 32: 101–136.
Manzella, I. & Labiouse, V. 2007a. Qualitative analysis
of rock avalanches propagation by means of physical

841
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Three-dimensional slope stability analysis by means

of limit equilibrium method

Shingo Morimasa & Kinya Miura

Toyohashi University of Technology, Aichi, Japan

ABSTRACT: We have conducted a series of stability analyses of cohesive soil slopes with limit equilib-
rium method to clarify the effect of three-dimensional shapes of sliding mass and slope on the slope stability.
In the analyses, straight and curved slopes consisting of frictionless soil were examined; the curved slopes
were defined with concentric circles instead of straight lines, and sliding surfaces were defined with elliptic
revolution. Through the analytical investigations some important features of three-dimensional slope sta-
bility were found. Even straight slope with finite width has a critical sliding depth with notably higher
stability factor than that of with infinite width which shows the lowest stability at infinite depth as in
ordinary two-dimensional analysis. Curved slopes have their own critical sliding width depending on its
curvature. We updated the stability diagram presented by Terzaghi with three-dimensional effects taken into
consideration.

1 INTRODUCTION within the limit equilibrium method. Through the

calculation of the driving and resisting moments of
Usually soil slope is modeled within two-dimensional the possible sliding surfaces, the critical sliding sur-
condition, and the slope stability against sliding is ana- faces are determined; therefore, the shape, size and
lyzed in plane strain condition. In the analysis the location of the sliding surface at critical condition are
slope is simplified to be straight and the width of obtained, and associated stability factor of the slope
sliding mass is assumed to be infinite. The model- is calculated. The calculation results are summarized
ing of slope in two-dimensional condition with the into several charts; some are for the description of
simplifications leads to lower slope stability; that critical sliding surface, and the others are for stabil-
is, lower safety factor is obtained, compared with ity factors presented by Terzaghi. Finally the three-
that in three-dimensional condition. Thus, stability dimensional effects of slope configuration and sliding
analysis conducted in two-dimensional condition is surface are discussed, and the important role of three-
apparently in conservative design side, and has been dimensional analysis in slope stability evaluation is
accepted in design practices. In reality, however, con- presented.
figuration of soil slope is not straight but curved, In the Terzaghi’s text he made a critical comment
and sliding mass has a finite width not as in two- on the role of stability chart as follows: ‘‘Owing to the
dimensional analysis. The effects of three-dimensional complexity of field conditions and to the important
shape of soil slope and sliding mass are notably difference between the assumed and the real mechan-
large and not negligible in the evaluation of the slope ical properties of soils, no theory of stability can be
stability. more than a means of making a rough estimate of the
In this study we have conducted a series of three- available resistance against sliding. If a method of
dimensional slope stability analyses on three types of computation is simple, we can readily judge the prac-
soil slopes: Straight Slope, Concave Curved Slope tical consequences of various deviations from the
and Convex Curved Slope. All the slopes considered basic assumption and modify our decisions accord-
are assumed to consist of homogeneous frictionless ingly. . . . ’’. We hope that the update of the Terza-
cohesive soil. In the analysis, sliding surfaces are ghi’s stability chart we make in this paper, con-
modeled with spheroid, elliptic revolutions. The slid- tributes to the reasonable evaluation of slope stability
ing mass was divided into thin slices, and driving and the engineering judgement in three-dimensional
and resisting moments were numerically calculated condition.

843
2 ANALYSIS METHOD where Or (xr , zr ) is the revolving axis, and r( y), a func-
tion of y, is the radius of the revolving solid. From
2.1 Spheroidal sliding surface Eq. 1 we employed a spheroid as a sliding surface as
shown in Fig. 2, because spheroid can be considered an
Figure 1 shows a cohesive straight slope with an incli-
appropriate shape of sliding surface and is easy to ana-
nation of β and an assumed three-dimensional sliding
lyze mathematically. The sliding surface is determined
surface and sliding mass. The y-axis is situated in
by the following equation:
the direction of slope width, and another horizontal
axis, x-axis, and vertical axis, z-axis, both are per-
(x − xr )2 y2 (z − zr )2
pendicular to the y-axis as shown in the figure; Ugai
2
+ 2 + =1 (2)
(1985) explained that the sliding surface must have a R wy R2
shape of revolving solid in accordance with the varia-
tion method presented by Baker et al. (1997), provided where R is the radius of circle in vertical x-z plane,
that the equilibrium of the forces on the sliding mass and wy is size of the spheroid in the y-direction. The
is satisfied, and the factor of safety against sliding radius of circle in the plane parallel with x-z plane can
is minimized. The revolving solid is defined by the be calculated as
following equation: 
r( y) = R 1 − ( y/wy )2 (3)
(x − xr )2 + (z − zr )2 = r 2 ( y) (1)
2.2 Slope shape
Ordinary slopes can be geometrically roughly clas-
sified into three types as shown in Fig. 3: (a)
Straight Slope, (b) Concave Curved Slope, and (c)
Convex Curved Slope. While Straight Slope has two-
dimensional configuration, the other two types of
Curved Slopes have three-dimensional configuration.
Toe and shoulder lines are straight and parallel each
other in Straight Slope, and are concentric circles in
Curved Slopes. All the slopes are symmetrical with
respect to x-z plane and defined by the following
parameters: slope angle β and slope height H . Curved
Slope has one more parameter, normalized curvature
radius Rt /H ; the value Rt represents the radius of toe
line.

Figure 1. Three-dimensional sliding mass assumed in cohe- 2.3 Analysis method

sive straight slope.
Figure 4 shows a Straight Slope in plan view and in
section. The shaded part in the figure indicates a slid-
ing mass whose width is 2wsld . The symmetrical slope
shape can cause a symmetrical slide, which allows us
to evaluate slope stability by computing only half of
the sliding mass, 0 ≤ y ≤ wsld . Calculating the ratio of
the resisting moment Mr to the driving moment Md , we
obtain the factor of safety FS of the slope with respect
to sliding along the sliding surface
 wsld 
Mr c H L( y)r( y) (dr/dy)2 + 1 dy
FS = = · 0  wsld
Md γH 0 Gz ( y)dy
(4)

where c is the cohesion of the soil, γ the unit weight,

Gz (y) geometrical moment of area with respect to
z  -axis of a slice at y (see Fig. 4), and L( y) and r( y)
Figure 2. Spheroid for determining sliding surface. is the length and the radius of sliding arc of the slice,

844
 respectively; the values of Gz ( y), L( y) and r( y) can
be calculated mathematically. Differentiating Eq. 3 by
y, we obtain

dr R·y
=−  (5)
dy wy wy2 − y2

From Eq. 4, the stability factor NS devised by

Terzaghi is
 wsld 
γH γH H L( y)r( y) (dr/dy)2 + 1 dy
NS =  = = 0  wsld
c c/FS 0 Gz ( y)dy
(6)

The mathematical integration in Eq. 6 is difficult

analytically. The sliding mass is, therefore, divided
into several slices and the value NS is calculated by
means of Simpson’s rule of numerical integration. This
investigation must be repeated for different revolving
axis Or in each analysis condition by trial and error.
The stability factor of the slope is equivalent to the
smallest value of the factor of safety FS thus obtained.
Two types of Curved Slopes can be analyzed in the
similar manner.
In the above calculations, we assume that the slid-
Figure 3. Three types of slope shape. ing mass is rigid body and moves toward the direction
perpendicular to the y-axis. For this reason, no share
forces is assumed to generated in the y-direction on the
sliding surface at sliding condition, that is, the shear-
ing resistance works only in the x- and z-directions as
shown in Fig. 5. It should be noted that shear and nor-
mal stresses on the interfaces between the slices both
have no influence on the calculation of the driving and
resisting moments as long as frictionless and cohesive
soil slopes are concerned.

Figure 4. Straight Slope in plan view and in section. Figure 5. Force acting on slice of sliding mass.

845
2.4 Analysis conditions slope condition, we can find that NS_3D is always
higher than NS_2D regardless of assumed sliding width.
Each slope is analyzed against two types of failure: toe
It has been indicated that the shearing resistance on the
failure and base failure. The depth of firm stratum on
ends of both sides of sliding mass in three-dimensional
which slope rests is expressed by the depth factor
sliding surface contribute to the increase in stability of
D slope. The end effect becomes notable with a decrease
nD = (D = nD · H ) (7) in sliding width and/or in firm stratum depth.
H In Fig. 7(a) to (e), the values of NS_3D are plot-
ted against the slope angle β for different values of
where D is the depth of firm stratum from the upper
wsld /H . It was confirmed that Fig. 7(f), plotted in
ground surface. While analysis on Straight Slope has
the same manner for NS_2D , was identical with Terza-
two parameters, slope angle β and assumed sliding
ghi’s stability diagram. Comparing Fig. 7(e) and (f),
width wsld , that on Curved Slopes have one more
we can recognize that the end effect is effective even
parameter, normalized curvature radius Rt /H .

3 CALCULATION RESULTS
AND DISCUSSION

3.1 Stability of straight slope in two- and three-

dimensional condition
3.1.1 End effect
Figure 6(a) to (f) shows the relations between normal-
ized sliding width wsld /H and stability factor NS_3D
for straight slopes with different slope angle β. In
order to investigate the effect of sliding width on
stability of slope, stability factor NS_2D obtained in
two-dimensional analysis is plotted on right-hand side
of the graphs. Comparing these values on the same

(a)
17
16 = 15 [Straight Slope]
15
nD = 1.0
14 nD : depth factor
13
2-D
nD' : critical depth factor
12 nD = 1.2
nD' = 2.0 1.0 (b)
11 11
10 10 = 30
nD = 1.5 1.2
nD = 1.0
2-D
9 9
nD' = 1.5
8 Toe Failure nD = 2.0 1.5 8 nD = 1.2
1.0
7 3.0 4.0 Toe 7 Toe Failure 1.5 1.2
5.0 nD = 4.0 2.0 3.0 Toe
6 6 2.0 4.0 2.0 1.5
Stability Factor, NS

4.0
All Type of Failure 6.0 All Type of Failure 4.0 2.0
5 5 4.0
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20

(c) (d)
10 10
9 = 45 9 = 60
nD' = 1.0 2-D
8 nD' = 1.2 2-D 8 1.2
nD = 1.0 1.5
7 Toe Failure 7
1.2 1.0 Toe Failure nD = 1.0 2.0
6 1.5 2.01.5 1.2(Toe) 1.2 4.0
2.0 3.0 4.0 1.5 1.5 2.0 nD' = 3.0 1.0
5 2.0 5 ~ 4.0
All Type of Failure 4.0 All Type of Failure (Toe)
4 4
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20

(e) (f)
9 9
8 = 75 8 = 90
nD' = 1.0 for all
7 7
2-D nD' = 1.0 for all
6 nD = 1.0 ~ 4.0 6 2-D
5 5 nD = 1.0 ~ 4.0
All Type of Failure (= Toe Failure) 1.0
~ 4.0
4 (Toe) 4 All Type of Failure (= Toe Failure) 1.0
~ 4.0
3 3 (Toe)
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
Figure 7. Stability diagrams for straight slope that are rela-
Sliding Width, w /H sld tions between slope angle β and stability factor NS for
different normalized sliding width wsld /H and depth factor
Figure 6. Relations between assumed normalized sliding nD ; (a) to (e) are for three-dimensional condition; (e) is
width wsld /H and stability factor NS for different slope angle for almost infinite sliding width; (f) is for two-dimensional
β and depth factor nD (straight slope). condition.

846
in the case of sliding mass with infinite width in (a) (b)
11 11
three-dimensional condition. 10 = 15 o
10 = 30
o

9 Rt/H = 5 10 20 9 Rt/H = 5 10 20
Toe Failure 100 Toe Failure 100
8 straight 8
3.1.2 Critical depth of sliding surface and critical 7 Rt/H = 5 7 Rt/H = 5
10 10
stability factor

Stability Factor, NS
20 100 20
6 straight 6 100
All Type of Failure All Type of Failure
5 5
Calculating stability factor for sliding mass with arbi- 0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 2 0
trary depth, we can obtain the relations as shown in
Fig. 8. While the stability factor in two-dimensional (c) (d)
10 10
condition decreases with increasing depth, that in 9 = 45
o
9 = 60
o

three-dimensional condition has the minimum value. 8 Rt/H = 5 10 20

Toe Failure 100
8 Rt/H = 5
Toe Failure 100
7 7 Rt/H = 5 10 20
The corresponding depth is the critical depth to which 6
Rt/H = 5 10
20 100 6
10
20
100

the sliding surface can develop. It is natural that the 5 All Type of Failure 5 All Type of Failure

appropriate critical depth and the corresponding crit- 4

0 2 4 6 8 10 12 14 16 18 20
4
0 2 4 6 8 10 12 14 16 18 2 0
ical stability factor NS should exist according to the
Sliding Width, wsld/H
sliding width. The values of NS are plotted in Fig. 6 [ Concave( ) Curved Slope ]
and 7 (all type of failure). We can thus find the min- (e) (f)
imum stability factor without knowing depth factor, 11 11
o o
= 15 = 30
which leads to an error on the safe side. 10 10
9 Rt/H = 10 9
straight Rt/H = 5 Toe Failure
8 20 8
10 20
7 100 7
Rt/H = 10 20 100
Stability Factor, NS
100 straight 6 10
3.2 Stability of curved slope 6
All Type of Failure Rt/H = 5
20
All Type of Failure
100
5 5
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
3.2.1 Critical width of sliding surface
Figure 9 shows the relations between assumed normal- (g) Toe Failure
(h)
10 10
ized sliding width wsld /H and stability factor of curved 9 = 45
o
9 = 60
o

slope for different slope angle β. Each figure contains 8 Rt/H = 5

Toe Failure 8 20
Toe Failure 100
7 10 20 7 Rt/H = 5 10
the values of stability factor against toe failure and 6 10 20
100
100 6
20
Rt/H = 5 100
those of critical stability factor against base failure 5 All Type of Failure 5 t = 5 10
R /H
All Type of Failure
4 4
for different normalized curvature radius Rt /H ; these 0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
values for straight slope are also plotted in the figure Sliding Width, wsld/H
by broken lines. It can be seen from the figure that [ Convex( ) Curved Slope ]
each curved slope has the minimum stability factor,
which indicates the existence of critical sliding width Figure 9. Relations between assumed normalized sliding
to which the sliding mass can expand. width wsld /H and stability factor NS against toe failure and
critical stability factor NS against base failure for different
slope angle β and depth factor nD ; (a) to (d) are for concave
3.2.2 Influence of curvature on stability of slope curved slope; (e) to (h) are for convex curved slope.
It is found in Fig. 9 that both types of curved
slopes indicate higher critical stability factors than
straight slope. On the other hand, for stability fac-
tors against toe failure, different behaviors appear
11 depending on the type of curvature; compared with
straight slope, concave type has higher value and con-
10
vex type has smaller value. Incidentally, the difference
wsld/H = 1.0 in stability factor between curved and straight slopes
Stability Factor, NS

9
becomes negligible with narrow sliding width and
8 large curvature.
7.08 The effect of curvature radius Rt /H on critical slid-
2.0 
7 ing width wsld and the corresponding stability factors
3.0
4.0 are shown in Fig. 10 for different slope angle β. It is
6 5.0 found from the figure that the increase in width of crit-
1.8 2-D ical sliding mass and the decrease in the corresponding
5 stability factors are almost in proportion to increase in
1.0 1.5 2.0 2.5 3.0 3.5
the logarithm of curvature radius.
Depth from Upslope, d/H
As a result of above discussion, stability diagram
Figure 8. Relation between normalized depth d/H of sliding for curved slope was finally obtained for differ-
mass and stability factor NS for different sliding width (slope ent curvature radius and sliding width as shown in
angle β = 30◦ ). Fig. 11.

847
 (a) (b)
14 14
13 wsld/H = 2.0 13 wsld/H = 5.0
nD=2.0
12 12
nD=1.0, 1.2, 1.5 nD=1.0, 1.2, 1.5, 2.0
11 11
10 Toe Failure 10
Toe Failure
9 nD=4.0 9
8 8
7 7
nD=4.0
6 6

Stability Factor, NS
All Type of Failure All Type of Failure
5 5
4 4
90 80 70 60 50 40 30 20 10 0 90 80 70 60 50 40 30 20 10 0

(c) Slope Angle, (deg.)

14
13 wsld/H = 10.0 nD : depth factor
12
11
10
9
nD=1.0
8 Toe Failure 1.2
1.5
7 2.0
4.0
6
5 All Type of Failure
4
90 80 70 60 50 40 30 20 10 0

Slope Angle, (deg.)

[ Curved Concave(-) Rt/H = 100 ]

(d) (e)
14 14
13 wsld/H = 2.0 13 wsld/H = 5.0
nD=2.0
12 12
nD=1.0, 1.2, 1.5 nD=1.0, 1.2, 1.5, 2.0
11 11
10 Toe Failure 10
9 9 Toe Failure
nD=4. 0
8 8
7 7
nD=4.0
6 6
Stability Factor, NS

All Type of Failure

5 5 All Type of Failure
4 4
90 80 70 60 50 40 30 20 10 0 90 80 70 60 50 40 30 20 10 0

(f) Slope Angle, (deg.)

14
13 wsld/H = 10.0 nD : depth factor
12

Figure 10. Relations between curvature radius and limit 11

10
sliding width, and corresponding stability factor; (a) and (b)
9
are for concave curved slope; (c) and (d) are for convex curved 8
nD=1.0
1.2
1.5
slope. 7
Toe Failure
2.0
4.0
6
5
All Type of Failure
4
4 CONCLUSIONS 90 80 70 60 50 40 30 20 10 0
Slope Angle, (deg.)
To clarify the effect of three-dimensional shapes of [ Curved Convex(+) Rt/H = 100 ]

sliding mass and slope on stability of the slope, we

conducted a series of stability analysis of straight and Figure 11. Stability diagram for curved slope (for curvature
curved slopes by means of limit equilibrium method. radius Rt /H = 100); (a) to (c) are for concave curved slope;
And we presented several slope stability diagrams (d) to (f) are for convex curved slope.
which make it possible to estimate the slope stabil-
ity considering three-dimensional effects of slope and
sliding surface. The following points were found as a
summary of this study.
base failure, but critical depth in accordance with
− In the analysis of slope stability with finite width assumed sliding width exists.
of sliding mass in three-dimensional condition, end − Though stability factor decreases with an increase
effect always increased the stability consistently of sliding width in the case of straight slope, sta-
both in straight and curved slopes. Sliding surface bility factor becomes minimal at a particular width
does not always touch firm base even in the case of of sliding mass in the case of curved slopes. At

848
the critical condition depth of sliding surface as REFERENCES
well as sliding width are determined as functions
of the curvature. As long as toe failure is con- K. Terzaghi, Theoretical Soil Mechanics, Chapter IX : Sta-
cerned, stability factor was less in convex slopes bility of Slopes: pp. 144–181.
than those in straight and concave slopes with Ugai, K. 1985, Three-Dimensional Stability Analysis of Ver-
common slope angle. In the case of base failure tical Cohesive Slopes, JGS Soils and Foundations, Vol.25,
No.3: pp. 41–48.
with small slope angles, stability factor becomes R. Baker and M. Garber 1997, Variational approach to slope
smallest in straight slopes with common slope stability, 9th ICSMFE, Vol.2: pp. 9–12.
angle.

849
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Embankment basal stability analysis using shear strength reduction

finite element method

Atsushi Nakamura
FORUM8 Co. Ltd., Meguro-ku, Tokyo, Japan

Fei Cai & Keizo Ugai

Gunma University, Kiryu, Gunma, Japan

ABSTRACT: The stability evaluation of the slope commonly uses limit equilibrium methods, for example,
Fellenius’ method in Japan. The limit equilibrium methods have two limitations. (1) It is necessary to divide
the sliding mass into slices, and to set up some additional equations to make the problem statically determinate.
(2) It is necessary to search sliding surface with a minimum safety factor. Shear strength reduction finite element
method (SSR-FEM) can solve the limitations of the limit equilibrium methods. This paper inspects superiority
of the slope stability analysis by SSR-FEM. And, as a model case, the failure of base ground of the embankment
is considered. Comparing the calculated results of SSR-FEM with limit equilibrium method, we point out the
problems inherent in limit equilibrium method for the model case when the slip surface is assumed to be circle.
We propose an approach to use SSR-FEM for slope stability analysis in practical design.

1 INTRODUCTION soil layer is changed, and the slip surface are analyzed
using SSR-FEM. Using these obtained slip surfaces,
The stability evaluation of the slope commonly uses we arranges the extent that the slip surface become a
limit equilibrium methods, for example, Fellenius’ circular arc or a non-circular arc. It is the main one
method, in most design codes and standards in Japan. of objectives t the stability examination that uses the
The limit equilibrium methods have two limitations. SSR-FEM from the stability examination by the limit
(1) It is necessary to divide the sliding mass into slices, equilibrium method in this report.
and to set up some additional equations to make the
problem statically determinate. (2) It is necessary to
search sliding surface with a minimum safety factor. 2 SSR-FEM FOR STABILITY ANALYSIS
Shear strength reduction finite element method (SSR-
FEM) can solve the limitations of the limit equilibrium 2.1 Basic concept of SSR-FEM
methods. In the finite element method with shear strength
SSR-FEM can obtain the safety factor and slip sur- reduction technique (SSR-FFM), a non-associated
face without the analyzer to assume any particular elasto-plastic constitutive law is adopted, where the
shape about the slip surface. SSR-FEM begins to be Mohr-Coulomb yield criterion is used to define the
used to analyze the slope stability though it is not com- yield function.
monly in the design. The limit equilibrium methods
are still used in almost cases. It is because the limit σ1 − σ 2 σ 1 − σ2
f = − c cos φ  − sin φ  (1)
equilibrium methods are simple. However, the limit 2 2
equilibrium methods are not fully verified for the fol-
lowing items. (1) Whether it is correct to assume that and the Drucker-Prager criterion to define the plastic
the slip surface is one circular arc. (2) How wide is the potential.
range of the grid for the center of slip circular arc to 
search the minimum safety factor. g = −αI1 + J2 − κ (2)
Then, embankment base failure of the support
ground of the fill was taken up as a model case in this Where
report. When the support ground consists of clayey soil tan ψ 3c
layer and bearing stratum, the thickness of the clayey α=  , κ=  (3)
9 + 12 tan2 ψ 9 + 12 tan2 ψ

851
 In Equation 1 and 2, c , φ  , and ψ are the effective condition, the increment of the shear strength reduc-
cohesion, friction angle, and dilatancy angle, respec- tion factor will make stresses at more Gaussian points
tively. I1 and J2 are the first invariant of the effective reach the yielding condition because of the residual
stress, and the second invariants of the deviatoric force induced by the decrease in the shear strength
stress, respectively. σ1 and σ3 are the major and minor soils.
principal effective stress, respectively. The shear strength reduction factor F increases
The global safety factor of slope in SSR FEM iden- incrementally until the global failure of the slope
tical to the one in limit equilibrium methods. The reaches, which means that the finite element calcu-
reduced strength parameters cF and φF are defined by lation diverges under a physically real convergence
criterion. The lowest factor of safety of slope lies
c F = c /F, φ  F = tan−1 (tan φ  /F) (4) between the shear strength reduction factor F at which
the iteration limit is reached, and the immediately pre-
vious one. The procedure described hereby can predict
In SSRFEM, firstly, the initial stresses in slope are
the factor of safety within one loop, and can be easily
computed using the elastic finite element analysis. The
implemented in a computing code.
vector of externally nodal forces consists of three parts:
One of the main advantages of SSRFEM is that
(1) surface force; (2) body force (total unit weight of
the safety factor emerges naturally from the analysis
soils); and (3) pore water pressure. Secondly, stresses
without the user having to commit to any particular
and strains are calculated by the elasto-plastic finite
from of the mechanism a priori. When the slope sta-
element analysis, where the reduced shear strength
bility is evaluated with the effective stress method, the
criterion. The shear strength reduction factor F is ini-
pore water pressure is usually predicted with the finite
tially selected to be so small, for example 0.01, that the
element seepage analysis or Biot’s consolidation the-
shear strength is large enough to keep the slope in elas-
ory. If the same finite element mesh is used for the
tic stage. Stresses at some Gaussian points reach the
seepage or consolidation analysis and SSRFEM, the
yielding condition with the shear strength reduction
water pressure, predicted in the seepage or consolida-
factor F in Equation 3 increased gradually. When the
tion analysis, can be directly used in SSRFEM. This
stress at anyone Gaussian point reaches the yielding
can simplify the slope stability analysis, and consider
more accurately the influence of the seepage force.

2.2 Predominance of SSRFEM

This example was firstly reported by Arai and Tagyo
and concerns a layered slope (Figure 3), where a layer
of low resistance is interposed between two layers of
higher shear strength. The material properties of var-
ious layers are listed in Table 1. Arai and Tagyo used
Figure 1. A point. the conjugate gradient method to search for the critical
slip surface and used simplified Janbu’s method to cal-
culate the safety factor, and obtained the lowest safety
factor of 0.405. This problem has also been exam-
ined by Sridevi and Deep using the random-search
technique RST-2, by Greco using pattern search and
Monte Carlo method, and by Malkawi et al. using

Figure 2. Mohr’s circle at A point. Figure 3. Slope in example.

852
Monte Carlo method (random walking). The slip sur- 3 APPLICATION OF SSR-FEM
faces located by various investigators are of significant
difference. In this study, the problem is analyzed and 3.1 Embankment basal failure model
Table 2 summarizes the results. Figure 4 shows the
This paper compared limit equilibrium methods
slip surface determined by SSRFEM. It is clear that
(Bishop’s simplified method) and SSR-FEM (ψ = φ)
the slip surface by SSRFEM is closer to that located
about embankment base failure of the support ground
by Greco.
of the fill. Figure 5 shows the model embankment. The
embankment was assumed 20.0 m in width of levee
crown, 10.0 m in height (h1). In two levels of strata,
Table 1. Material properties for example. the upper layer was soft, the lower layer was strong.
The thickness (h2) of the upper layer was changed from
Layer φ  (Deg) c (kN/m2 ) γ (kN/m2 ) 2.0 m, 4.0 m, 6.0 m, 8.0 m, to 10.0 m, to compare the
results of both analytical methods.
1 12 29.4 18.82 Material properties is show in Table 3. The nec-
2 5 9.8 18.82 essary matrix of a plasticity calculation was defined
3 40 294.0 18.82 using Young’s modulus and Poisson’s ratio. The influ-
ence of these coefficients before the failure is great,
but the influence on the total safety factor is very small
(Zienkiewicz et al.1975, Griffiths and Lane 1999).
Therefore, we assumed Young’s modulus of 2 ×
Table 2. Minimum safety factors for example. 104 kN/m2 , Poisson’s ratio of 0.3 regardless of the
Method for Method for
soils. As for the embankment, ϕ = 35.0 degrees,
safety factor slip surface Safety factor the adhesive strength of the soft layer assumed
35 kN/m2 , the adhesive strength of the bearing stratum
Arai and Tagyo (1985) 100 kN/m2 .
··
Simplified Janbu Conjugate 0.405, 0.430 ·×
·
gradient
Sridevi and Deep (1992)
·· 3.2 Analysis result
Simplified Janbu RST-2 0.401, 0.423 ·×
·
Greco (1996) 3.2.1 h2 = 2.0 m
Spencer Pattern search 0.388 Figure 6(a) shows an analysis result of h2 = 2.0 m. In
Spencer Monte Carlo 0.388 SSR-FEM, the slip surface occurred in soft layer base
Malkawai et al. (2001) and was a non-circular arc. In contrast, circular slip
Spencer Monte Carlo 0.401 was searched by the limit equilibrium method located
Rocscience Inc. (2002)
at the embankment slope. The sliding mechanics was
Spencer Random search 0.401
·· different. As for the safety factor, SSR-FEM gave 1.24,
Simplified Janbu Random search 0.410, 0.434·×
· and the limit equilibrium method gave 1.21.
Kim et al. (2002)
Spencer Random search 0.44
Lower-bound Automatic 0.40
Upper-bound Automatic 0.45 20.0
Cai et al. (2003)
SSRFEM (ψ = 0) Automatic 0.417
SSRFEM (ψ = φ) Automatic 0.423 10.0
softground
··
·×
· Corrected the safety factor calculated by simplified Janbu. bearing stratum

Figure 5. Embankment basal destruction model.

Table 3. Material Properties for Test Model.

Layer φ (Deg) c(kN/m2 )

Embankment 35.0 0.00

Soft ground 0.0 35.0
Bearing stratum 0.0 100.0
Figure 4. Slip surface by SSRFEM for example.

853
 Figure 6(b). Result of SSRFEM and LEM h2 = 4.0 m.

Figure 6(a). Result of SSRFEM and LEM h2 = 2.0 m.

3.2.2 h2 = 4.0 m
Figure 6(b) shows an analysis result of h2 = 4.0 m.
In SSR-FEM and limit equilibrium method, the slip
surface occurs in soft layer base and become a non-
circular arc. As for the safety factor, SSR-FEM gave
1.16, and limit equilibrium method gave 1.22, and
some differences occurred between the two methods.
3.2.3 h2 = 6.0 m
Figure 6(c) shows an analysis result of h2 = 6.0 m.
In SSR-FEM and limit equilibrium method, the slip
surface occurs in soft layer base. As for the safety
factor, SSR-FEM gave 1.16, and LEM gave 1.22.
3.2.4 h2 = 8.0 m
Figure 6(d) shows an analysis result of h2 = 6.0 m.
In SSR-FEM and limit equilibrium method, the slip
surface occurs in soft layer base. As for the safety
factor is both 1.10.
3.2.5 h2 = 10.0 m
Figure 6(e) shows an analysis result of h2 = 10.0 m.
In SSR-FEM and limit equilibrium method, the slip
surface occurs in soft layer base. As for the safety Figure 6(c). Result of SSR-FEM and LEM (h2 = 6.0 m).
factor is both 1.08.

Figure shows that both analysis results consistent

3.3 The comparison of the analysis result
with each other when the thickness of the soft layer is
Table 4 shows the results of both methods. Figure 7 as above 6.0 m (h2/h1  0.6).
shows the results between the safety factor and the ratio However, in the situation that a soft layer is thin,
of thickness (h2) of the soft layer and embankment a difference occurs in SSRFEM and limit equilibrium
height (h1 = 10.0 m). method.

854
 Table 4. An analysis result table every layer thickness.

h2(m) 2.00 4.00 6.00 8.00 10.00

h2/h1 0.20 0.40 0.60 0.80 1.00

SSR 1.211 1.165 1.132 1.097 1.075
LEM 1.240 1.220 1.130 1.100 1.080

1.300
1.250
1.200
1.150 SSR

Fs
1.100 LEM
1.050
1.000
0.950
0.200 0.400 0.600 0.800 1.000
h2/h1

Figure 7. A figure of comparison.

Figure 6(d). Result of SSR-FEM and LEM (h2 = 8.0 m).

Slip surface in the straight line occurs to go along
the soft layer lower part from a result of SSR-FEM. In
other words this case becomes the non-circular arc.
The limit equilibrium method assumes that it is
circular arc sliding. As a result, it is thought that a
safety factor searched the lowest circular arc sliding.
However, the limit equilibrium method was not able
to search the smallest safety factor and the sliding
shape, because a safety factor of SSR-FEM is 1.21,
and a safety factor of the limit equilibrium method
is 1.24.
In h2 = 4.0 m, embankment base failure occurs
both in SSRFEM and limit equilibrium method. The
slip surface occurs in soft layer base and become a non-
circular arc. As for the safety factor, the latter was 1.16
former 1.22. Among both, some differences occur. In
h2 = 6.0 m are regarded as non-circular arc about
the sliding shape in SSR-FEM about 8.0, 10.0 m as
follows. But, the ratio that the tangent holds shrinks
as a soft layer thickens. Therefore, for a safety factor,
is approximately equal.
Table 5 shows the result that measured the length of
the straight line of the non-circular arc in SSR-FEM.
According to the list, the length of the straight
line gets longer so that the thickness of the soft layer
increases, but the ratio for the layer thickness under-
stands a thing becoming small. In other words the glide
Figure 6(e). Result of SSR-FEM and LEM (h2 = 10.0 m). plane suffers from the shape that is near to an arc so
that a soft layer is thick.
When a soft layer is very thin h2 = 2.0 m(h2/h1 = Figure 8 is a result of non-arc sliding by the limit
0.2), in SSR-FEM, bottom failure occurs. On the other equilibrium method that considered a straight line cal-
hand, embankment slope failure occurs by limit equi- culated by SSR-FEM about layer thickness 4.0 m. It
librium method. The calculated failure mechanics was was 1.17, and the safety factor almost equal to that of
different with each other. SSR-FEM.

855
Table 5. Length of the straight line every layer thickness. Although the series of analyses was performed, a
good parameter was not found to indicate the layer
h2(m) 2.00 4.00 6.00 8.00 10.00 thickness of the soft layer under which non-arc sliding
should occur. A future search should be performed.
L1 (m) 6.3 6.3 7.0 7.5 7.5
At present, we advise that the sliding shape was cal-
L1/h2 3.15 1.57 1.16 0.93 0.75
culated using SSR-FEM, the safety factor of the calcu-
lated slip surface was evaluated using limit equilibrium
method wanted by design codes and standards.

REFERENCES

Arai, K. & Tagyo, K (1985): Detemination of noncircular

slip surface giving the minimum factor of safaty in slop
stability analysis. Soils and Foundation., The Japanese
Geotechnical Society, Vol 1.25, No.1, pp.43–51.
Sridevi, B. & Deep, K. (1992): Application of globalopti-
mization technique to slop-stability analysis, Proc. 6th
Inter. Symp. on Landslides, pp.573–578.
Figure 8. A non-arc by limit equilibrium method slips. Greco, V. R. 1996. Efficient Monte Carlo technique for locat-
ing critical slip surface, J. Geotech. Eng. Div., ASCE, 122
(7): 517–525.
4 CONCLUSION Malkawi, A.I.H., Hassan, W.F. & Sarma, S.K. (2001): Global
search method for locating general slip surface using
Monte Carlo technigues, J. Geotech. Geoenviron. Eng.,
By this report, we compared an analysis result of SSR- Vol.127, No.8, pp.668–698.
FEM with the limit equilibrium method. With that in Rockscience Inc. (2002): Verification manual for Slide, 2D
mind, we wanted to grasp such soil layer condition that limit equilibrium slop stability for soil and rock slopes,
sliding surface became the non-arc from the analysis Version 4.0, Rocksicence Inc., Canada
result of SSR-FEM. Kim, J., Salgado, R. & Lee, J. (2002): Stability analysis
It is necessary to give the center of slip circular act of complex soil slopes using limit analysis, J. Geotch.
by the judgment of the designer after having supposed Gepenvion. Eng., Vol. 128, No. 7, pp.547–557.
sliding shape to be an arc by the limit equilibrium Cai, F. & Wakai, A. (2003): Use of finite element method in
landslide analysis., The Japan Landslid Society, Vol. 40,
method. Therefore, the limit equilibrium method can-
No. 3. pp.76–80.
not cope with a non-arc even if the minimum safety Griffiths, D.V. & Lane, P.A. (1999): Slope stability anal-
factor was able to search the circular act. ysis by finite elements, Geotechnique, Vol. 49, No.,
SSR-FEM does not have to assume an examination pp. 387– 403.
condition, in a given condition, there is the big advan- Zienkiewicz, O.C., Humpheson, C. & Lewis, R.W. (1975):
tage that it is slippery, and can search shape to become Associated and non-associated visco-plasticity and plas-
the minimum safety factor. ticity in soul mechanics, Geotechnique, Vol. 25, No. 4,
When the thickness of the layer was thin, SSR-FEM pp.671–689.
could search a non-arc as sliding line associated with Cai, F. & Ugai, K. (2001): The suggestion of the method
to evaluate a total safety factor by elasto-plastic FEM
the minimum safety factor, and, in the example, the
of the slope having the measures mechanic. Soils and
difference with the limit equilibrium method became Foundation., Tokyo, 149 (4): 16–18 (in Japanese).
great.
By this report, SSR-FEM analyses were performed
to find the parameter which was ascertained whether
the sliding shape was non-circular. The soft layer
thickness was changed and a series of analysis was
carried out.

856
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Back analysis based on SOM-RST system

H. Owladeghaffari
Department of Mining & Metallurgical Engineering, Amirkabir University of Technology, Tehran, Iran

H. Aghababaei
Faculty of Mining Engineering, Sahand University of Technology, Tabriz, Iran

ABSTRACT: This paper describes application of information granulation theory, on the back analysis of
‘‘Jeffrey mine- southeast wall-Quebec’’. In this manner, using a combining of Self Organizing Map (SOM) and
Rough Set Theory (RST), crisp and rough granules are obtained. Balancing of crisp granules and sub rough
granules is rendered in close-open iteration. Combining of hard and soft computing, namely Finite Difference
Method (FDM) and computational intelligence and taking in to account missing information are two main
benefits of the proposed method. As a practical example, reverse analysis on the failure of the southeast wall-
Jeffrey mine—is accomplished.

1 INTRODUCTION function). Application of direct approach, to evalu-

ation of problem based non-linearity and employing
Back analysis is a reverse procedure, which is to of heuristic optimization methods such Ant Colony
solve the external load or partial material parameters, (ACO), Particle Swarm (PSO), and Simulated Anneal-
based on the known deformation and stresses at lim- ing (SA) are two main results. So, back analysis based
ited points and the partially known material parameters on interactions of hard computing and soft comput-
(Zhu & Zhao, 2004). ing methods can be settled in direct approach. As an
In most geotechnical engineering problems, it is initial over view, 1–1 mapping methods can be sup-
often necessary to know the in situ field, mate- posed as hard computing such analytical, numerical
rial mechanical parameters, and even the mechanical or hybrid methods, whereas soft computing methods
model, by utilizing the monitored physical information (SC), as main part of not 1–1 mapping methods, is
such as deformation, strain, stress, and pressure during a coalition of methodologies which are tolerant of
construction. The back analysis method, based on the imprecision, uncertainty and partial truth when such
required input physical deformation, can be divided tolerance serves to achieve better performance, higher
in to deformation back analysis method, stress back autonomy, greater tractability, lower solution cost.
analysis method and coupled back analysis method The principles members of the coalition are fuzzy
(Sakurai et al, 2003). The complex feature of rock logic (FL), neuro computing (NC), evolutionary com-
mass, associated structures and the difficulty of inter- puting (EC), probabilistic computing (PC), chaotic
pretation of the interaction of excavation and rock computing (CC), and machine learning (Zadeh, 1994).
mass are the main reasons to deploying of back anal- Utilizing of soft computing methods besides hard
ysis methods in rock engineering. Generally, two computing methods, such support vector machine
main procedures of back analysis methods have been (Feng et al, 2004) and neural network (Pichler et al,
pointed in the literature: inverse and direct approach 2003), in the reverse procedures have been expe-
(Cividini et al, 1989). In reverse method, the mathe- rienced, successfully. Advancing of soft computing
matical formalization is inverse of ordinary analysis. methods, under the general new approach, namely
In this process, the number of monitored data is more information granulation theory, has been opened new
than the unknown parameters, so that by utilizing of horizons on the knowledge discovery data bases.
optimization techniques, the unknown parameters are Information granules are collections of entities that
obtained. The main advantage of such method is an are arranged due to their similarity, functional adja-
independent spirit in regard of the iteration operations, cency, or indiscernibility relation. The process of
which causes to reducing of calculation time, sensi- forming information granules is referred to as infor-
bly. Direct approach, is a process associated with the mation granulation (Zadeh, 1997). There are many
cycling optimization (aim is to minimize the fitness approaches to construction of IG, for example Self

857
Organizing Map (SOM) network, Fuzzy c-means
(FCM), rough set theory (RST). The granulation level,
depend on the requirements of the project. The smaller
IGs come from more detailed processing. On the other
hand, because of complex innate feature of informa-
tion in real world and to deal with vagueness, adopting
of fuzzy and rough analysis or the combination form
of them is necessary. In fact, because of being the
complex innate of crisp, rough or fuzzy attributes, in
the natural world, extraction of fuzzy or rough infor-
mation inside the crisp granules, can give a detailed
granulation. To develop back analysis based new
soft computing approaches and taking in to account
of a mathematical tool for the analysis of a vague Figure 1. Kohonen’s SOM.
description of objects, as well as rough set theory,
a combining of self organizing map-neural network-
(SOM), and rough set theory (RST) associated with For supervised learning, both input and output are
the hard computing methods is proposed. necessary for training the neural network, while the
The rough set theory introduced by Pawlak (1982, unsupervised learning needs only the inputs. In unsu-
1991) has often proved to be an excellent mathemat- pervised learning, neural network models adjust their
ical tool for the analysis of a vague description of weights so that input data can be organized in accor-
object. The adjective vague, referring to the quality of dance with statistical properties embedded in input
information, means inconsistency, or ambiguity which data. Kohonen’s SOM includes two layers such as input
follows from information granulation. The rough set layer and mapping layer, in the shape of a one or two-
philosophy is based on the assumption that with every dimensional grid. The number of nodes in the input
object of the universe there is associated a certain layer is equal to the number of features associated with
amount of information, expressed by means of some input data. The mapping (output) layer acts as a distri-
attributes used for object description. The indiscerni- bution layer. Each node of the mapping layer or output
bility relation (similarity), which is a mathematical node also has the same number of features as there
basis of the rough set theory, induces a partition of the are input nodes. Both layers are fully connected and
universe in to blocks of indiscernible objects, called each connection is given an adjustable weight. Fur-
elementary sets, which can be used to build knowl- thermore, each output node of the mapping layer is
edge about a real or abstract world. Precise condition restricted smaller distance around the cluster center.
rules can be extracted from a discernibility matrix. The For a Kohonen’s SOM, suppose that the number of
rest of paper has been organized as follow: a part from output nodes is m, the number of input nodes is n,
details, section 2 covers a brief on the RST and SOM and wi = (wi1 , wi2 , . . ., wim ) (1 ≤ i ≤ m) is the con-
and in section 3 the details of the proposed algorithm nection weight vector corresponding to output node i.
has been presented. Finally, we discuss an example in wi can be viewed as a center of the cluster i. when-
which we show how the proposed technique behaves. ever new input data x = (x1 , x2 , . . ., xn ) is presented to
SOM during the training phase, the output value for
output node i is computed by square of the Euclidean
2 RST & SOM distance denoted by oi between x and wi , as shown in
Equation 1:
Here, we present some preliminaries of self organizing
feature map—neural network- and rough set theory 
n

that are relevant to next section. oi = (di )2 =  x − wj  2 = (xj − wij )2 ,

j=2

1≤i≤m (1)
2.1 Self organizing map-neural network
Kohonen’s (1987) SOM algorithm has been well If the node i∗ satisfies Equation 2 then it is declared
renowned as an ideal candidate for classifying input as a winner node:
data in an unsupervised learning way.
Figure 1 shows Kohonen’s SOM topology, where (di∗ )2 = min oi , 1≤i≤m (2)
two layers such as input layer and mapping layer
are working together to organize input data into an Adjustable output nodes including the winning
appropriate number of clusters (or groups). node i∗ and its neighbor nodes are determined by the

858
neighborhood size of the winning node i∗ (|i∗ |). Sub- containing X To constitute reducts of the system, it will
sequently, connection weights of the adjustable nodes be necessary to obtain irreducible but essential parts
are all updated. The learning rule of SOM is shown in of the knowledge encoded by the given information
Equation 3: system. So one is, in effect, looking for the maxi-
mal sets of attributes taken from the initial set (A, say)
wij = η(xj − wij ), i ∈ | i∗ |, 1≤j≤n (3) that induce the same partition on the domain as A. In
other words, the essence of the information remains
Where η is a learning rate. To achieve a bet- intact, and superfluous attributes are removed. Con-
ter convergence, η and the neighborhood size of the sider U = {x1, x2, . . ., xn} and A = {a1, a2, . . ., an} in
winning node, should be decreased gradually with the information system S = U, A. By the discerni-
learning time. bility matrix M(S) of S is meant an n∗ n matrix such
To achieve a better convergence, and the neighbor- that cij = {a ∈ A : a(xi )  = a(xj )}
hood size of the winning node, should be decreased A discernibilty function fs is a function of m Boolean
gradually with learning time. SOM has been success- variables a1, . . ., am corresponding to attributes
fully employed in different fields of applied science. a1, . . ., am, respectively, and defined as follows:
Specially, in geomechanics, for example, in clustering
of lugeon data (Shahriar & Owladeghaffari, 2007) and fs (a1 , . . . , am )
joint sets (Sirat & Talbot, 2001).
= ∧{∨(cjj ) : 1 ≤ i, j ≤ n, j ≺ i, cij  = ϕ}

2.2 Rough set theory where ∨(cij ) is the disjunction of all variables with a.
a ∈ cij (Pal et al, 2004). With such discriminant matrix
The rough set theory introduced by Pawlak (Pawlak, the appropriate rules are elicited.
1991) has often proved to be an excellent mathematical The existing induction algorithms use one of the
tool for the analysis of a vague description of object. following strategies:
The adjective vague referring to the quality of informa-
tion means inconsistency, or ambiguity which follows a. Generation of a minimal set of rules covering all
from information granulation. The rough set philoso- objects from a decision table;
phy is based on the assumption that with every object b. Generation of an exhaustive set of rules consisting
of the universe there is associated a certain amount of of all possible rules for a decision table;
information, expressed by means of some attributes c. Generation of a set of ‘strong’ decision rules,
used for object description. The indiscernibility rela- even partly discriminant, covering relatively many
tion (similarity), which is a mathematical basis of the objects each but not necessarily all objects from the
rough set theory, induces a partition of the universe in decision table (Greco et al, 2001). In this study we
to blocks of indiscernible objects, called elementary use first approach in combining with initial crisp
sets, which can be used to build knowledge about a granules, adaptively.
real or abstract world. Precise condition rules can be
extracted from a discernibility matrix. 3 THE PROPOSED METHOD BASED
An information system is a pair S = U, A, where ON SOM-RST SYSTEM AND HARD
U is a nonempty finite set called the universe and A COMPUTING METHODS
is a nonempty finite set of attributes. An attribute
a can be regarded as a function from the domain U Figure 2 shows a general procedure, in which the
to some value set Va. An information system can be information granulation theory accompanies by a pre-
represented as an attribute-value table, in which rows defined project based rock engineering design. After
are labeled by objects of the universe and columns by determining of constraints and the associated rock
attributes. With every subset of attributes B⊆A, one engineering regards, the initial granulation of infor-
can easily associate an equivalence relation IB on U: mation as well as numerical (data base) or linguistic
format is accomplished. Developing of modeling
IB = {(x, y) ∈ U : for every a ∈ B, a(x) = a(y)} instruments based on IGs, whether in independent
or associated shape with hard computing methods
Then, IB = ∩a∈B Ia . (such fuzzy finite element, fuzzy boundary element,
If X ⊆ U , the sets {x ∈ U : [x]B ⊆ X } and stochastic finite element . . .) are new challenges in the
{x ∈ U : [x]B ∩ X = ϕ}, where [x]B denotes the current discussion.
equivalence class of the object x ∈ U relative to IB, Thus, one can employ such method as a new mythol-
are called the B-lower and the B-upper approximation ogy in designing of rock engineering flowcharts. The
of X in S and denoted by BX and BX , respectively. It main benefits are considering of the roles of the
may be observed that BX is the greatest B-definable set expert’s experiences and educations, the missing or
contained in X and BX is the smallest B-definable set vagueness information and utilizing of advantages of

859
 Figure 3. A proposed procedure based on hard computing
and SOM-RST system.

Figure 2. A general methodology for back analysis based

on information granulation theory.

soft and hard computing methods. A new advantages First step is to collect data sets by using 1-1
of the mentioned method, is to considering of the out- modeling (here, hard computing methods as one of
ward changes in ‘‘words format’’ and ‘‘calculation the main part of the direct modeling methods are
with words and perception’’ (CWP). selected). Applying of SOM as a preprocessing step
Under mentioned methodology and to develop and discretization tool is second process. For contin-
modeling instruments, authors have (are) proposed uous valued attributes, the feature space needs to be
(proposing) new algorithms with neural networks, discretized for defining indiscernibilty relations and
fuzzy logic (possibility theory), rough set theory, and equivalence classes. We discretize each feature in to
meta heuristic optimization methods which are accom- three levels by SOM: ‘‘low, medium, and high’’; finer
panied by close-open world idea. Random selection of discretization may lead to better accuracy at the cost of
initial precise granules can be set as ‘‘close world’’ a higher computational load. Because of the generated
assumption (CWA). But in many applications, the rules by a rough set are coarse and therefore need to
assumption of complete information is not feasible be fine-tuned, here, we have used the preprocessing
(CWA) nor realistic, and only cannot be used. In step on data set to crisp granulation by SOM, so that
such cases, an open world assumption (OWA), where extraction of best initial granules and then rough gran-
information not known by an agent is assumed to be ules is rendered in close-open iteration. In open world
unknown, is often accepted. The aim of open-world is iteration phase, we use a simple idea associated with
to achieve complete knowledge of the universe by set human’s cognition of the surround world:‘‘simplicity
of classified rules or by modifying rules. The close- of rules’’ (dominant to the problem) whether in num-
open iteration accomplishes the balancing of crisp and bers, applied operators or/and the length of rules. Here,
sub rough granules by some random selection of initial we take in to account three main parameters in rules
granules and increment (or decreasing) of supporting generation: length of rules, strength of object and num-
rules, gradually. Figure 3, shows one of our proposed ber of rules. By setting the adaptive threshold error
algorithms, apart from contributing of FL and free- level and the number of close-open iteration, stabil-
derivative optimization methods. One may employ ity of the algorithm is guaranteed. After obtain best
other shape of open world assumption‘s implementa- rules, the monitored data set as real decision parts
tion. For instance to raise the quality of approximation are compared with the extracted rules and best con-
by much more categories, but the optimum numbers ditional parts are picked up. These parts, associated
of such scaling could be approached by an algorithm, with the reduced data sets by SOM, are the approxima-
so that an adaptive disceritization besides balancing of tion of the wanted parameters as well as back analysis
the initial granules with open world is accomplished. results.

860
4 AN ILLUSTRATIVE EXAMPLE: shear zone present. Sliding in the overburden was first
FAILURE ON THE SOUTHEAST observed, followed by local wedge failure and finally
WALL- JEFFREY MINE a major slide in 1971, involving some 33 million ton of
rock (figure 4). Failure was believed to have standard
This section describes how one can acquired approx- in the weak shear zone which then leads to failure of
imated values of the effective parameters under a upper portions of the slope movement rates of the order
reverse analysis. To this aim and by using finite dif- of 1500 m/month were recorded. After this failure,
ference method, as well as FLAC (), the failure on some reactivation of the 1971 of the 1971 year failure
the southeast wall in Jeffrey mine (Quebec) has been occurred (Sjoberg, 1996).
evaluated. By using finite difference method under FLAC4, as
The Jeffrey mine is located at Asbestos; Québec a hard computing method, almost 30 different mod-
As Asbestos fiber is mined from ultra basic host els on the mentioned slope and by considering of
rocks dominated by periditotes, dunites and serpan- Mohr-Cloumb plastic law have been constructed. The
tites. The rocks mass is intersected by several thick aim was to detect the material properties, while the
shear zones and smaller scale discontinuous. Strength slope movement rate satisfies the monitored records,
and deformability for the rock materials vary widely approximately. Table 1 shows the part of the obtained
from very soft and weak to moderato stiff and strong results while 8 effective parameters changes.
rock. In 1970, the slope height was 180 m plus 60 m of The performance of 3 ∗ 3 SOM (nx = 3, ny = 3;
overburden (clay, silt, & sand). Failure occurred in rel- Matrix of neurons −nx . ny determines the size of 2D
atively fractured, serpantinized peridotite with a major SOM.) on the maximum velocity vectors of the tested
data have been depicted in figures 5 and 6 (The dis-
critization procedure using 3 ∗ 3 SOM on the attribute
has been rendered). This step by accounting of dif-
ferent structures of SOM and in interaction with RST
was iterated. The effective rough set procedure param-
eters were selected as follows: minimum rule strength:
60%; maximum length of rule:2 and maximum num-
ber of rules: 5. It must be noticed that decreasing of
the rules is accomplished, gradually (depend on the
error level).
Error level, here, is settled in true classified test
data (percentage). Since total of analysis is low (30),
we set n = 1, EL = 80% and κ = 2 (maximum
closed-iterations for any incensement of rules).
After four closed-open iterations, algorithm satis-
fies conditions of the test data. Figure 7 illustrates the
Figure 4. Cross section through the Jeffry mine showing extracted rules which are matched with the recorded
the 1971 failure surface (Sjoberg, 1996). data by the proposed algorithm.

Table 1. Subset of the obtained results by FLAC4- Cp (kg/cm∧ 2): coherent of Periditotets; Sz: shear zone; T: tensile strength
(kg/cm∧ 2); TMD: total max. Displacement (m); MVV: max. Velocity vector (m/s).

No Cp∗ 10∧ 5 Phi p Cb Phi b C sz Phi sz Tp Tb TMD MVV

1 2.00 25 3.00E + 05 35 500 15 428901.4 42844.4 2.41E-05 4.46E-14

2 3.20 35 2.20E + 05 25 1000 5 428901.4 42844.4 5.82E-06 9.86E-16
3 2.50 30 3.50E + 05 40 1500 10 428901.4 42844.4 2.66E-06 1.84E-14
4 2.75 25 3.75E + 05 35 1500 15 428901.4 42844.4 4.27E-06 1.68E-22
5 3.00 35 4.00E + 05 45 1000 20 428901.4 42844.4 2.52E-08 2.06E-22
6 2.00 25 3.00E + 05 35 1500 15 428901.4 42844.4 2.61E-06 4.24E-14
7 2.00 25 3.00E + 05 35 1500 15 1.00E + 06 6.80E + 05 2.20E-11 4.10E-22
8 2.00 25 3.00E + 05 35 1500 5 1.00E + 06 6.80E + 05 2.58E-06 1.73E-13
9 2.00 25 3.00E + 05 35 500 15 2.71E + 06 1.13E + 06 7.80E-16 1.92E-22
10 2.00 25 3.00E + 05 35 800 7 2.71E + 06 1.13E + 06 3.71E-08 4.95E-13
11 2.00 25 3.00E + 05 35 800 8 2.71E + 06 1.13E + 06 1.76E-07 1.70E-15
12 2.00 25 3.00E + 05 35 500 8 2.71E + 06 1.13E + 06 1.16E-06 8.14E-16

861
 5 CONCLUSIONS

The role of uncertainty and vague information in

geomechnaic analysis is undeniable feature. Indeed,
with developing of new approaches in information
theory and computational intelligence, as well as
soft computing approaches, it is necessary to con-
sider these approaches within and inside of current
and conventional analysis, especially in geomechanic
field. Under this idea and to fining best informa-
tion granules which are picked off inside each other,
close-open worlds (cycle) procedure has been pro-
Figure 5. The performance of 3 ∗ 3 SOM on the maximum posed. Utilizing this algorithm in interaction with
velocity vectors of the tested data (series 1: real; series 2: finite difference method and inferring back analy-
deduced data set). sis results are main advantages of our method. So,
we could obtain a becoming approximation of the
parameters (in failure of southeast wall of Jeffry
mine -1971). It must be notice that application of
such process ensues a (or some) reduct set of the
attributes where the reduct set(s) can be supposed as
soft sensitive analysis on the model. Decreasing of
time consuming and extractions of effective parame-
ters, behind the core(s)-most effective parameters- of
attributes (in condition parts) are other benefits of our
model.

REFERENCES

Cividini, A., Jurino, L., Gioda, G. 1989.Some aspects of

characterization problems in geomechanics, Int J Rock
Mech Min Sci. 18: 487–503.
Feng, X.T., Zhao, H., Li, S. 2004. A new displacement back
Figure 6. Discritization on the TMV in three clusters: low analysis to identify mechanical geomaterial parameters
(3), medium (2), and high (1) by SOM. based on hybrid intelligent methodology. Int J Numer Anal
Meth Geomch 28: 1141–1165.
Greco, S., Matarazzo, B., Slowinski, R. 2001. Rough sets the-
ory for multi criteria decision analysis, European Journal
Rule 1. (cb<=220000.000000) => (mvv at of Operational Research 129: 1–47.
most 1); Kohonen, T. 1987. Self-organization and associate memory.
Rule 2. (phib<=25.035000) => (mvv at most 2nd ed. Springer —Verlag. Berlin .
1); Pal, K.S., Polkowski, L.A., Skowron, A. 2004. Rough-
Rule 3. (csz<=999.790000) => (mvv at most
neural computing: techniques for computing with words.
1);
springer-verlag.
Rule 4. (phisz<=5.035400) &
(tb<=42844.000000) => (mvv at most 1);
Pawlak, Z. 1982. Rough sets. Int J Comput Inform Sci 11:
341–356.
Pawlak, Z. 1991. Rough sets: theoretical aspects reasoning
Figure 7. Part of the extracted rules by the proposed about data. Kluwer academic. Boston.
algorithm. Pichler, B., Lackner, R., Mang, H.A. 2003.Back analysis of
model parameters in geotechnical engineering by means
of soft computing. Int J Numer Anal Meth Geomch 57:
1943–1978.
Sakurai, S., Akutagawa, S., Takeuchi, Shimiziu, N. 2003.
So, we could extract some best rules where the deci-
Back analysis for tunnel engineering as a modern obser-
sion parts of them satisfy the monitoring rate of the vational method. Tunneling & underground space tech-
slope. Thus, Application of such system under regular nology 18 (2–3): 185–196.
monitoring data set can be rendered. The main point in Shahriar, K., Owladeghaffari, H. 2007.Permeality analy-
such data is to granulate the real decisions and compare sis using BPF, SOM & ANFIS. 1st Canada–U.S. Rock
with the deduced decision parts rules. Mechanics Symposium (in print).

862
Sirat, M., Talbot, C.J. 2001. Application of artificial neural Zadeh, L.A. 1997. Toward a theory of fuzzy information
networks to fracture analysis at the Aspo HRL, Swe- granulation and its centrality in human reasoning and
den:fracture sets classification. Int J Rock Mech Min Sci. fuzzy logic. Fuzzy sets and systems 90: 111–127.
38: 621–639. Zane, L.A. 1994. Fuzzy logic, neural networks, and soft
Sjoberg, J. 1996. Technical report: large scale slope sta- computing. Commun ACM vol. 37: 77–84.
bility in open pit mining. LULEA University. ISSN Zhu, W., Zhao, J. 2004. Stability analysis and modeling of
0349-3571. underground excavation in fractured rocks. Elsevier.

863
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Temporal prediction in landslides – Understanding the Saito effect

D.N. Petley
Department of Geography, University of Durham, Durham, UK

D.J. Petley
School of Engineering, University of Warwick, Coventry, UK

R.J. Allison
Department of Geography, University of Sussex, Sussex, UK

ABSTRACT: Of all natural hazards, landslides offer the best potential for prediction of the time of an event,
and indeed successful predictions of the time of large-scale slope collapse have now been made for a number
of landslides. In general these predictions are based upon the use of the approach first proposed by Saito in
which the inverse of landslide strain rate is plotted against time, and the time of final failure determined by
fitting a straight line to the data, and extrapolating the time at which v−1 = 0. In this paper, an investigation
of Saito linearity is undertaken using laboratory testing. Using undisturbed samples of materials from the basal
shear region of landslides, stress path testing has been undertaken. 38 mm diameter samples have been tested
using the pore pressure reinflation (sometimes termed the field) stress path, in which the deviatoric and the
total mean stresses are kept constant, whilst pore pressure is elevated. This triggers failure, during which the
development of axial strain with time is monitored. The results clearly demonstrate that Saito linearity is the result
of brittle deformation processes associated with the formation of the shear surface within the landslide mass.
In landslides in which no shear surface forms, the linear trend is not observed. The data suggest that the state
and rate dependent friction model, which has been implicated as a mechanism in some studies, is not the
cause of this behavior. Our tests show that the linear phase of deformation in v−1 —time space is associated
with stress concentration during the late stage of shear surface formation. These results provide enhanced
understanding of the processes of shear surface formation, and the circumstances under which Saito linearity
can be observed.

1 INTRODUCTION
provision of fully engineered solutions to the large
of majority slope threats is just not viable given the
1.1 Context
limited resources, the abundance of dangerous slopes,
Landslides represent one of the most destructive of and the likelihood of environmental degradation. In
geological phenomenon, being responsible for over many of the more developed countries the use of
100,000 fatalities in the period 1990–2006 (Petley engineering solutions is often opposed due to the
et al. 2005a). The impact in terms of economic loss and incumbent damage to the environment. However, at
damage is undoubtedly substantial, possibly reach- the same time there is an increasing lack of toler-
ing as much as 2–3% of GDP per annum in many ance of risk. Thus, there is an increasing emphasis
mountainous developing countries (Brabb 1991). As being placed on the development of non-engineered
a result, a host of techniques have been developed approaches to slope problems, and in particular to
to permit the analysis of the stability of a slope and the development of reliable and robust techniques
to allow suitable engineering designs to be devised that permit the prediction of the time of final fail-
to mitigate the hazard. These techniques have proven ure of a slope, allowing the implementation of an
to be effective, such that the incidence of landslide effective warning system. In this paper, we seek
fatalities in many developed countries is now low to investigate the most widely used predictive tool,
(Petley et al. 2005a). However, such approaches are that first proposed by Saito (1965), in which the
generally expensive and are often environmentally inverse of velocity is used to estimate the time of final
insensitive. In mountainous developing countries, the failure.

865
1.2 The ‘‘Saito’’ approach to the prediction
of failure
Saito (1965) first suggested that the pattern of change
in the rate of movement of a landslide can be used to
predict the time of final failure. This was based on an
observation that the inverse of velocity (v−1 ) defines
a straight line when plotted against time in the period
leading up to failure. The time of final failure can thus
be predicted by extrapolating the linear trend to the
time when v−1 = 0, i.e. the rate of displacement is
infinite. This approach has been used successfully on
a number of occasions (Zvelebil 1984; Suwa 1991;
Hungr and Kent 1995; Rose and Hungr 2007). How-
ever, unfortunately there has also been a number of Figure 1. The generalized form of the three phase move-
occasions when this methodology has proven to be ment model.
unsuccessful (Hungr et al. 2005). The major problem
is that to date the actual mechanisms that cause the
linear trend in v−1 —time (t) space are not well under- processes dominate, although exactly how this occurs
stood. Use of this technique should only be attempted is contentious.
when these processes have been explained.

2.2 The applicability of the Saito approach

2 EXPLORING THE SAITO METHOD
Petley et al. (2002) explored the general applicability
of the Saito technique by examining the development
2.1 Landslide movement
of the final movement phase of a number of landslides.
It is well established that during the development of It was clearly shown that Saito linearity is not seen in
first time failure in landslides three distinct ‘‘phases’’ all landslides. The linear trend was only observed in
of movement (Fig. 1) are often. This pattern of move- failures in which the controlling processes were brit-
ment is often referred to as ‘‘three phase creep’’, tle, i.e. in which final failure occurred through the
although in this context the term ‘‘creep’’ may be a formation of a new shear surface. Where the con-
misnomer. In the three phase model, the key types of trolling processes were not brittle (i.e. they did not
movement are:- rely on the formation of a new shear surface), such
as when the movement occurs as a result of ductile
1. Primary movement, which represents the initiation
deformation, flow processes or sliding along a pre-
of displacement. Primary creep is characterized by
existing, (comparatively) weak surface, a non-linear
an initially high, but declining, displacement rate.
trend was noted in v−1 —t space. Thus, Petley et al.
2. Secondary movement, in which the rate of move-
(2002) and Kilburn and Petley (2003) proposed that
ment is low. The actual movement rate often fluc-
the Saito technique is only applicable to landslides in
tuates according to the stress state of the landslide,
which a shear surface is being formed. This is entirely
and may stop altogether at times.
consistent with the model for ‘‘progressive failure’’
3. Tertiary movement, in which the displacement rate
(i.e. the failure of a landslide without mobilizing the
increases rapidly as the landslide accelerates to
peak strength of the basal materials) proposed by Bjer-
final failure. It is during the tertiary movement
rum (1967). Here, first time failure of landslides in
phase that Saito linearity is observed.
materials that show distinctly different peak and resid-
The form of the three phase movement model is ual strengths (i.e. that show brittleness) was envisaged
the result of the interaction of strain hardening and to occur through the generation of a shear surface
strain softening processes. During primary creep, that grows due to stress concentration at the tip of
strain hardening dominates, resulting in reductions the cracks. Shear surface growth could occur because
in strain rate with time. In secondary creep, strain the stress concentration exceeded peak strength. Such
hardening and strain softening are in an approximate a model is consistent with the observations of Pet-
balance, although it seems likely that the evolution of ley et al. (2002) in terms of the importance of brittle
this phase reflects a slow but progressive change in processes.
dominance from hardening to softening mechanisms. Within the geophysical community a different
Thus, it is likely that the gradient of the displace- explanation is often advanced to explain three phase
ment—time graph passes through an inflection point creep and Saito linearity. This is a concept of rate-
during this phase. In tertiary creep, strain softening and state-dependent friction, in which the resistance

866
to movement is considered to reduce as strain rate Table 1. Experimental parameters for the Foxmould tests.
increases (Helmstetter et al. 2003; Helmstetter and
Sornette 2004). Such a model is not consistent with Experiment State Type p0 PPR rate
the model proposed by Bjerrum (1967). Clearly there
K0 Undisturbed K0 n/a n/a
is a need to investigate this behavior further both to
MDC100 Undisturbed MDC 100 n/a
determine which of the crack growth and the state- and
MDC200 Undisturbed MDC 200 n/a
rate-dependent friction models is correct and, more
MDC400 Undisturbed MDC 400 n/a
importantly, to provide a better basis for the applica-
MDC600 Undisturbed MDC 600 n/a
tion of the Saito technique to the provision of warning
MDC1000 Undisturbed MDC 1000 n/a
systems. As full scale monitoring of the failure of land-
slides is problematic, the best way to approach this is PPRU100 Undisturbed PPR 100 10 kPa h−1
to use laboratory experimentation. PPRU200 Undisturbed PPR 200 10 kPa h−1
PPRU400 Undisturbed PPR 400 10 kPa h−1
PPRU1000 Undisturbed PPR 1000 10 kPa h−1
PPRD100 Remoulded PPR 100 10 kPa h1
3 METHODOLOGY PPRI100 Undisturbed PPR 100 10 kPa h−1

To investigate the behavior of materials in the basal NB. PPR = pore pressure reinflation.
regime of a landslide we have undertaken an experi-
mental study on samples of the Foxmould. This is a
unit within an Upper Cretaceous greensand formation These latter two tests were briefly described in Petley
found on the south coast of England that consists of et al. (2005a), although a detailed analysis was not
a yellowish-brown, weakly-bonded, glauconitic sand- provided.
stone. In terms of landslides the significance of this
material is that it is the unit containing the shear
zone of the controlling rotational component of the
Black Ven landslide complex in southern England 4 RESULTS
(Brunsden 1969; Petley et al. 2005b). In this pro-
gramme of tests we have undertaken ‘pore pressure 4.1 Conventional tests
reinflation’ (field stress path) tests, in which the sam- Clear peak and residual strength envelopes were
ple is consolidated and then subjected to drained shear defined by the MDC tests, both having a curved form
to a pre-determined stress/strain state. The key part in q-p space (Fig. 2). A Mohr’s circle analysis showed
of the test occurs at this point, when the sample is that the Foxmould has a comparatively low low cohe-
subjected to increasing pore pressure under constant sion intercept (c = 35 kPa), but with a peak strength
total stress conditions. In these tests we have used effective angle of internal friction of 34◦ , which is typ-
38 mm × 76 mm cylindrical undisturbed cores of ical for a weakly-cemented sand. The residual strength
Foxmould, prepared using a soils lathe. Testing has envelope indicates no cohesion and an effective angle
been undertaken in a standard GDS stress path cell of friction of 29◦ . At failure each sample displayed a
(Menzies 1988). set of conjugate shear surfaces.
In most of the tests reported here we have consoli-
dated the sample along a K0 stress path as this produces
the most realistic stress and strain state (Table 1),
4.2 Pore pressure reinflation tests
although two PPR tests were undertaken after isotropic
consolidation. We have undertaken five monotonic Experiments PPRD100 and PPRU100 were under-
drained compression (MDC) experiments at a dis- taken to compare the behavior of an undisturbed and
placement rate of 0.05 mm/hour in order to define the a remoulded sample of the Foxmould. At the start of
peak and residual strength envelopes. Six pore pres- both experiments the density of the samples was the
sure reinflation experiments have been completed, in same. Isotropic consolidation was used for these two
which pore pressure was increased until the sample samples as the different K0 stress paths that the sam-
failed. In four of these (PPRU100, 200, 400, 1000) ples would follow would prevent comparison between
pore pressure reinflation was undertaken from the K0 them. The results of the PPR portion of these tests are
stress path. An additional two experiments (PPRD100 shown in Fig. 3.
and PPRI100) were conducted in which the samples It is clear that the shape of the axial strain—-
were isotropically consolidated to p = 100 kPa before time plot is quite different for these two samples.
being subjected to drained shear at 0.004 mm min−1 to First, the undisturbed sample failed much later than
a deviatoric stress of 105 kPa, at which point pore pres- did the remoulded sample. This is unsurprising given
sure was increased at 10 kPa h−1 until failure occurred. that it was much weaker than was the undisturbed

867
 1600
Peak strength 60000 PPRI100 Undisturbed
1400 envelope PPRD100 Remoulded
Residual strength
envelope 50000

1 / Axial strain rate (secs/%)

Deviator stress (kPa)

1200

1000
40000
800 k0 envelope
1000
30000
600
800
400
400 20000
200 200
100
10000
0
0 200 400 600 800 1000 1200 1400
Mean effective stress (kPa) 0
0 5000 10000 15000 20000
Time (min)
Figure 2. Stress path diagram for the MDC and K0 stress
path experiments, showing the peak, residual and K0 lines.
Figure 4. 1/Axial strain rate—time plot for the isotropically
consolidated PPR tests.

20
PPRI100 Undisturbed 2
18 PPRD100 Remoulded
1.8
16
1.6
Axial strain (%)

14
12 Axial strain (%) 1.4

10 1.2
Tertiary phase
8 1
6 0.8
4 Secondary phase
0.6
2 0.4 Primary phase
0
0.2
0 5000 10000 15000 20000
Time (secs) 0
0 5000 10000 15000 20000
Time (secs)
Figure 3. Axial strain—time plot for the isotropically
consolidated PPR tests.
Figure 5. Axial strain—time plot for the first 2% of axial
strain for the undisturbed isotropically consolidated PPR test,
showing three phase creep.
sample. Note also though that the undisturbed sam-
ple fails over a much shorter time period than did 1000 Peak strength Residual strength
the remoulded sample. This is evident in the Saito 900 envelope envelope
plot as the undisturbed sample shows the linear trend 800
1000
whilst the undisturbed sample shows the non-linear
Deviator stress (kPa)

700
trend described by Petley et al. (2002) (Fig. 4). This 600
k0 envelope

suggests that the linear trend is associated with the 500

presence of bonding in the sample, as this is the key 400
400
difference between the samples, as suggested by Petley 300
et al. (2005b). 200 200
Interestingly, in a plot of the development of axial 100 100
strain against time, the undisturbed sample clearly 0
shows the three phase movement pattern (Fig. 5). 0 200 400 600 800 1000 1200
Mean effective stress (kPa)
This is entirely consistent with the observed patterns
of movement in real landslide systems, suggesting
that this type of stress path testing is suitable for the Figure 6. Stress path diagram for the PPR experiments,
showing the peak, residual and K0 lines. Sample numbers
simulation of landslide movements. (value of p0 ) are indicated.
The stress paths for the four PPR experiments are
shown in Fig. 6. In each case the deviatoric stress was
successfully controlled until failure was initiated. In of internal friction than does the conventional failure
every case final failure occurred at a mean effective envelope.
stress value slightly lower than that for the conven- The pattern of axial strain against time for the
tional tests—i.e. the experiments seem to imply that 1000 kPa PPR experiment is shown in Figs. 7
the PPR failure envelope has a slightly higher angle and 8. Once again a clear primary movement phase is

868
 1.00 1600000
0.90 Secondary phase
1400000
0.80
Primary phase

1 / axial strain rate (secs / %)

1200000
0.70
Axial strain (%)

0.60 1000000
0.50 Secondary phase Tertiary phase
800000
0.40
0.30 600000
Primary phase
0.20 400000
0.10
200000
0.00
60000 90000 120000 150000 180000 0
Time (secs) 60000 90000 120000 150000 180000
Time (secs)

Figure 7. Axial strain against time (0–1% strain) for the

1000 kPa experiment showing the primary and secondary Figure 9. Saito plot for the 1000 kPa experiment showing
movement phases. the primary, secondary and tertiary movement phases. Note
the non-linear form of the secondary movement phase.

3.00

2.50
800000
Tertiary phase
2.00
Axial strain (%)

700000 Secondary phase

1 / axial strain rate (secs / %)

1.50 600000
Tertiary phase
Secondary phase
500000
1.00

400000
0.50
300000
0.00
200000
60000 90000 120000 150000 180000
Time (secs)
100000

0
Figure 8. Axial strain against time (0 to 3% strain) for the
120000 150000 180000
1000 kPa experiment showing the secondary and tertiary Time (secs)
movement phases.
Figure 10. Saito plot for the 1000 kPa experiment showing
the onset of the tertiary movement phase.
observed (Fig. 7), which is consistent with that seen in
the isotropically consolidated sample described above.
The final failure is rapid as per the isotropic sample,
although in this case the point of onset of tertiary move- 200000

ment is not easy to determine (Fig. 8). Note that in this 180000

experiment, as with the others described here, fail-

1 / axial strain rate (secs / %)

160000

ure occurs at a surprisingly low value of axial strain 140000 Secondary phase
(<3%). 120000

The Saito plot for this 1000 kPa test is interesting 100000 Tertiary phase

(Fig. 9). Here, the primary movement phase is clearly 80000

evident, and there appears to be a sharp termination 60000

40000
into what is apparently the secondary phase. During
20000
this secondary movement period the Saito trend is non-
0
linear, resembling the trend seen for the remoulded 30000
sample. Time (secs)
However, a detailed examination of the later stages
of the experiment shows that the sample did appear to Figure 11. Saito plot for the 200 kPa experiment showing
display a clear linear Saito trend (Figure 10). The onset the secondary and tertiary movement phases. Note the non-
of this linearity occurred at about 14,000 seconds, linear form of the secondary movement phase and the lack of
which is consistent with that indicated by Fig. 8. a discernable primary movement phase.

869
 The other experiments showed only a secondary and Landslide block
tertiary movement phase, with no primary movement
(Fig. 11). This is probably because this phase of move-
ment could not be resolved by the equipment in use
here, given the small levels of strain involved. In each
case a very clear non-linear secondary creep phase was
evident in the Saito plot, followed by a linear tertiary
phase.

5 DISCUSSION Figure 12. Modified Bjerrum (1967) model to account for

Saito linearity.
The experiments described here clearly demonstrate
that Saito linearity in v−1 − t space can be replicated
experimentally in the stress path cell. These exper- of stress concentration on the unsheared material. As
iments have been undertaken on a material that is the area of this zone tends towards zero, the shear stress
known to fail in a brittle manner through the gen- acting upon it tends to infinity, explaining the observed
eration of a conjugate pair of shear surfaces. In the acceleration behavior.
experiments in which the samples had intact bond-
ing the Saito plots clearly demonstrated a linear trend
in v−1 − t space. In the experiment in which the 6 KEY CONCLUSIONS
bonding had been destroyed through remoulding no
linear trend was evident. In this case the sample dis- 1. The Saito technique is valid when, and only when.
played the same non-linear (exponential) trend that the landslide is brittle;
Petley et al. (2002 and 2005b) described. This appears 2. The Bjerrum (1967) model of progressive failure
to be consistent with the suggestion made by Petley is consistent with the Saito technique, and poten-
et al. (2002) and Kilburn and Petley (2003) that it tially provides an explanation for it through stress
is brittle crack growth processes that are responsi- concentration.
ble for the linearity in v−1 − t space first observed
by Saito (1965). These results do not appear to be
consistent with the model proposed by Helmstetter
et al. (2003) and Helmstetter and Sornette (2004) that REFERENCES
state- and rate-dependent friction is responsible for the
observed behavior. In particular, the state- and rate- Bjerrum, L. 1967. Progressive failure in slopes of overcon-
solidated plastic clay and clay shales. Journal of the Soil
dependent friction model does not seem to be able to Mechanics and Foundations Division, Proceedings of the
explain why linearity would be present in undisturbed American Society of Civil Engineers 93 (SM5): 2–49.
samples but not present in samples that are identi- Brabb, E. 1991. The world landslide problem. Episodes 14:
cal except for the destruction of their interparticle 52–61.
bonds. Brunsden, D. 1969. Moving cliffs in Black Ven. Geographi-
Thus, the Saito technique for the prediction of fail- cal Magazine 41(5): 372–374.
ure is valid where failure of the slope occurs through Helmstetter, A., D. Sornette, J.-R. Grasso, J.V., Andersen, S.
brittle processes (i.e. shear surface development). Gluzman, S. & Pisarenko, S. 2003. Slider-block friction
Where movement occurs through a different mecha- model for landslides: implication for prediction of moun-
tain collapse. Journal of Geophysical Research, 108(B2):
nism the Saito approach cannot be applied. On the 409–432.
other hand, if linearity is seen then it is clear that shear Helmstetter, A. & Sornette, D. 2004. Slider block fric-
surface formation is occurring. tion model for landslides; application to Vaiont and La
Bjerrum (1967) suggested that the process of pro- Clapiere landslides. Journal of Geophysical Research,
gressive failure in a slope can be associated with the 109(B2): 210–225.
development of the shear surface of the landslide. Hungr, O. & Kent, A. 1995. Coal mine waste dump failures
This is a brittle cracking mechanism. The experimen- in British Columbia, Canada. Landslide News, 9: 26–28.
tal data presented here support the Bjerrum (1967) Hungr, O., Corominas, J., and Eberhardt, E. 2005. Estimating
model. Interestingly, with slight modification the landslide motion mechanism, travel distance and velocity.
In: Hungr, O., Fell, R., Couture, R., and Eberhardt, E.
Bjerrum model might also explain why Saito linearity Landslide Risk Management. Taylor and Francis, London,
is observed (Fig. 12). Here, Bjerrum’s growing crack pp. 99–128.
is modified by the influence of a tension crack (alter- Kilburn, C.J. & Petley, D.N. 2003. Forecasting giant, catas-
natively, the shear surface moving towards the surface trophic slope collapse: lessons from Vajont, Northern
will also achieve the same effect). This creates a zone Italy. Geomorphology, 54(1–2): 21–32.

870
Menzies, B.K. 1988. A Computer Controlled Hydraulic Tri- Rose, N.D. & Hungr, O. 2007. Forecasting potential rock
axial Testing System. In: Donaghue, R.T., Chaney, R.C. slope failure in open pit mines using the inverse-velocity
and Silver, M. (eds), Advanced Triaxial Testing of Soil method. International Journal of Rock Mechanics and
and Rock, ASTM STP 977, American Society for Testing Mining Sciences 44: 308–320.
and Materials, Philadelphia, USA, pp. 82–94. Saito, M. 1965. Forecasting the time of occurrence of a
Petley, D.N., Bulmer, M.H.K. & Murphy, W. 2002. Patterns slope failure. Proceedings of the 6th International Con-
of movement in rotational and translational landslides. ference on Soil Mechanics and Foundation Engineering
Geology, 30(8), 719–722. 2: 537–541.
Petley, D.N., Dunning, S.A. & Rosser, N.J. 2005a. The anal- Suwa, H. 1991. Visually observed failure of rock slope in
ysis of global landslide risk through the creation of a Japan. Landslide News 5: 8–10.
database of worldwide landslide fatalities. In: Hungr, O. Zvelebil, J. 1984. Time prediction of a rockfall from a sand-
Fell, R., Couture, R., and Eberhardt, E. Landslide Risk stone rock slope. In: Proceedings of the IVth International
Management, A.T. Balkema, Amsterdam, 367–374. Symposium on Landslides 3: 93–95.
Petley, D.N., Higuchi, T., Petley, D.J., Bulmer, M.H. & Carey,
J. 2005b. The development of progressive landslide failure
in cohesive materials. Geology 33(3): 201–204.

871
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

3D landslide run out modelling using the Particle Flow Code PFC3D

R. Poisel & A. Preh

Institute for Engineering Geology, Vienna University of Technology, Vienna, Austria

ABSTRACT: Rockfalls are modelled as the movements of single rock blocks over a surface or as the movement
of a viscous mass over a surface (e.g. DAN). In reality a mass of discrete, interacting rock blocks is moving
downslope. Thus the program PFC (Particle Flow Code) based on the Distinct Element Method was modified
in order to model rock mass falls realistically in 3 dimensions based on physical relations. PFC models the
movement and interaction of circular (2D) or spherical (3D) particles and wall elements using the laws of
motion and of force-displacement. In the course of the calculation the contacts between particles and particles
or particles and walls are detected automatically. The particles may be bonded together at their contact points,
and the bondage can break due to an impact. For realistic modelling of the run out a viscous damping routine
in case of a particle—wall contact was introduced. Numerical drop tests and back analyses of several rock mass
falls provided appropriate damping factors. Thus, the movement types bouncing, sliding, rolling and free falling
of single rock blocks and the interaction between the blocks occurring in a rock mass fall can be modelled
realistically by using the adapted code of PFC. The application of this method is demonstrated by the example
of Aaknes (Western Norway).

1 INTRODUCTION 2004). After detachment of the moving mass, the run

out is modelled automatically.
Landslide run outs are modelled as the movements of In the ‘‘Ball Wall model’’ the underlying bedrock
single rock blocks over a surface or as the movement is simulated by linear (2D) and planar (3D) wall ele-
of a viscous mass over a surface (e.g. DAN). In reality ments (Roth 2003). Therefore, an estimate or a model
a mass of discrete, interacting rock blocks is mov- of the failure mechanism of the slope (Preh 2004)
ing downslope. Thus the program PFC (Particle Flow and of the detachment mechanism is needed as an
Code) based on the Distinct Element Method was mod- input parameter. However, in the ‘‘Ball Wall model’’
ified in order to model landslide run outs realistically the detached mass can be modelled, using more and
in 3 dimensions based on physical relations. smaller balls with the same computational effort in
PFC models the movement and interaction of circu- order to approach reality better.
lar (2D) or spherical (3D) particles and wall elements
using the laws of motion and of force-displacement. In
the course of the calculation the contacts between par- 2 RUNOUT RELEVANT PARAMETERS
ticles and particles or particles and walls are detected
automatically. The particles may be bonded together According to observations in nature, several kinds of
at their contact points, thus modelling a solid, and the movements of the rock fall process (Broilli 1974) have
bondage can break due to an impact. Thus, PFC can to be distinguished during the computation (Bozzolo
simulate not only failure mechanisms of rock slopes, 1987):
but also the run out of a detached and fractured rock
• free falling,
mass (Poisel & Roth 2004).
• bouncing,
Rock mass falls can be modelled as an ‘‘All Ball
• rolling and
model’’ and as a ‘‘Ball Wall model’’. An ‘‘All Ball
• sliding.
model’’ simulates the slope as an assembly of balls
bonded together. The simulation shows the failure In order to achieve an appropriate simulation
mechanism of the slope due to gravity (Poisel & Preh of these different kinds of movements by PFC,

873
some modifications have been necessary using the
implemented programming language Fish.

2.1 Free falling

In order to model the free falling of blocks, neither the
acceleration nor the velocity (ignoring the air resis-
tance) is to be reduced during fall as a consequence of
mechanical damping.
PFC applies a local, non-viscous damping pro-
portional to acceleration to the movement of every
single particle as a default. The local damping used
in PFC is similar to that described by Cundall (1987). Figure 1. Viscous damping activated at a contact with the
A damping-force term is added to the equations of linear contact model (Itasca 1999).
motion, so that the damped equations of motion can
be written
to the contact force, of which the normal and shear
F(i) + Fd(i) = M(i) A(i) ; i = 1...6 (1) components are given by

 Di = Ci · |Vi | (5)
mẍ(i) for i = 1 . . . 3;
M(i) A(i) = (2)
Iω̇(i−3) for i = 4 . . . 6 where Ci (i = n: normal, s: shear) is the damping
constant, Vi (i = n: normal, s: shear) is the rela-
where F(i) , M(i) , and A(i) are the generalized force, tive velocity at contact, and the damping force acts
mass, and acceleration components, I is the principal to oppose motion. The damping constant is not spec-
moment of inertia, ω̇ is the angular acceleration and ẍ ified directly; instead, the critical damping ratio βi
is the translational acceleration; F(i) includes the con- (i = n: normal, s: shear) is specified, and the damping
tribution from the gravity force; and Fd(i) is the damping constant satisfies
force
Ci = βi · Ccrit
i (6)
Fd(i) = −α|F(i) |sign(ν(i) ) i = 1...6 (3)
where Ccrit
i is the critical damping constant, which is
expressed in terms of the generalized velocity given by
 
ẋ(i) for i = 1 . . . 3;
ν(i) = (4) Ccrit = 2 mωi = 2 mki (7)
ω(i−3) for i = 4 . . . 6. i

where ωi (i = n: normal, s: shear) is the natural fre-

The damping force is controlled by the damping con- quency of the undamped system, ki (i = n: normal,
stant α, whose default value is 0.7 and which can be s: shear) is the contact tangent stiffness, and m is the
separately specified for each particle. effective system mass.
This damping model is the best suited for a quick In rock fall programs, the rebound height of blocks
calculation of equilibrium. There arises, however, the touching the bedrock is calculated using restitution
disadvantage of the movements of the particles being coefficients. The restitution coefficient Ri (i = n: nor-
damped as well. Therefore, the local damping has mal, s: shear) is defined as the ratio of the contact
been deactivated for all kinds of particle movements. velocity before and after the impact and can be
defined as
2.2 Bouncing
vif
Elastic and plastic deformations occur in the contact Ri = (8)
zone during the impact of a block. Both the kinetic vii
energy of the bouncing block and the rebound height
are reduced by the deformation work. The reduction of where vfi (i = n: normal, s: shear) is the velocity of the
the velocity caused by the impact is modelled with the block after impact and vii (i = n: normal, s: shear) is
help of a viscous damping model integrated in PFC. the velocity of the object before impact. The relation
The viscous damping model used in PFC introduces between the restitution coefficient Ri and the critical
normal and shear dashpots at each contact (Fig. 1). damping ratio βi can be estimated by simulating drop
A damping force, Di (i = n: normal, s: shear), is added tests (Preh & Poisel 2007).

874
 v

Figure 3. Deformation of the surface and distribution of

contact stresses.

Figure 2. Rebound angle influenced by the particle spin

(black line—rebound course; green, yellow and red balls—
particle position after rebound).

Spin has an impact on both the direction and the

velocity of the rebounding block. Therefore, it is
essential to consider the spinning when modelling the
run outs of rock falls. PFC determines the motion of
each single particle by the resultant force and moment
vectors acting upon it, and describes it in terms of the
translational motion of a point in the particle and the
rotational motion of the particle (Equations 1 and 2).
Figure 2 depicts the flight trajectories of three particles
bouncing at different spins.
Furthermore, with PFC the interaction of friction
and spin is considered, since the influence of the spin
increases with the increase of frictional resistance.
By modelling rock mass falls, it was shown to Figure 4. Calculation of the rolling resistance.
be necessary to distinguish between ball-ball contacts
and ball-wall contacts. This was done by using the
programming language Fish. of translational momentum (Equation 9) and angular
momentum (Equation 10).
2.3 Rolling
m · ẍs = −Frr (9)
The most important run out relevant effect is rolling
resistance, because it is known that pure rolling of 2
−I · ω̇rr = Mrr , Isphere = · m · r2 (10)
blocks in the model leads to more extensive run outs 5
than observed in nature.
The rolling resistance is caused by the deforma- where Mrr is the resulting moment caused by the
tion of the rolling body and/or the deformation of the rolling resistance, I is the principal moment of inertia
ground (Fig. 3) and depends strongly on the ground and ωrr is the angular deceleration.
and the block material. The kinematic link is established by the condition
Due to these deformations, the distribution of con- of pure rolling (Equation 11).
tact stresses between the ground and the block is
asymmetric (Fig. 4). Replacing the contact stresses ẍs = ω̇ · r (11)
by equivalent static contact forces results in a normal
force N, which is shifted forward by the distance of The angular acceleration is defined by a finite dif-
crr , and a friction force Frr , opposing the direction of ference relation in order to express the increment of
the movement. the angular velocity per time increment (Equation 12).
The deceleration of the angular velocity caused by Thus, the friction force Frri is defined by the conser-
the rolling resistance is calculated using conservation vation of momentum

875
 ωrr 2.4 Sliding
Frr = −m · ·r (12)
t Sliding is calculated by the slip model implemented in
PFC without any further adaptation.
Equation 10 and equation 12 yield

2 ωrr 3 THE AAKNES ROCK SLOPE FAILURE

− · m · r2 · = Frr · rrr − N · crr
5 t
2 ωrr ωrr Large rock slope failures are common events in the
− · m · r2 · = −m · · r · rrr inner fjord areas of Western Norway and represent
5 t t one of the most serious natural hazards in Norway.
− m · g · crr . (13) Rock avalanches and related tsunamis (Harbitz et al.
1993) have caused serious disasters and during the last
Therefore, the angular deceleration is 100 years more than 170 people have lost their lives
in Western Norway. The Tafjord disaster occurred in
−g · crr 1934 when 3 million m3 rock mass dropped into the
ωrr =
· t fjord. The tsunami generated by the avalanche reached
r · rrr − 25 · r a maximum of 62 m above sea level, several inhabited
 villages along the fjord were destroyed and 41 people
rrr = r 2 − c2rr . (14) were killed (Blikra et al. 2005).
The unstable rock slope at Aaknes (Fig. 5) is sit-
The rolling resistance is implemented by adding the uated in a steep mountain slope in Storfjorden and is
calculated increment of the angular velocity to the built up by gneisses with an overall dipping foliation
angular velocity calculated automatically by PFC at parallel to the slope (Tveten et al. 1998). The slide
every time step (Equation 14). planes are probably following weak zones along the
foliation planes. Geophysical data from 2D resistiv-
ity, refraction and reflection seismics and penetrating
ωi(t) = ωi(t) + ωrr,i (15) radar indicate that the slide is covering an area of some
800.000 m2 maximum and that the thickness of the
According to these considerations, the rolling resis- unstable area is between 40 and 140 m (Blikra et al.
tance is an eccentricity crr or sag function urr . The 2007).
deeper the block sags, the greater is the rolling The Aaknes/Tafjord project was initiated in 2004
resistance ωrr . In classical mechanics, the rolling with the aim of investigating, monitoring and provid-
resistance is a function of the ratio of the eccentricity ing early warning of the unstable areas at Aaknes and
crr to the radius r. Hegguraksla in Tafjord in More og Romsdal County.
The monitoring system includes extensometers, lasers,
crr GPS, a total station, ground based radar and borehole
μr = [−] (16) instrumentation (Blikra et al. 2007). The displacement
r
velocities are in the order of 3–10 cm/year at present.
This means that spherical blocks of different sizes
have the same run out for the same rolling resistance
coefficient.
In nature, however, it can be observed that large
blocks generally have a longer run out than smaller
ones. Therefore, according to the damping model
described, the run out is calibrated by the sag urr .
Calculations carried out by PFC, using the model
of rolling resistance described above and modelling a
detached rock mass as an irregular assembly of parti-
cles of two different sizes (r1 = 0.8 m, r2 = 1.6 m)
employing the same sag of urr = 25 cm for both par-
ticle sizes have shown that the larger particles have a
longer run out than smaller ones and that within the
deposit mass the smaller particles (r = 0.8 m) rest
at the bottom and the larger particles (r =1.6 m) at
the top (Preh & Poisel 2007). This model behaviour Figure 5. The Aaknes rock slope. Unstable area marked by
corresponds closely to observations in nature. dashed line. Photo: Th. Sausgruber.

876
The geometry and structure of the failure is complex
and the instable area seems to be composed of several
individual blocks.
Based on the relatively frequent slide events docu-
mented in the fjord areas, it has been estimated that
a flank collapse in this part of the fjord system in the
order of 1–8 million m3 may have a probability of less
than 1 event /1.000 years (Blikra et al. 2005).

3.1 Numerical investigations of the Aaknes run out

using PFC
The mapping of geological structures dividing the
moving area into separate blocks (Blikra et al. 2005),
displacement vectors showing different displacement Figure 7. Detached rock volume modelled by 2583
directions of these blocks as well as the wavy mor- particles.
phology of sliding planes of already occurred slides
in the surrounding area led to the conclusion, that the
slope failure will take place rather as a fall of a mass
of blocks than as a slide of a coherent mass. Thus PFC
was chosen to simulate the Aaknes rock slope failure
numerically.
The digital terrain models (DTMs) provided by the
Geological Survey of Norway were used to generate
the wall elements simulating the detachment and the
terrain surface of the smallest scenario (Fig. 6) and
the detached rock volume (some 5 million m3 ) was
modelled by 2583 particles (balls) with rmin = 5 m
and rmax = 7.5 m (Fig. 7).
First it was assumed that the moving mass is frac-
tured completely due to the sliding displacements
before the run out starts. Thus the run out process of
completely unbonded particles was started by delet-
ing the wall elements above the detached rock volume
using the PFC routine described above. Figure 8. Unbonded material, distribution of particles at the
Figure 8 shows the position of the unbonded par- moment the first particle hits the water surface.
ticles at the moment the first particle hits the water

surface, figure 9 the accumulated mass of particles

hitting the water surface over time.
In a second approach it was assumed that the mov-
ing mass is composed of several blocks built up by
an assemblage of particles. Thus bonds with a shear
strength of 1 GN and with a tension strength of 1 GN
were introduced at the particle contact points. These
bonds partly broke during the following run out.
Figure 10 shows the position of the initially com-
pletely bonded particles at the moment the first particle
hits the water surface, figure 11 the accumulated mass
of particles hitting the water surface over time.
Comparison of Figures 8–11 reveals completely
different distributions in space as well as in time. The
impacts of the unbonded material are restricted more
or less to the channel on the orographic right side of
Figure 6. Detachment and terrain surface of the smallest the moving mass while the initially completely bonded
scenario modelled by wall elements. material runs down distributed over the whole width

877
Figure 9. Unbonded material, accumulated mass of balls
hitting the water surface over time. Figure 11. Bonded material, accumulated mass of balls
hitting the water surface over time.

falls in 3D. The adapted code was used to model the run
out of the Aaknes rock slide in Storfjorden (Western
Norway) by unbonded and by bonded particles. The
simulations showed different particle distributions in
space and time which are expected to cause different
tsunamis. Thus investigations of the sliding rock mass
disintegration as a function of sliding displacements
are essential.

REFERENCES

Blikra, L.H., Longva, O., Harbitz, C.B. & Lovholt, F. 2005.

Quantification of rock avalanche and tsunami hazard in
Storfjorden, western Norway. In K. Senneset, K. Flaate,
Figure 10. Bonded material, distribution of particles at the J.O. Larsen (eds.), Landslides and avalanches. Proc. of
moment the first particle hits the water surface. the 11th Int. Conf. and Field Trip on Landslides (ICFL),
Norway, September 1–10, 2005: 57–63. London: Taylor
and Francis.
Blikra, L.H., Jogerud, K., Hole, J. & Bergeng, T. 2007.
of the sliding mass. The run out of the unbonded mate- Åknes/Tafjord prosejktet. Status og framdrift for over-
rial lasts only some 100 seconds, the run out of the vaking og beredskap. Report 01–2007 Åknes Tafjord
initially bonded material some 400 seconds. prosjektet (in Norwegian).
These results were transformed into a smoothed Bozzolo, D. 1987. Ein mathematisches Modell zur Beschrei-
particle hydrodynamics (SPH) based depth integrated bung der Dynamik von Steinschlag. Dissertation Nr. 8490
numerical model with coupling between the solid and an der ETH Zürich.
the fluid phases by Pastor (2007) in order to simulate Broilli, L. 1974. Ein Felssturz im Großversuch. Rock
the tsunami caused by the rock mass run out. Mechanics, Suppl. 3: 69–78.
Cundall, P.A. 1987. Distinct Element Models of Rock and
Soil Structure. In E.T. Brown (ed.), Analytical and Com-
putational Methods in Engineering Rock Mechanics: Ch.
4 CONCLUSIONS 4, 129–163. London: Allen & Unwin.
Harbitz, C.B., Pedersen, G. & Gjevik, B. 1993. Numerical
With the help of the adapted PFC code it is possible simulations of large water waves due to landslides. J. of
to create a mechanically correct model of rock mass Hydraulic Engineering 119(12): 1325–1342.

878
Hoek, E. 1987. Rockfall—A program in basic for the analysis Preh, A. & Poisel, R. 2007. 3D modelling of rock mass
of rockfalls from slopes. Dept. Civil Eng., University of falls using the Particle Flow Code PFC3D . Proceedings
Toronto, Toronto. of the 11th Congress of the International Society for Rock
Itasca 1999. PFC2D (Particle Flow Code in 2 Dimensions) Mechanics, Lisbon, July 9–13, 2007. Specialized Session
User’s Guide. Itasca Consulting Group, Inc., Minneapolis. S01—Rockfall—Mechanism and Hazard Assessment.
Pastor, M. 2007. Numerical simulation of the tsunamis Roth, W. 2003. Dreidimensionale numerische Simulation
caused by the Aaknes rock slide due to run outs modelled von Felsmassenstürzen mittels der Methode der Distink-
by PFC. Unpublished report. ten Elemente (PFC). PhD Dissertation, Inst. for Eng.
Poisel, R. & Preh, A. 2004. Rock slope initial failure Geology, Vienna University of Technology.
mechanisms and their mechanical models. Felsbau 22: Spang, R.M. & Rautenstrauch, R.W. 1988. Empirical and
40–45. mathematical approaches to rockfall protection and their
Poisel, R. & Roth, W. 2004. Run Out Models of Rock Slope practical applications. Proceedings of the 5th Interna-
Failures. Felsbau 22: 46–50. tional Symposium on Landslides, Lausanne, 1988: Vol. II,
Preh, A. 2004. Modellierung des Verhaltens von Massen- 1237–1243.
bewegungen bei großen Verschiebungen mit Hilfe des Tveten, E., Lutro, O. & Thorsnes, T. 1998. Bedrock map
Particle Flow Codes. PhD Dissertation, Inst. for Eng. Alesund. 1:250,000. Geological Survey of Norway.
Geology, Vienna University of Technology.

879
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Double-row anti-sliding piles: Analysis based on a spatial framework

structure

Tonghui Qian & Huiming Tang

Faculty of Engineering, China University of Geosciences, Wuhan, Hubei, China

ABSTRACT: Since the single-row anti-sliding piles could not able to provide an efficient bracing structure
under larger sliding-force, the double-row anti-sliding piles, which have the advantages of less displacement
in the top of the piles and large anti-force, can be used to reduce the pile deformation. However, the existing
calculation models of double-row piles ignored the interaction features of piles, beams and soils, and the
calculation results based on the models are not accurate enough. In this paper, a spatial force model of the
double-row anti-sliding piles was presented. When taking double-row anti-sliding piles as a single-layer multi-
span frame affected by piles, beams and soils, the spatial synergetics interaction between top ring beam and
linking beam, as well as the affection of soil-arch on the frame were analyzed. A deformation equation was
established according to pile border condition, continuous deformation, static force balance and deformation
coordination relationship between top ring beam and pile. An analysis program was developed on the basis
of FEM and the theory of Winkler Elastic Foundation Beam. Finally, the model was applied in the practice
of Three-Gorge Reservoir Prevention and Treatment project, and the result shows it is suitable for similar
engineering.

1 INSTRUCTION In this paper, a spatial force model of the double-

row anti-sliding pile is presented. Taking double-row
Anti-sliding piles are important retaining structure anti-sliding pile as a single-layer multi-span frame
for landslide prevention. Since single-row anti-sliding affected by piles, beams and soils, the spatial syner-
piles is not capable of ensuring good anti-sliding getics interaction between top ring beam and linking
effect under larger sliding-force, the double-row anti- beam, as well as the affection of soil-arch on the
sliding piles based on spatial framework structure, frame are analyzed. A deformation equation is estab-
which have the advantages of less displacement in lished according to pile border condition, continuous
the top of the piles and large anti-force, are devel- deformation, static force balance and deformation
oped as a new retaining structure, and can be used coordination relationship between top ring beam and
to reduce the piles deformation. Due to the com- pile. An analysis program is developed based on
plicated stress mechanism and imperfect design and FEM and the theory of Winkler Elastic Foundation
calculation, researches about double-row anti-sliding Beam. Finally, the model is applied in the practice
piles in large landslide prevention are just limited. Xu of Three-Gorge Reservoir Prevention and Treatment
(1998, 2004) calculated the rigid frame anti-sliding project, and the result shows it is suitable for similar
piles by Finite Difference Method according to the engineering.
theory of elastic foundation beam. Xiong (2002) sup-
posed double-row piles as separate cantilever piles and
discussed the distribution of anti-sliding push-force
2 CALCULATION METHOD
on different piles. Zhou (2005) presented an action
model for calculation of soils between piles on the
2.1 Presumption
front piles. However, the previous calculation models
of double-row piles ignored the deformation coordina- For the double-row anti-sliding piles under large push-
tion action between linking beams and ring beams, the force, the push-force acts on the back-row piles,
interaction between piles and beams. Further more the showed in Figure 1. The height of pile above slid-
calculation results based on the models are not accurate ing surface is h and the whole height of pile is H .
enough. Simplify the foundation resistance on front and back

881
 Figure 2. Calculation sketch for the double-row anti-sliding
piles.


Figure 1. The section of double-row anti-sliding piles. uqi = uqi (Fp , Fli , Fqi , Mlq , Mqq , Fa , Fb )
(1)
θqi = θqi (Fp , Fli , Fqi , Mlq , Mqq , Fa , Fb )

uhi = uhi (Fp , Fli , Fqi , Mlh , Mqh )
rows as elastic supporting and the horizontal resis- (2)
tance coefficient is calculated by m method. The basic θhi = θhi (Fp , Fli , Fqi , Mlh , Mqh )
presumptions are listed below (SS & DI, 1983).
According to the basic presumptions, the anti-
1. Piles, beams and linking beams satisfy the super- sliding piles are linear elastomer and satisfy the super-
position principle of force and displacement. position principles. So the above equations (1) and (2)
2. Soils below the sliding plane are thought as Winkler can be modified as following:
Discrete Linear Spring. ⎧
3. Tensile strength above the soils is 0 (Jiang, 2004). ⎪
⎪ uqi = upi + ubi + uqi − δPFii
u
(Fli )
⎪
⎪
4. The horizontal resistance coefficient is expressed ⎪
⎨ +δPMii (Mlq − Mqq )
u
as k = k0 +mz, m is the increasing proportional fac- (3)
tor of horizontal resistance coefficient with depths ⎪
⎪ θ
θqi = θpi + θbi + θqi − δPFii (Fli )
⎪
⎪
and is of different values for upper and below soils. ⎪
⎩ θ
5. When the ratio of pile space and pile diameter or +δPMii (Mlq − Mqq )
pile width (b/B) is smaller than 8, soil arch effects 
uhi = upi + δPFii
u
(Flh ) − δPMii
u
(Mqh + Mlh )
on front row should be considered. (4)
θ θ
θhi = θpi + δPFii (Flh ) − δPMii (Mqh + Mlh )

Where, upi and θpi are the displacement and turning

2.2 The mechanics model angle of the pile top in the vertical direction under the
θ
The push-force and active soil pressure distribute condition of only push-force existing, δPFiiu
and δPFii
along the piles length and the soil arch effects of Fa are the displacement and turning angle of the pile top in
and Fb are shown in Figure 2. the vertical direction under the condition of only axial
θ
Under the interaction of push-force (Fp ), active soil force existing, and δPMii
u
and δPMii are the displacement
pressure (Ea ), axial force (Fli ) from linking beam to and turning angle of the pile top in the vertical direc-
piles, bending moment Mla from linking beams to tion under the condition of only bending force moment
front-row piles, bending moment Mlh from linking existing.
beams to back-row piles, axial force Fi and bending
moment Mqq from front-rows ring beam on piles, axial
2.3 Stress analysis of ring beam
force Fi and bending moment Mqh from back-rows ring
beam on piles, active soil pressure Fa and soil arch Under the interaction of spatial synergetics, the res-
force Fb , the horizontal displacements on piles top are traint of ring beam provides a certain horizontal force
expressed by uqi and uhi angles by θqi , θhi . and bending moment to the anti-sliding piles, which

882
 2. Because of the linking beam, the bending moment
on pile top is transmitted to linking beam and
changed to the axial force, which is only horizontal
force on pile top without turning angle. The turning
angle of pile top is 0.

θ θ
θpi + θbi + θqi − δPFii (Fli ) + δPMii (Mlq − Mqq ) = 0
(8)
Figure 3. Calculation sketch for ring beam.
θ θ
θpi + δPFii (Flh ) − δPMii (Mqh + Mlh ) = 0 (9)

increases the stability of anti-sliding piles and reduces 

n 
n
δqfij Fhj = δQFij Fqj (10)
the trends of deformation.
j=1 j=1
Basic presumptions:
3. The effect of turning angle of the linking beam is
1. The crunodes at two ends of ring beams are quite
ignored. Under the lateral force the both front and
rigid, and seen as the fixed bearing ends.
back row anti-sliding piles are connected by linking
2. Within the span of ring beams, the piles exert hori-
beam. So the pile top deformation of both front and
zontal force on ring beams, and the stress and
back row are the same.
displacement satisfy the superposition principle.
It must be noticed that the flexibility coefficient
3. The torque from piles to ring beam is ignored.
of back-row piles is different from that of front-
The calculation sketch of ring beam is shown in row, and the same thing maybe to the ring beams
Figure 3. of double rows.
The displacement can be calculated by The flexibility coefficient of ring beams is
  

n
b2i xi2 2aj
ui = δQFij Fj (5) δGFij = 3aj − 1 + xi (i ≤ j)
6EIL2 L
j=1
(11)
Where δQFij is the horizontal displacement of ring
beam at point i when only existing horizontal force where, E is elastic modulus and I is section moment
Fj = 1. of inertia, bj , xi , aj , L, are presented in Figure 4.
4. The additional stress produced by soil arch
between front-row and back-row piles is Fb. When
2.4 The spatial analysis method 0 < b < 8B (Zhou, 2005),
After establishing of the calculation model for the 
double-row and ring beam, the inner force and defor- 2 bB B
Fb = + arctan qb (12)
mation can be calculated by the deformation col- π b2 + B2 b
laborative relationship between boundary condition,
deformation continuity of supporting points and static qb is the anti-force that soil exerted on the back-
force balance condition (Xu, 1982, Bao, 1998). row piles.
1. Based on deformation collaborative relationship,
the displacement of double-row piles and ring
beams are the same at joint i.

upi + ubi + uqi − δPFii

u
(Fli ) + δPMii
u
(Mlq − Mqq )

n
= δQFij Fqj (6)
j=1

upi + δPFii
u
(Flh ) − δPMii
u
(Mqh + Mlh )

n
= δqfij Fhj (7) Figure 4. Calculation sketch for the flexibility coefficient
j=1 of the ring beam.

883
 {F} = [K]{u} (14)

Where {F} is load vector, {u} is displacement vector

and {K} is whole rigidity matrix.

4 CASE STUDY

A FEM program was develop for analyzing double-

row anti-sliding piles based on the principles and meth-
ods above, by which the inner force and deformation
could be calculated.
Huangtupo Landslide Prevention Engineering
Project is located in the programming area for Badong
Country in Three Gorges Reservoir Region. Accord-
ing to geological investigation, the potential sliding
plane is of the shape of trumpet with front-edge width
Figure 5. Finite element calculation model of double-row. of 210 m, middle width of 340 m, back-edge width of
400 m and average portrait length of 280 m. From back
edge to front edge, the fragile degree of soils reduces
The active soil pressure between front-row and and the soil permeability increases. Three types of
back-row piles is Fa .When 0 < b < 8B (Zhou, anti-sliding piles were adopted: single-row cantilever
2005), piles, pre-stress anchored piles and double-row anti-
sliding piles. In order to compare the anti-sliding
effect, the rigidity is equivalent to square double-row
β(b/B) = 2b/b0 − (b/b0 )2 (13)
and single-row piles. The bearing condition at pile bot-
tom is gemel. Using m calculation method, 280 KN /m
5. Through equation (3) to (13), the 7n unknown surplus push-forces are distributed in a form of rectan-
forces (Fli , Fa , Fb , Mlq , Mqq , Mlk ) and Mqh could gle on the back-row piles. The calculation parameters
be figured out. for piles are listed in Table 1.
The bending moment diagrams for three types of
anti-sliding piles are shown in Figure 5 to Figure 8.
Figures 6, 7, 8, show that the bending moment, stress
3 FEM CALCULATION MODEL and displacement of double-row anti-sliding piles are
FOR DOUBLE-ROW ANTI-SLIDING PILES smaller obviously than those of single-row cantilever
piles and the pre-stress anchored piles when their
Figure 5 show, opposed that the soils below the sliding rigidity is with the same value, and the distribution
plane are continuous spring (Winkler Assumption), of bending moment of double-row anti-sliding piles
push-force behind piles is distributed as the pattern are more reasonable. The distribution of inner forces
of uniform load, the upper part of the pile above between front-row and back-row in double-row anti-
sliding plane is a beam, and the part of the pile sliding piles are different, and the maximal value of
below sliding plane is an elastic foundation beam. The bending moment located on the 1/3 of pile body in
rigidity coefficient can be calculated by simplifying back-rows is larger distinctly than that in front-row.
the horizontal foundation resistance on each joint of This means that the back-row piles are subject to larger
double-row piles. The distribution patterns of push- outer forces for the anti-sliding piles due to the spatial
force and active soil pressure on back-row piles are effect. It is also shown that the maximal displacement
determined by the properties and thickness of sliding of back-rows is smaller than that of front-rows. Defor-
body while the active soil pressure on front-row piles mation gradually decreases from pile top to bottom and
(Fa ) and the additional soil pressure (Fb ) can be gained the inner force of linking beam is larger. That means
by equations (12) and (13). Based on deformation col- that the spatial interaction has influence on double-row
laborative relationship, there are distributed force and anti-sliding piles. In addition, the lateral deformation
moment on pile top. Considering the static balance of of double-row piles is smaller than that of single-row
retaining structure, the following equation is obtained cantilever piles but close to that of pre-stress anchored
according to the method of direct rigidity coefficient piles. Thus, double-row anti-sliding piles can reduce
expressed in FEM (Zhu, 1998) the deformation effectively.

884
Table 1. Calculation parameters for piles.

Basic Single-row Pre-stress Double-row

parameters cantilever pile anchored pile anti-sliding pile Remarks

Pile Size Length (m) 28.000 28.000 28.000

Embedded Depth (m) 5.000 5.000 5.000
Width (m) 2.800 2.500 2.000 The same equi-
Height (m) 3.200 3.000 2.500 valent rigidity
Spacing (m) 6.000 6.000 Spacing 6.000
Distance 6.000
Horizontal Tension Linking Beam 1.0 × 2.0
965 KN (width × height)
Sliding Plan C (KPa) Cd = 22 Sliding Body C(kPa) Cd = 220
Cs = 25 Cs = 180
φ(◦ ) φd = 19.1 φ(◦ ) φd = 30
φs = 15.8 φs = 25

ρm (kN/m3 ) ρd = 19.7, ρs = 20.2

Figure 6. Bending moment of single-row cantilever pile. Figure 8. Bending moment of double-row anti-sliding piles.

Figure 7. Bending moment of anchored—pile.

linking beam is larger. This means that the defor-

mation of double-row piles is restrained greatly due
5 CONCLUSIONS to the spatial action, which enhances the ability of
anti-deformation. In addition, the spatial effects are
1. The stress of double-row anti-sliding piles is dif- existing among the anti-sliding piles, ring beams
ferent from that of single-row cantilever piles and and linking beams.
pre-stress anchored piles. Bending moment of pile 3. Double-row anti-sliding pile has the advantages
top is not 0 for both of the front-row and back-row of symmetrical stress, small bending moment and
piles. The maximal bending moment of double-row less displacement so that it is suitable for similar
anti-sliding piles is smaller than that of single-row projects.
cantilever piles and pre-stress anchored piles. 4. Pile, beam and soil were thought as an integral part
2. The maximal side displacement of double-row anti- in this paper and the effect of soil arch was also
sliding piles is smaller than that of single row considered. The calculation results showed with the
cantilever piles but close to pre-stress anchored clarity that the calculation model for double-row
piles. The displacement of front-row and back- anti-sliding piles based on the spatial framework
row in double-row piles is close and inner force of structure are reasonable and practicable.

885
REFERENCES Jiang Chusheng. 2004. Computation of Inner Forces for
Chair-shaped Anti-sliding Piles. Roadbed Engineering,
Xu Fenghe & Wang Jinsheng. 1988. Design and Construc- 112 (01): 57–59.
tion of Anti-sliding Rigid Framework in Luoyixi Project. Xiong Zhiwen, Ma Hui & Zhu Haidong. 2002. Force Distri-
Proceedings of Landslides. Beijing: China Railway Pub- bution of Buried Double-row Anti-sliding Piles, Roadbed
lishing House. Engineering, 102(03): 5–10.
Min Shunnan & Xu Fenghe. 1990. The Research of Chair- Zhou Cuiying, Liu Zuoqiu & Shang Wei. 2005. A New Mode
shaped Piling Wall in Shirongxi. Proceedings of Land- for Calculation of Double-row Anti-sliding Piles, Rock
slides. Beijing: China Railway Publishing House. and Soil Mechanics, 26(03): 441–444.

886
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Centrifuge modeling of rainfall-induced failure process of soil slope

J.Y. Qian
State Key Laboratory of Hydroscience and Engineering, Tsinghua University, Beijing, China

A.X. Wang
Department of Civil and Environment Engineering, University of Science and Technology Beijing, China

G. Zhang & J.-M. Zhang

State Key Laboratory of Hydroscience and Engineering, Tsinghua University, China

ABSTRACT: Two centrifuge model tests were carried out to investigate the rainfall-induced deformation and
failure process of Beijing Olympic Forest Park. A new rainfall generator was developed to realize rainfalls
during the centrifugal model tests. The deformation process of slopes subjected to a heavy rain was measured
using an image-based measurement system. The tests results indicated that the water content of the soils has
a significant effect on the failure of slopes: the slope with high initial water content is more inclined to fail.
The displacements of the slope, both in horizontal and vertical directions, increased with increasing rainfall; the
area where significant displacements occurred also enlarged. The deformation of the slope was significantly
dependent on the wetting front of the slope during a heavy rain.

1 INTRODUCTION 2 DEVICES

Soil slopes are inclined to fail due to the rainfall appli- The model tests were conducted using the TH-50 g-t
cation in the rainy season. The failures of theses slopes geotechnical centrifuge machine. The model container
claim many human lives and cause serious economic for tests is 500 mm long, 200 mm wide and 500 mm
loss. Therefore, a study is needed to explore a better high. Two transparent lucite windows are installed on
understanding of the mechanism of failure of slopes both sides of the model container.
subjected to rainfall; this can help to diminish these A new rainfall generator was developed to real-
disasters so far as possible (Kimura, T., et al. 1991, ize rainfalls during the centrifuge model tests. The
Tamate, S & Takahashi A. 1998). device mainly consists of three parts: the water supply
A 50-meter-high soil hill was completed in Beijing module, the water dispersing module, and the cali-
Olympic Forest Park, China, in 2007. Millions cubic bration module. Water is stored in a water tank that
meters of soil was constructed within only 108 days. is fixed on the radial beam of the centrifuge machine
Significantly, the Olympic Games 2008 will be held before the test. It is transported to the dispersing mod-
is pluvial August .Therefore, to evaluate the stability ule through the plastic pipes once an instruction is
level of the high soil hill of the Beijing Olympic Forest accepted during the test. The dispersing module is
Park, behavior of deformation and failure process of used to simulate the rainfall. It is made up of a lucite
a soil slope under different rainfall conditions should container with two copper pipes and a layer of semi-
be concerned. permeable cloth beneath the copper pipes (Figure 1).
In this paper, two centrifugal model tests are The calibration module is used to determine the rain-
described to investigate the process and mechanism fall intensity during the test; it is made up of a few
of failure of soil slopes subjected to a heavy rainfall; measuring cups.
the influence of the initial water content and density of The rainfall is slantwise relative to the fills because
the soil are also preliminarily discussed (Liu, D., et al. the radial velocities are significantly different between
2006). the rainfall generator and the slope. Moreover, the

887
rainfall direction is significantly influenced by the history of an arbitrary point on the soils can be mea-
wind due to rapid circumrotation of the centrifuge sured with sub-pixel accuracy. Only a colorful region
machine. A series of technical measures are used to with a random distribution is needed for this system;
assure a homogeneous rain on the slope, including: such a region can be obtained by embedding white
1) the rainfall generator is made a bit wider than the particles in the lateral side of the soils. For the test
container; 2) the lucite windows are cut in a given way conditions in this paper, the measurement accuracy
with an angle of 45◦ to keep the rain away from the can reach 0.02 mm based on the model dimension.
blocking of the lucite windows; 3) two pieces of wind-
proof cloth are fixed outside the container to prevent 3 CENTRIFUGE MODEL TEST
the wind from the rainfall (Figure 2).
An image-record and displacement measurement 3.1 Soils
system was used to record the images of soils dur-
The soils used in the centrifuge tests were taken
ing centrifuge model tests. 48 frames of image can
directly from the soil hill of Beijing Olympic Forest
be captured per second. An image-correlation analy-
Park. The average grain size of the soil is 0.03 mm.
sis algorithm was used to determine the displacement
The plastic limit and liquid limit are 5% and 18%,
vectors of soil without disturbing the soil itself (Zhang,
respectively; thus the plastic index is 13.
G. et al. 2006, Mu, T. et al. 2006). The displacement
3.2 Models
The soils were compacted to a layer with 6 cm in thick-
ness in the container until the total thickness reached
36 cm. A slope with the inclination of 45◦ was formed
by cutting out the redundant soil. A six-cm-high hori-
zontal soil layer under the slope was set to diminish the
influence of the bottom container plate on the defor-
mation of the slope. Moreover, the silicone oil was
painted on both sides of the container to decrease the
friction between slope and container sides. The images
of slope were recorded using the image-based mea-
surement system with which the displacement field
and its change can be furthermore obtained.
The model slope was installed on the centrifuge
machine and the centrifugal acceleration gradually
increased to 50 g, a centrifugal acceleration that was
maintained during raining tests. The rainfall started
after the deformation of the slope due to increasing
centrifugal acceleration became stable. The intensity
Figure 1. The dispersing module of rainfall generator. of rain produced was 15 mm/min based on the model
dimension. Two different centrifuge model tests were
conducted by varying the dry density and initial water
content of the soil (Table 1).

4 RESULT

4.1 The behavior of the failed slopes

The drier slope (No. 1 test) failed when the rainfall was
64 mm, while the wetter one ( No. 2 test) failed when

Table 1. The soils and conditions of two tests.

Initial water
No. Dry density content Rainfall

1 1.55 g/cm3 10% 64 mm

2 1.4 g/cm3 14% 12 mm
Figure 2. Photograph of model container.

888
the rainfall was only 12 mm. This indicated that the
slopes with higher water content were more inclined to
fail during a heavy rain. The failure behavior of the two
slopes was also significantly different (Figures 3–4).
A shallow slip surface was washed out in the wetter
slope, while a relative deep landslide occurred in the
drier one. Several cracks also appeared on the top of
the drier slope, while the top soils of the wetter one
were washed out. The toes of the fills were deposited
by the silts washed out by the rain; this implied that Figure 5. Contour lines of residue deformation of the drier
the surface runoff appeared during the tests. slope (No. 1 test). unit, mm.
Figure 5 shows the residue deformation due to rain-
fall. It can be seen that the horizontal displacement and
settlement of the drier slope were significantly larger
than those of the wetter one. It should be noted that
all the displacements of the slope are based on model
dimension. The increasing flexibility of the soil due
to increase of water content may be one of the reasons
of different displacement. Moreover, the rainfall influ-
ence depth, within which the significant deformation
occurred, of the drier slope was larger than that of the
wetter one. This would explain the different failure
behaviors of the two slopes. Thus, it can be concluded
Figure 6. Contour lines of residue deformation of the wetter
that the water content of the soils has a significant
slope (No. 2 test). unit, mm.
effect on the failure of slopes.

4.2 Deformation and failure process

The drier slope (No.1 test) was selected to discuss the
deformation and failure process of a slope due to rain-
fall. The downward movement of the wetting front can
be determined using the recorded photos (Figure 7).
It can be seen that the shape of wetting front was

Figure 7. Movement of wetting front obtained from pho-

tographs in test No. 1. r, rainfall.

Figure 3. Photographs of the failed drier slope (No. 1 test).

approximately parallel to the slope. And the displace-
ments were fairly small in the area without rainwater
penetrated.
Figure 8 shows the images and displacement con-
tour lines of the slope at several typical times. It can be
seen that the cracks occurred when the rainfall reached
17.5 mm. The horizontal and vertical displacements
both increased with increasing rainfall. The area where
significant displacements occurred also enlarged with
increasing rainfall; this exhibited a significant relation
Figure 4. Photographs of the failed wetter slope (No. 2 test). to the development of wetting front. It can be noticed

889
 image

3
2 8 10
u (mm)

1 1 6 6
0 0 2
2
0 0

14 20
v (mm)

2 4 16
10 12
2 6
1 1 8
2
4

r =17.5mm r =23.5mm r =37.5mm r =64mm

Figure 8. Images and displacement fields under different rainfall of the drier slope (No.1. test). u, horizontal displacement;
v, vertical displacement; r, rainfall.

0 At the beginning of the rainfall, small settlements

2 occurred at all the three points. Their difference
4
became significant when the rainfall reached 17.5 mm.
A The significant deformation occurred in the borders of
u(mm)

6
B the slope firstly (point A); it developed to the internal
8
C with increasing rainfall (point B & C). This is because
10 that the rain water penetrated the slope from border
12 A B C
to the internal (Figure 7). When the rainfall reached
60 mm, the horizontal displacement near the slope bor-
14
der (Point A) became straight nearly. In other words,
0 20 40 60 80
the displacement of the slope increased rapidly at this
0
time; this indicated the failure of the slope occurred.
2 The vertical displacement also increased with increas-
4
ing rainfall. When the rainfall got to 60 mm, unlike
6
v (mm)

horizontal displacement, the vertical displacements of

8 all the three points exhibited an significant increments,
10 this may be because a significant settlement occurred
12 at the bottom of the slope.
14
16
0 20 40 60 80
r (mm) 5 CONCLUSIONS

Figure 9. Displacement histories of typical points of the A new rainfall generator was developed to realize rain-
drier slope (No.1 test). u, horizontal displacement; v, vertical falls during the centrifugal model tests. Using the
displacement; r, rainfall. equipment, two centrifugal model tests were carried
out to investigate the deformation and failure process
of slopes subjected to a heavy rain and the influence
that the shape of the zero contour lines of the horizontal of the initial water content of fills.
displacement was similar to that of the wet front. The tests results indicated that the water content
In the slopes, three typical points were selected to of the soils has a significant effect on the failure of
discuss the deformation process of the slope (Fig. 9). slopes. The slope with high initial water content is

890
more inclined to fail, with a relatively shallow slip REFERENCES
surface washed out.
The displacements of the slope, both in horizontal Kimura, T., et al. 1991. Failure of fills due to rain fall.
and vertical directions, increased with increasing rain- Centrifuge’91: 509–516.
fall; the area where significant displacements occurred Tamate, S & Takahashi A. 1998. Slope stability test. Cen-
also enlarged. The shape of the zero contour lines of trifuge’98: 1077–1082
Liu, D., et al. 2006. Model study on the irreducible water sat-
horizontal displacement is similar to that of the wet uration by centrifuge experiments. Journal of Petroleum
front; this exhibits a significant relation between the Science and Engineering, 53, 77–82.
deformation and wetting front of the slope. Zhang, G., Liang, D. & Zhang, J-M. 2006. Image anal-
ysis measurement of soil particle movement during a
soil—structure interface test. Computers and Geotech-
ACKNOWLEDGEMENTS nics, Vol. 33 (4–5): 248–259.
Mu, T., Zhang, G. & Zhang, J-M. 2006. Centrifuge mod-
The project is supported by National Basic Research eling of progressive failure of soil slope. International
Program of China (973 Program)(No. 2007CB714108) Conference on Physical Modeling, Vol. 1: 373–377.
and National Natural Science Foundation of China
(No. 50679033). The support is gratefully acknowl-
edged.

891
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

A GIS-based method for predicting the location, magnitude and occurrence

time of landslides using a three-dimensional deterministic model

C. Qiu, T. Esaki & Y. Mitani

Institute of Environmental Systems, Kyushu University, Fukuoka, Japan

M. Xie
Civil and Environmental Engineering School, The University of Science and Technology Beijing, China

ABSTRACT: To predict the location, magnitude and occurrence time of landslides that occur in wide moun-
tainous area, a new GIS-based three-dimensional method is proposed. A GIS raster-based 3D slope stability
analysis model is delivered to calculate the 3D safety factor in regional landslide assessment. In order to locate
the potential failures, a Monte Carlo technique is used by means of minimizing the 3D safety factor through an
iterative procedure, based on a simulation of ellipsoid for the 3D shape of slip surfaces. A new GIS-based model
is developed by coupling a dynamic rainfall-infiltration model with the GIS-based 3D model to quantify the
varying safety factors during rainfall infiltration. All proposed methods are applied by developing a GIS-based
comprehensive system. The effectiveness of the method and the practicality of the developed system are verified
by a practical application for landslide-prone area in Japan.

1 INTRODUCTION Chameau 1983, Hungr 1987, Huang et al. 2002),

however, applications of the 3D models are generally
Any landslide hazard predicting work that is com- limited for site-specific slopes, but seldom being used
monly used for hazard precaution should give answers for a wide area due to the difficulties in processing
to three key questions: (1) the magnitude, (2) the and managing a vast amount of complex informa-
location and (3) the time of failure occurrence. The tion of natural slope, and in identifying unknown slip
factors that determine a landslide can be summarized surface(s).
as falling into two categories: quasi-static variables On the other hand, concerning the effect of the
and dynamic variables (Dai & Lee 2001). The quasi- dynamic variables on slope stability, the influence of
static variables, such as geology, terrain, and geotech- rainfall on landslide has been a subject of research
nical properties, contribute to landslide susceptibility for many years (Al-Homoud et al. 1999, Gasmo et al.
and determine the location and the magnitude of the 2000, Kim et al. 2004). When a wide natural area
failure. The dynamic variables, such as rainfall and is an object of study, much research has traditionally
earthquake, tend to trigger landslides in an area of been focused on the relationship between the probabi-
given susceptibility and thus have a direct impact on lity of landslide occurrence and the rainfall threshold,
the time of occurrence of the landslide. It is therefore in which the rainfall threshold, defined as the rainfall
necessary to evaluate the effects of both the quasi-static required to cause one landslide in an area observed
variables and the dynamic variables on slope stability during past rainfall events, was used to predict the
(Wu & Sidle 1995, Crosta 1998). occurrence of slope failure. An intrinsic hypothesis
Once the slope geometry and sub soil conditions underlying such methods is that the landslides ana-
have been determined, the stability of a slope may be lyzed exhibit consistent behavior. Therefore, a model
assessed using the deterministic methodologies. The established for a specific site is generally unreliable
deterministic models can be one-dimensional (1D), for application to a different site due to variations in
two-dimensional (2D), or three-dimensional (3D). In geotechnical conditions. Furthermore, such methods
case of a wide mountainous area where the geome- utilize rainfall as the only factor to provide warning
try varies significantly and the material properties are and do not take other important variables influencing
highly inhomogeneous, a 3D analysis may become landslide, such as geology, geometry, and ground-
necessary. Although many 3D models have been water, into account. Because rainfall-induced land-
proposed in the literature (Hovland 1977, Chen & slides occur as the result of the infiltration of water

893
from intense rainfalls, it is reasonable to evaluate this
type of slope failure using physically based models to
simulate the transient hydrological and geotechnical
processes responsible for slope stability (Wu & Sidle
1995).
To efficiently analyze a regional landslide hazard
which relates a large number of spatial data, a useful
tool for processing spatial data is essential. Recently,
the Geographic Information System (GIS), with its
excellent spatial data processing capacity, has attracted
great attention in landslide hazard assessment. If a
GIS-based deterministic model can be established, the
computing power provided by GIS will significantly
revolutionize the 3D slope stability analyses in two
ways: (1) the 3D models can be generated precisely
and automatically, based on the GIS data for topogra-
phy and stratum; and (2) a large number of slip surfaces Figure 1. GIS-based 3D slope stability analysis model.
can be analyzed, making it possible to locate the critical
slip surfaces with a high degree of reliability. Integrat-
ing a conventional column-based 3D model with GIS and then converted to GIS raster layers through spa-
raster data, the authors have proposed a GIS-based tial interpolation, as shown in Figure 1b. Each raster
3D slope stability analysis model to calculate the 3D layer represents a certain type of information such as
safety factor of a predefined failure mass and to further topography, stratum, or groundwater. The superposed
locate the critical slip surface by means of minimizing construction of the strata can be presented by overlay
the 3D safety factor through a Monte Carlo simulation of the multiple raster layers. The overlay extends the
(Xie et al. 2003, Qiu et al. 2007). 2D raster layers to a third dimension, which makes the
This paper describes the development of the pro- application of 3D engineering models possible. Here,
posed model into a new GIS-based 3D model by the GIS raster model is manipulated to accommodate
incorporating an infiltration model and taking account the traditional column-based 3D model by discretiza-
of geomechanical changes of soil strength during rain- tion of the study mass, as shown in Figure 1c. Slope
fall in the calculation. Using this model, the location stability analysis-related information for each soil col-
and the shape of critical slip surfaces can be identi- umn, such as topography, strata, groundwater, faults,
fied through a random search procedure by varying can thus be obtained from the corresponding cell of
the geometrical parameters of the ellipsoid that is raster layers (Fig. 1d) and used to calculate a safety
used to simulate the shape of the slip surface. Further- factor.
more, the time of occurrence of failures can be forecast Integrating Hovland’s (1977) 3D model and the GIS
by mapping the changing distribution of safety factors raster-based database, the safety factor can be derived
during rainfall. A GIS system is developed to effi- from the horizontal force equilibrium in the direction
ciently implement all of the computational procedures of sliding (Xie et al. 2003):
as well as the data preparation and the result visual-
ization. The effectiveness of the proposed method was 
(cA + W cos θ tan φ) cos θAvr
verified by a practical application for a landslide-prone J I
area in Japan. SF3D =  (1)
W sin θAvr cos θAvr
J I

where SF3D = 3D safety factor of the slope; c =

2 A NEW COUPLING RAINFALL-SLOPE
cohesion (kN/m2 ); A = area of the slip surface (m2 );
STABILITY ANALYSIS MODEL
W = weight of one soil column (kN); φ = friction
angle (◦ ); θ = inclination of the slip surface (◦ );
2.1 The GIS-based 3D model and identification
θAvr = angle between the direction of movement and
of critical slip surface
the horizontal plane (◦ ); and J , I = numbers of rows
The mechanism underlying the proposed GIS-based and columns of the cell within the range of failure
3D model is illustrated by Figure 1. A potentially slid- mass.
ing mass, which will be the study object, is shown To identify critical slip surfaces from the study area,
in Figure 1a. Initially, all the discrete investigation a random search is performed by means of a mini-
data for the study site can be represented as a num- mization of the 3D safety factor. The shape of the slip
ber of GIS vector layers (e.g. a line layer of contour), surface is assumed to be the lower part of an ellipsoid

894
and will vary in depth, dip angle, and direction of incli- and the actual infiltration rate is equal to the rainfall
nation. The direction of inclination is calculated as the rate:
mode value from the dip direction of all raster cells
in the range of the sliding mass. The uncertain para- f (t) = P (7)
meters, such as dip angle and the a, b, and c axes, are
F = Pt (8)
specified by Monte Carlo simulation from the range
they may take. For each cell that is taken as the cen-
tral point of a randomly chosen trial ellipsoid from the It is widely known that rainfall causes a rise of
range of sliding mass, a minimum safety factor and a groundwater level as well as an increase in pore water
critical slip surface can be obtained after enough time pressure that results in slope failure. However, in many
of calculations. situations where shallow failures are concerned, it
has been noted that the failure was attributed to the
advance of a wetting front into the slope instead of
2.2 Coupling the slope stability analysis model a rise of the groundwater level (Cho & Lee 2002).
with a rainfall-infiltration model The wetting front causes a reduction in soil suction (or
negative pore pressure) and an increase in the weight
Infiltration is a function of soil properties, rainfall and
of soil per unit volume. These result in a process
local settings. There are numerous models formulated
in which soil resistant strength decreases while total
on the basis of soil characteristics that have been pro-
stress increases, until failure occurs on the potential
posed to evaluate infiltration. The widely used Green
failure surface where equilibrium cannot be sustained.
& Ampt model (1911) is adopted in this study. In this
Under these conditions, there are four possible situ-
model, the infiltration rate at any time t is calculated by
ations regarding the slip surface that can be anticipated,
and four models are thus proposed to calculate the
ψf (θs − θi ) corresponding safety factors (Fig. 2).
f (t) = Ks + Ks (2)
F Model 1: The slip surface forms in the unsaturated
zone between the wetting front that is advancing from
The expression of F(t) can be stated as follows: the ground surface and the groundwater table (Fig. 2a).
In this situation, the horizontal resistance force F1 and
1   the horizontal sliding force F2 acting on the slip sur-
t = tp + F − Fp + ψf (θs − θi ) ln (n) (3)
Ks face can be calculated using Equation 9 and Equation
ψf (θs − θi ) + Fp 10, respectively:
n= (4)
ψf (θs − θi ) + F F1 = {ci A + [γi z + (γsat − γi )Hw ]
tn and Fn can be calculated from the following equa- × cos θ tan φ  } cos θAvr (9)
tions: F2 = [γi z + (γsat − γi ) Hw ] sin θAvr cos θAvr (10)
Fp
tp = (5) Model 2: The slip surface forms in the saturated
P zone between the ground surface and the wetting front
ψf Ks (θs − θi ) that is advancing from the ground surface (Fig. 2b).
Fp = (6)
P − Ks In this situation, the horizontal resistance force F1 and
the horizontal sliding force F2 can be calculated using
where f (t) = infiltration rate (m/h) at time t Equation 11 and Equation 12, respectively:
(h); Ks = soil saturated hydraulic conductivity (m/h);  
ψf = matrix suction at the wetting front (m); F = F1 = cw A + (γsat z cos θ − uw ) tan φ 
cumulative amount of infiltrated water (m); θs = soil
× cos θAvr (11)
saturated volumetric water content; θi = initial soil
volumetric water content; tn = time when water F2 = γsat z sin θAvr cos θAvr (12)
begins to pond on the soil surface (h); Fn = amount
of water that infiltrates before water begins to pond at Model 3: The slip surface forms in the saturated
the surface (m); and P = rainfall rate (m/h). zone under the groundwater table and the wetting front
It should be noted that a necessary condition for that has reached the groundwater table (Fig. 2c). The
Equation 2 to Equation 6 is that the rainfall rate must be horizontal resistance force F1 and the horizontal slid-
greater than the soil hydraulic conductivity. If the rain- ing force F2 can be calculated using Equation 11 and
fall rate is not greater than the potential infiltration rate Equation 12, respectively.
(P ≤ Ks) or no surface ponding occurs (t ≤ tn ), then Model 4: The slip surface forms in the saturated
all rainfall will infiltrate into the soil without runoff, zone under the groundwater table and the unsaturated

895
 zone exists between the wetting front and the ground-
water table (Fig. 2d). In this situation, the horizontal
resistance force F1 and the horizontal sliding force F2
can be calculated using Equation 13 and Equation 14,
respectively:

F1 = {cw A + [(γi (zw − Hw ) + γsat (Hw + z − zw ))

× cos θ − uw ] tan φ  } cos θAvr (13)
F2 = [γi (zw − Hw ) + γsat (Hw + z − zw )]
× sin θAvr cos θAvr (14)

Assuming that the vertical sides of each soil column

are frictionless, the 3D safety factor can thus be cal-
culated by summing F1 and F2 of all soil columns of
failure mass:

F1
J I
SF3D =   (15)
F2
J I

where SF3D = 3D safety factor of the slope; F1 =

horizontal resistance force (kN); F2 = horizontal slid-
ing force (kN); γsat = saturated unit weight of soil
(kN/m3 ); γi = initial unit weight of soil (kN/m3 );
ci = initial effective cohesion of soil (kN/m2 ); cw =
saturated effective cohesion of soil (kN/m2 ); φ  =
effective friction of soil (◦ ); zw = depth of the wet-
ting front (m); Hw = depth of the groundwater table
(m); z = depth of the slip surface (m); A = area of the
slip surface of the soil column (m2 ); θ = inclination
of the slip surface (◦ ); θAvr = dip angle of the main
sliding direction (◦ ); and J, I = numbers of rows and
columns of the cell in the range of failure mass.

3 COMPUTATIONAL COMPLEMENTATION

Figure 3 illustrates the computational procedures for

estimation of rainfall-induced slope stability. For a
given rainfall intensity, the permeation depth at a
specified time can be predicted by the infiltration
model described above. At the same time, taking a
cell of the raster dataset to be the central point of an
ellipsoid, a trial slip surface can be formed using the
Monte Carlo simulation model. A safety factor is cal-
culated subsequently by the 3D slope stability model,
depending on the relationship between the permeation
depth and the location of the slip surface after many
trials, the critical slip surface and the associated safety
factor, which is the minimum of all the trial calcu-
lations, can be obtained. The safety factor is then
saved to a point dataset as the safety factor of the
raster cell. Also, if the safety factor is < 1, the
Figure 2. Four models for rainfall-induced slope stability information for the corresponding critical slip sur-
analysis. face, such as location, shape, and volume, will be

896
 Figure 4. 3D view of the distribution of 3D safety factors
with critical slip surfaces.

The bedrock of the study area is represented by

Mesozoic granites, which is hard in a fresh state but
often heavily weathered to form deep residual deposits.
The soil cover in the area, with an average depth of
about 2 m, is composed mainly of colluvium and resid-
Figure 3. Flow chart of rainfall-induced slope stability ual deposits formed by weathering of granites and
caculation. accumulation of debris as result of landslide activ-
ity. It is well known that landslides are common in
colluvium and residual soils, particularly during peri-
ods of intense rainfall. The well-developed internal
saved in the format of a polygon. In this way, the
drainage of the soils is conducive to water infiltration,
safety factor at a specified time can be calculated
subsequent reduction in pore-water tension, and con-
for all raster cells inside the slide mass range. As a
sequent sliding. According to the characteristics of the
result, the time of occurrence of a landslide can be
geological formation, the possible collapse mode was
predicted by means of the decrease of the safety fac-
considered to be a shallow slope failure with a sliding
tor along with the amount of rainfall, and the location
surface along the boundary surface of the bedrock or
and the magnitude of the potential slip surfaces can
inside the soil layer.
be known from the polygon dataset of critical slip
The surface digital elevation model (DEM) data
surfaces.
in grid form with 2 m mesh size was produced
All of the above calculations have been incorporated
from airborne laser scanning. The geomechanical and
within a GIS-based system, called the 3D slope stabil-
hydraulic parameters of the soil surface consisted of
ity analysis system (3DSSAS), which is programmed
information from a specific geotechnical field inves-
by Visual Basic language using Microsoft component
tigation, triaxial test results of soil samples collected
object model technology (COM). In 3DSSAS, ArcOb-
from three spots, and related literature, as ci = 14.5
jects, the framework that forms the foundation of the
ArcGISTM applications (GIS software developed by (kN/m2 ), cw = 3.2 (kN/m2 ), φ  = 40.4 (◦ ), γi = 15.3
ESRI) with more than 2000 COM-based components (kN/m3 ), γw = 16.5 (kN/m3 ), Ks = 5.2 (cm/h),
has been used. ψf = 40.8 (cm), θs − θi = 0.31, respectively. In addi-
tion, in a range of 370 × 520 m, detailed data for soil
depth, with a precision of one value per 10 m2 , were
4 A PRACTICAL APPLICATION obtained by a special surveying method using a probe
stick. This range is used as the analytic range of this
The Goto section of National Route 49 is located near study. The soil depth data is then interpolated using
the city of Iwaki in Japan, about 10 km westward, the Gausses method of the Kriging to create a raster
where slope disasters have happened frequently. To data with 2 m mesh size to match mesh size of the
forecast slope failure and to provide support for decid- DEM data. The bedrock surface was then abstracted
ing on suitable countermeasures, the system described as a GIS raster dataset by subtraction of the surface
here was used for landslide mapping. elevation and the soil depth.

897
 of a randomly produced ellipsoid is lower than the
boundary surface of the bedrock, the confined surface
of the bedrock will be prioritized for selection as one
part of the assumed slip surface. After 100 trial cal-
culations for each raster cell taken as the central point
of a trial ellipsoid in the study range, finally, the crit-
ical slip surfaces were identified, and the variation of
safety factors over time was mapped. A 3D view of
the distribution of the critical slip surfaces that have a
safety factor < 1 is shown in Figure 4. The critical slip
surfaces are seen frequently in parts where the slope
angle is around 35◦ –50◦ . Figure 5 illustrates six distri-
bution maps of the safety factors changing over time.
From these maps, a high correlation between rainfall
and the decrease of the safety factor can be recognized.

5 CONCLUSIONS

By combining an infiltration model with an improved

column-based 3D slope stability analysis model, a new
GIS-based 3D model has been developed for evaluat-
ing the variation of surficial slope stability during a
rainfall event. A GIS-based program has been devel-
oped to carry out all the processes, including data
handling, calculation, and expression of the results.
The introduction of the GIS technology enables both
a convenient management of enormous amounts of
slope-related data and an easy implementation of
complicated calculations.
The suitability of the proposed method for large
complex and data-limited natural terrains was demon-
strated by a practical case study, in which the distri-
bution of the safety factors changing over time and
the location of the critical slip surfaces were mapped.
Such a spatio-temporal hazard map of rainfall-induced
landslide has profound implications for the identifica-
tion of the resultant processes that control slope stabi-
lity and specification of remedial and stabilization
measures.

REFERENCES

Al-Homoud, A.S., Prior, G. & Awad, A. 1999. Modeling the

effect of rainfall on instabilities of slopes along highways.
Environmental Geology 37 (4): 317–325.
Chen, R.H. & Chameau, J.L. 1983. Three-dimensional limit
equilibrium analysis of slopes. Geotechnique, Institution
Figure 5. Distribution maps of 3D safety factors changing
of Civil Engineers (ICE) 33 (1): 31–40.
over time.
Cho, S.E. & Lee, S.R. 2002. Evaluation of surficial stability
for homogeneous slopes considering rainfall character-
istics. Journal of Geotechnical and Geoenvironmental
Engineering 128 (9): 756–763.
For the study area, a uniform rainfall event with
Crosta, G. 1998. Regionalization of rainfall thresholds: an aid
duration of 10 h and an intensity of 8 cm/h was to landslide hazard evaluation. Environmental Geology 35
assumed, and the method proposed above was applied. (2–3): 131–145.
The detection of the critical slip surface was achieved Dai, F. & Lee, C. 2001. Terrain-based mapping of landslide
through trial searching and calculation of the 3D safety susceptibility using a geographical information system: a
factor. In the trial searching process, if the lower part case study. Canadian Geotechnical Journal 38: 911–923.

898
Gasmo, J.M., Rahardjo, H. & Leong, E.C. 2000. Infiltration Kim, J., Jeong, S., Park, S. & Sharma, J. 2004. Influence
effects on stability of a residual soil slope. Computers and of rainfall-induced wetting on the stability of slopes in
Geotechnics 26: 145–165. weathered soils. Engineering Geology 75: 251–262.
Green, W.H. & Ampt, G.A. 1911. Studies of soil physics: Qiu, C., Xie, M. & Esaki, T. 2007. Application of GIS
I. The flow of air and water through soils. J Agric Sci technique in three-dimensional slope stability analysis, In
4:1–24. Z. Yao & M. Yuan (Eds), International Symposium
Hovland, H.J. 1977. Three-dimensional slope stability anal- on Computational Mechanics, China, Springer Press:
ysis method. Journal of the Geotechnical Engineering, 703–712.
Division Proceedings of the American Society of Civil Wu, W. & Sidle, R.C. 1995. A distributed slope stabil-
Engineers, ASCE 103 (GT9): 971–986. ity model for steep forested basins. Water Resources
Huang, C., Tsai, C. & Chen, Y. 2002. Generalized method Research 31 (8): 2097–2110.
for three-dimensional slope stability analysis. Journal of Xie, M., Esaki, T., Zhou, G. & Mitani, Y. 2003. GIS-based
Geotechnical and Geoenvironmental Engineering, ASCE 3D critical slope stability analysis and landslide hazard
128 (10): 836–848. assessment. Journal of Geotechnical and Geoenvironmen-
Hungr, O. 1987. An extension of Bishop’s simplified method tal Engineering (ASCE) 129 (12): 1109–1118.
of slope stability analysis to three dimensions. Geotech-
nique 37 (1): 113–117.

899
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Application of a rockfall hazard rating system in rock slope cuts

along a mountain road of South Western Saudi Arabia

B.H. Sadagah
King Abdulaziz University, Jeddah, Saudi Arabia

ABSTRACT: Mountain roads play a vital role in the development of southwestern Saudi Arabia, as they were
built in a difficult terrain. This part of the country is characterized by high-rising and steep slope mountains. This
forms a natural obstacle to man-made road alignment and the engineered slope-cuts. Abha-Al-Darb mountain
road of almost 50 km long lies at one of the harshest terrains in Saudi Arabia. A 12 km portion of such mountain
road, in this study, that lies along sharp cliff suffers from frequent rock falls, mainly in rainy seasons, in addition to
various types of slope failures. The rock masses are mainly schist of high grade metamorphism. The folia-
tion of the schistosity dipping towards the rock slope cuts, causing a frequent occurrence of mainly rockfalls.
The man-made rock slope cuts are 70◦ −90◦ , and reach up to 40 m height. Many of these slopes are dangerous
and potentially unsafe due to rockfalls. The source of rock blocks is from upper slope elevations. Absence of
ditches and meshes aggravate the conditions under the slopes. In this research, the rock masses were studied in
order to identify the decisive parameters of rockfalls, and applied the Rockfall Hazard Rating System on such
steep man-made and natural rock slopes. Parameters such as slope angle, restitution coefficient, rolling friction
coefficient, bounce coefficient, trajectories, effect of block size and geometry were studied. Colorado Rockfall
Simulation Program (CRSP) modeled the most important rolling factors on rockfalls. According to this study,
remedial measures were suggested according to the site conditions and the dominant parameters utilizing the
latest technology to arresting fallen blocks.

1 INTRODUCTION process. Characterizing geology is important, because

this factor and its relation to the roadway controls
Rockfall occurrences along road cuts create consid- whether material is available to be shed as a rockfall.
erable risk for human injury and property damage, Approaches to incorporating geology in hazard assess-
posing problems for transportation across the moun- ment have included 1) defining hazard by association
tain road. Consequences of rockfall include impact with rock type (Hadjin, 2002, Vandewater et al. 2005),
damage to pavement from falling rocks, rocks on 2) ascertaining whether geologic discontinuities are
roads posing hazards to commuters, pedestrians, road oriented favorably or unfavorably with respect to pro-
closure, and environmental impact due to collisions moting rockfall (Abbott et al., 1998), and 3) describing
with vehicles (Moore, 1986, Wyllie & Norrish, 1996). the rockfall process for rock and soil slopes (Lowell &
Consequently, as the demand for rockfall protection Morin, 2000).
increases (Flatland, 1993). The goal of this study is to investigate 1) the role
Rockfalls occur when rock or debris is shed from of geological setting in causing rockfalls, 2) apply-
a road cut or nearby steep slope by processes such ing the Rockfall Hazard Rating System on selected
as planar sliding, wedge failure, toppling, differ- dangerous part of a road cut along the mountain-
ential weathering, and raveling onto the catchment ous road to evaluate the safety conditions, 3) apply
and/or road (Norrish & Wyllie, 1996, Sadagah, 1989). the latest used technique Colorado Rockfall Simu-
Characterization of rockfall hazard along road cuts is lation Program (CRSP) to model the incidents of
necessary for identifying hazard level and prioritizing rockfall failures, and 4) provide the transportation
remediation measures. The characterization includes departments and future investigators with correla-
attributes such as vehicular traffic patterns, roadway tions and recommended remedial measures that could
geometry, and rock slope geometry (Wyllie & Norrish, be used to consider likely rockfall modes and block
1996). However, a considerable attention was pointed sizes for a prospective road cuts in similar geologic
to characterize the role of geology in the rockfall characteristics.

901
2 STUDY AREA area. In this region, the contour line elevation indicate
that the escarpment edge generally has a drop of about
The study area is located in the southwestern part of 900 meters along a horizontal distance of 700 meters,
Saudi Arabia. In order to perform the development which forms a slope angle of >45◦ at the escarpment
plans, the Ministry of Transportation has to design, edge to about 35◦ at the bottom of the valley, overall.
construct, and execute roads in the mountainous high- These slope angles decrease toward the mouth of the
land areas. The purpose of these urban highways are (i) valley. Locally much greater slopes are found in the
to connect the road networks located on the plateau on field areas of study.
the mountainous areas with the lowland areas across One of the existing descents is Al-Dilaa descent
the Red Sea escarpment by creating these descents (Figure 2); it is located along Wadi Dilaa on the south
(Figure 1), to render service to residential areas and fringe of Abha city, and has constructed in a diffi-
commuters, (ii) to improve effectively the developing cult terrain, this descent connects directly the cities of
economic and social life, (iii) to improve an absolute Abha and Jizan.
solution to transportation and trade problems between The descent is located along the edge of the high-
the western, south western and southern regions, and land forming the eastern part of what is called Arabian
(iv) to supplement the existing road network which Shield. This area is formed of difficult and inaccessible
by the end of 70’s decade became insufficient for the terrains mostly of hard rocks, generally to moderately
high volume of traffic generated by the modernization weathered, with slightly to moderately jointed rock
of the country. masses, and moderate to steep rock slopes that are
covered in variable amounts of soils and debris. The
characteristics are different from one place to another
2.1 Location and physiography according to the changes in rock type, lithology, and
origin. The rocks in this descent are mainly igneous
The topography in the western, south western, and
and metamorphic in origin. The slope angle of the
southern regions of the kingdom along the Red
rocks is as mentioned above.
Sea coast is hilly and rugged. These mountains are
bounded by sharp cliffs especially in the western parts.
These cliffs form the so-called Red Sea escarpment
and they reach an altitude of more than 3,000 meter 2.2 Climate
above sea level at the southern regions of the escarp- The climate in the study area (western highlands)
ment, this altitude decreasing northwards. Along the ranges from hot and dry to humid and rainy. The daily
Red Sea escarpment lie many descents, especially in and seasonal temperature variations are significant by
the western and south western parts of the mountainous about 15◦ .
Because the areas of this study lie on escarpment
which rise to heights of 3,000 meters and above, they
are very similar to the mediterranean region. In win-
ter the temperature falls to below zero at nights. The
average annual temperatures are about 13.7◦ .
The winds in the study area come from the
Ethiopian highlands across the Red Sea.

Figure 2. The entrance of Al-Dilaa descent at Abha city,

Figure 1. Location of the study area. where the escarpment edge is shown.

902
2.3 General setting A major phase of igneous activity culminated in the
formation of monzogranite, granodiorite, diorite and
The Al-Dilaa descent lies at north of Wadi Dilaa. It
gabbro, renamed collectively by Greenwood (1985) as
connects Abha city at the escarpment with Al-Darb
biotite monzogranite. The field observations show that
town at downhill. The project begin at Abha city where
the contacts are not sharp, and the monzogranite is par-
the upper part of the plateau (chainage km 0.0 to km
tially intruded by the granodiorite. Greenwood (1985)
14.24); this is where the study area is, and extend for
grouped the diorite and the gabbro as one unit. Field
50 km to Al-Darb town.
observations show that the granodiorite is younger
The descent starts at the edge of the Red Sea escarp-
than the monzogranite and older than the diorite.
ment on a plateau at an elevation of 2,250 meters above
A schematic cross section at the Al-Dilaa descent is
sea level and continues down to the beginning of the
given in Fig. 3.
Dilaa valley at an elevation of about 1,300 meters
The monzogranite at the descent crest is pink,
above sea level. The valley sides at the north end of the
coarse grained, massive to weakly cleaved rock, locally
descent are very steep because of the sharp decrease
prophyritic. This granite body forms a mixed-rock
of the elevation at the escarpment. The valley runs
complex with the diorite, granodiorite and gabbro.
almost NNW-SSE, meandering between the mountain
The granodiorite is light to dark grey coarse grained
slopes. The north side of the valley viewed from the
and partially gneissed. The diorite is fine grained to
valley floor rises sharply up to 500 meters above the
medium grained, massive to weakly cleaved, medium
valley, and then rises another 250 meters in more gen-
to dark greenish grey. The gabbro forms a small plu-
tle slopes. At the south end of the valley the mountain
tonic body, massive to moderately cleaved, equigran-
foothills start at an elevation of about 600 meters.
ular, medium to coarse grained.

3 BRIEF GEOLOGICAL DESCRIPTION 3.3 Geology and erosion effect on rock jointing
As the Arabian Shield was uplifted 25–30 Ma ago,
3.1 Introduction
and because of the intrusive nature of the rocks form-
The study area is located along the escarpment edge ing the mountains, the western side of the escarpment
of Al-Dilaa descent. It is located in the south west- is very steep toward the Red Sea and forms a natu-
ern part of the Arabian Shield. Al-Dilaa descent is ral obstacle for communications. This escarpment is
bounded by Longitudes 42◦ 29 21 and 42◦ 31 18 E not sharp all over the continental divide; the different
and Latitudes 18◦ 9 20 and 18◦ 11 42 N. the descent intensities of the erosional cycles and the mechanical
is easily accessible; Al-Dilaa descent is forming the properties of the rocks certainly affect their response
south fringe of Abha city. to erosion processes. Linking the crest with the low
The Arabian Shield is an extensive occurrence of lands is a difficult engineering problem.
Precambrian crystalline and metamorphic rocks in
the western part of the Arabian Peninsula. It covers
an area of 610,000 km2 and is separated from the
African Shield in the west by the Red Sea graben
about 25–30 Ma ago (Stoeser and Camp, 1985). Strati-
graphic units of the Shield are given by Schmidt
et al. (1973) and Greenwood et al. (1980). They
are grouped from older to younger in the follow-
ing lithostratigraphic assemblages: (1) biotite schist
and amphibolites; (2) metabasalt-greywacke-chert;
(3) metabasalt-andesite; and felsic volcanic rocks.

3.2 Brief geologic description

The rocks at Al-Dilaa descent is partially metamor-
phosed intrusive plutonic rocks. The tonalite forms a
large suite of rocks that intruded layered rocks of the
Bahah group (Greenwood et al., 1982; and Stoeser,
et al., 1984). The tonalite is light to medium green-
ish grey and generally medium to coarse grained, fine
grained locally. In the field, tonalite is weakly to
strongly foliated and has gneissic layering defined by Figure 3. Cross section of the rocks located at Al-Dilaa
biotite and hornblende concentrations. descent.

903
 The strength and the degree of jointing in the 5 ENGINEERING GEOLOGY OF THE FIELD
rocks influenced by the processes and cause the alti- AREAS
tude of the rock mass to decrease, even to retreat
back the escarpment divide line along the high ter- 5.1 Introduction
rains. Al-Dilaa descent is a good example of this
Al-Dilaa descent is generally of steep slopes descend-
process.
ing from an escarpment edge, and dissected by
The escarpment edge line of Al-Dilaa descent is
the drainage patterns. The descent mountainous is
sharp due to (i) the high strength of the rock materi-
strongly fissured in the upper areas of Wadi Al-Dilaa.
als, (ii) less fractured rocks along the escarpment line
The road alignment runs through and area which has
(diorite and granite).
a difference in elevation of 1,000 m. most of the prob-
lems associated with the construction of the roads
arose from (i) the steep escarpment slope, (ii) the
drainage patterns, (iii) the unstable natural slopes and
4 ROCK MASSES (iv) the need to force the route corridors through highly
schistosed rock masses.
4.1 Field data acquisition
The main target of the engineering geological study
5.2 Geotechnical parameters
was (1) to record the existing engineering geological,
geodynamic and rock conditions of the mapped areas, At al-Dilaa descent, the weathering grade of tonalities
(2) to demonstrate the suitability of the designed slopes is generally grade II (slightly weathered), but in some
and tendency of rock slope stability, and (3) to map the localities the rocks are moderately to highly weath-
past and probable future active hazards to the road and ered; in these sections these grades are associated with
to the other existing engineering structures. These data the existence of groundwater seepage and landslides.
were then to be processed as a case history for creating The higher value of the weathering grade is associ-
a safer road design in other areas of similar rock types. ated with the low value for Schmidt hammer rebound
The road was divided into sections and stations number (29). At the higher elevations of the escarp-
as the geological and geotechnical characteristics of ment, the weathering grade is classified as II (slightly
the rocks and engineering structures varied along the weathered).
motorways. The engineering solution specified for a rock mass
Classical geological mapping pf all the present containing weak and weathered rock is shotcrete. The
geological elements was carried out, in addition to thickness of the shotcrete is 5 cm as setted by Ministry
mapping of the geotechnical properties of the existing of Transportation, in recognition of the active modes
rocks. The engineering geological description of the of weathering in progress in the mountainous areas.
rocks was based upon the systems of ISRM (1978), The mode of failure is controlled by the orientation
Geological Society of London (1977), and Geotech- of the joints with respect to the loading direction, and
nical Control Office (1984). CSIR, NGI classification by the spacing of the joints in relation with respect to
systems by Barton et al. (1974) and Bieniawski (1976) the dimensions of the loaded area. Therefore, joints
were applied in assessing the rock conditions in every spacing was measured at the example given.
section (cut slopes, tunnels, and bridges) along the JRC value for each joint set of the studied rock
road alignment at descent. masses was calculated in addition to the friction angle
The collected data comprised (1) for rock slopes; b for every rock type, in order to calculate the peak
joint survey, joint spacing survey, and joint rough- shear strength of the rock types.
ness. A vertical profile was measured for each section The results shows that tonalite rocks shows the high-
along the road cut alignment. In addition, the dimen- est shear strength value, however these rocks have low
sions and the geometry of each rock slope cut was JRC values in the field. The granites and the gran-
registered and the spacing between the rock bolts and odiorites show shear strength values lower than the
their number was measured; (2) For every studied tonalite although the granites and granodiorites have
section, measurements were made on discontinuities higher JRC values in the field.
using the Schmidt hammer, shear tests, and point load Degree of jointing in the study area being related to
were covering all rock types along the road; (3) Land- (1) rock type, (2) geological history, and (3) mechani-
slide hazards, failures, and damage were observed and cal properties of the rock. The critical height of steep
recorded; (4) The water level in wells, seepage and wet slopes is dependent not only on the intact strength
areas were marked; and (5) non-destructive test such but also on the orientation and strength of issues and
as density and sonic pulse velocity were done for dif- joints.
ferent rock types. Example of a studied section will be At Dilaa descent, the granitic rock masses at the
given below. edge are widely jointed, the tonalite rocks are closely

904
jointed and schistosed. Although the rock masses at 6.2 Rockfall events
Al-Dilaa descent are classified as low and highly
Incidents of rockfalls reported frequently along the
jointed, the rocks vary locally in the intensity of joint-
higher parts from 2260 to 2220 and drop on elevation
ing. The general cohesion and hence the stability of
2200 m to 2020 m of the descent road, especially after
the rock masses close to the escarpment edge are
rainfall. The fallen rock blocks vary in number, size
likely to be higher than those of the tonalite. Accord-
and shape according to the technical characteristics
ingly, the slopes of the granitic rocks are higher in
of the rock material, rebound coefficient, slope sur-
elevation and in the slope angle the rock slopes of the
face roughness. The fracture information, including
tonalities.
information on size, roughness and other parameters
Toppling is another type of rock slope failure. It is
were measured in the field and exported for analyses
controlled by the ratio between the height to the base
to CRSP version 4.0 created by Jones et al. (2000).
of the rock blocks, weight of the blocks, the angle of
Results are given in Figures 5 and 6 assuming that the
friction and the angle of the joint plane where the block
fallen rock blocks are of 1 m diameter.
will topple. Toppled blocks and loose boulders even-
The results show that the velocity and bounce
tually accumulate in the form of scree. These screes
increase after about 37 meters from the maximum ele-
may be small but can contain big pieces of rock which
vation of the slope profile. Further, it reaches the road
have become detached from the rock mass and which
with a high value of velocity and bounce height. If
have fallen as an individual piece.
the diameters of the blocks are smaller, the blocks will
At Al-Dilaa descent, the toppled rock boulders
have a higher velocity and bounce height. The results
are not restricted to any particular rock type. Many
shows that as the start point of y-coordinate of rockfall
loose blocks have fallen from above the support of
are further high above the hit point (road), the kinetic
slope face; and generally accumulating in the gullies
energy bounce height, and velocity increase. This will
where they could be forming the basis of a debris flow
cause more damage to the road.
or a landslide. At Al-Dilaa descent, the free fall of the
blocks is triggered mainly by the rain fall and could
cause a road closure for clearance. The fallen boulders
make holes in the road pavement. 25

20

Maximum
6 MODELING OF THE ROCKFALLS Bounce 15
Height (m)
6.1 Using Rockfall Hazard Rating System 10

From above description of the geotechnical character- 5

istics at the study area, an example was taken (Fig. 4).
The RHRS shows a high values at this section which 0 0 37 75 113 151 189 227
indicate the necessity of quick solution to take the Horizontal Distance Along Slope (m)
appropriate remedial measures.
Figure 5. Horizontal distance along slope versus max.
bounce height along failure path.

30

25

Maximum 20
Velocity
(m/sec) 15

10

0 0 37 75 113 151 189 227

Horizontal Distance Along Slope (m)

Figure 4. Arial photo showing the steep slopes above the Figure 6. Horizontal distance along slope versus max.
descent highway. velocity along failure path.

905
7 RECOMMENDATIONS Saudi Arabian Directorate General of Mineral Resources
Geoscience Map GM-77A, p. 27, scale 1:250,000.
According to the results of applying HRHS system and Greenwood, W.R., Anderson, R.E., Fleck, R.J. & Roberts,
CRSP program based on the geotechnical characteris- R.J. 1980. Precambrian geologic history and plate tectonic
tics of the rock masses, it is highly recommended to of the Arabian Shield: Saudi Arabian Directorate General
of Mineral Resources Bull. 24, 25p.
install a wire mesh at different elevations on the slopes Greenwood, W.R., Stoeser, D.B., Fleck, R.J. & Stacey,
parallel to the road alignment. Elevations where the J.S. 1982. Proterozoic island-arc complexes and tectonic
wire meshes to be installed depend on the results of belts in the southern part of the Arabian Shield, King-
1) number of blocks stopped along the profile, which dom of Saudi Arabia: Saudi Arabian Directorate General
will be provided by CRSP, 2) size of the blocks, and of Mineral Resources Open-file Report USGS-OF-02-8,
3) slope profile and roughness. 46p.
Hadjin, D.J. 2002. New York state Department of transporta-
tion rock slope rating procedure and rockfall assessment:
Transportation Research Record, No. 1786, pp. 60–68.
8 CONCLUSIONS International Society for Rock Mechanics. 1978. Suggested
methods for determining the strength of rock materials in
triaxial compression. ISRM Commission on Standardiza-
The Arabian Shield was uplifted 25–30 Ma ago, since tion of Laboratory and Field Tests. Intnl. J. Rock Mech.
then it has been subjected to severe erosional processes Min. Sci. & Geomech. Abstr. Vol. 15, No. 2, pp. 47–51.
leading to the removal of about 1 km of rock, which is Jones, C.L., Higgins, J.D. & Andrew, R.D. 2000. Colorado
greatly decreases the quality of the rock masses. rockfall simulation program version 4.0 for windows
The nature, the mode of formation and the mechan- manual. 127p.
ical properties of the rocks at the topmost elevation Moore, H.L. 1986. Wedge rockfalls along Tennessee high-
along the escarpment are the major factors responsible ways in the Appalachian region: their occurrence and cor-
for shaping the morphology of the descent. rection. Bulletin Association of Engineering Geologists.
Some of the fallen blocks could fall behind the road Vol. 23, No. 4, pp. 441–460.
Norrish, N.I. & Wyllie, D.C. 1996. Rock slope stability analy-
and fall in the valley. sis. In Turner, A.K. and Schuster, R.L. (eds.), Landslides:
CRSP program shows a significant ability to per- Investigation and Mitigation: Transportation Research
form simulation of the rockfalls, in addition to back Board Special Report 247, National Research Council,
analyses of the unseen rockfall events. Washington, DC, pp. 391–425.
Sadagah, B.H. 1989. Engineering geological maps for road
design and construction in Saudi Arabia. Unpublished
PhD. Thesis, Imperial College of Science, Technology
REFERENCES and Medicine, University of London. 417p.
Schmidt, D.L., Hadely, D.G., Greenwood, W.R.,
Abbott, B., Bruce, I., Savigny, W., Keegan, T. & Oboni, F. Gonzalez, L., Coleman, R.G. & Brown, G.F. 1973.
1998. Application of a new methodology for the man- Stratigraphy and tectonism of the southern part of the
agement of rockfall risk along a railway, In Moore, D.P., Precambrian Shield of Saudi Arabia: Saudi Arabian
Hunger, O. (eds.), 8th Congress of the International Asso- Directorate General of Mineral Resources Bull. 8, 13p.
ciation of Engineering Geologists: Vancouver, BC, A.A. Stoeser, D.B., Stacey, J.S., Greenwood, W.R. & Fischer, L.B.
Balkema, Netherlands, pp. 1201–1208. 1984. U/Pb zircon geochronology of southern portion of
Barton, N., Lien, R. & Lunde, J. 1974. Engineering classi- the Nubian mobile belt and Pan-African continental colli-
fication of rock masses for the design of tunnel support. sion in the Saudi Arabian Shield. Saudi Arabian Direc-
Norwegian Geotech. Inst. Publ. 106, 48p. torate General of Mineral Resources Technical Record
Bieniawski, Z.T. 1976. Rock mass classification in rock engi- USGS-TR-04-5, 88p.
neering. Proc. Symp. Exploration for Rock Engineering, Stoeser, D.B. & Camp, V.E. 1985. Pan-African microplate
Johannesberg, Vol. 1, pp. 97–106. accretion of the Arabian Shield. Bulletin of the Geological
Flatland, R. 1993. Application of the Rockfall Hazard Rat- Society of America, Vol. 96, pp. 817–826.
ing System to the rock slopes adjacent to US50 and state Vandewater, C.J., Dunne, W.M., Mauldon, M., Drumm,
route 28 in the east side of Lake Tahoe, Nevada: Unpub- E.C. & Bateman, V. 2005. Classifying and assessing the
lished M.S. thesis, University of Nevada, Mackay School geologic contribution to rockfall hazard. Journal of Envi-
of Mines, Reno, 316p. ronmental & Engineering Geosciences, Vol. 11, No. 2,
Geological Society of London. 1977. The description of rock pp. 141–154.
masses for engineering purposes. Report by the Geo- Wyllie, D.C. & Norrish, N.I. 1996. Stabilization of rock
logical Society Engineering Group Working Party. Q. J. slope. In Turner, A.K. and Schuster, R.L. (ed), Landslides:
Engng. Geol., Vol. 10, No. 4, pp. 355–388. Investigation and Mitigation: Transportation Research
Geotechnical Control Office. 1984. Geotechnical manual Board Special Report 247, National Research Council,
for slopes. Engineering Development Department, Hong Washington, DC, pp. 474–504.
Kong, 295p.
Greenwood, W.R. 1985. Geologic map of Abha quadran-
gle, sheet 18F, Kingdom of Saudi Arabia: GM-75C.

906
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Model tests of collapse of unsaturated slopes in rainfall

N. Sakai
National Research Institute for Earth Science and Disaster Prevention (NIED), Tsukuba, Japan

S. Sakajo
Kiso-Jiban Consultants Co., Ltd., Tokyo, Japan

ABSTRACT: The authors conducted two model tests of sandy slope collapse by rainfall in 1G field. The
mechanism was investigated very precisely by installing many sensors. The saturation degrees, surface displace-
ment and ground deformation were monitored very well. Through these test results, it was approved that slope
collapse could be predicted prior to the failure by monitoring surface displacements and saturation degrees.
However, it is not easy in case that the initial water flow was made in the slope before rainfall start. Fur-
thermore, a numerical simulation was conducted coupling seepage analysis and non-linear deformation analysis
alternatively, based on the finite element method. The limitation and applicability of the analysis were discussed.

1 INTRODUCTION service through the internet of the past record of slope

collapses.
Landslide is one of severe natural disasters all over the The authors have studied this slope collapse from
world. Actually, this can be divided into the 3 research the points of water and deformation to establish a
fields in general. 1) The first one is landslide, where is
currently sliding in the mountain, with its sliding zone
is specified by geologists based on soil investigations.
2) The second one is land collapse by the rainfall or
earthquake. 3) The third one is mud water flow mixing
soils and gravels. The three fields have been studied
individually by the different research groups. The first
one has been studied mainly by the geologists. The sec-
ond one has been studied by a few soil engineers and
researchers. The third one has been studied mainly by
the national institutes, universities and governmental
researchers.
The authors have studied the second one mechan-
ically and experimentally for these years, because of
believing this study could contribute to the third field
in the future. Slope collapse is very important because
it occurs quite often every year and the repairs are
very costly, in not the local but also urban areas all
over Japan. As the prediction service of slope col-
lapse, several information systems are available by
different organizations, which is based on rainfall
volume or water infiltration to the ground, although
actually it is difficult to predict as shown in the photo-
graph 1. As seen in this photograph, only particular
places are damaged in the mountainous area with
same rainfall. Therefore, for example, NIED (National
Research Institute for Earth Science and Disaster Pre-
vention) provides the real time rainfall information Photo 1. Typical slope collapse by rainfall.

907
proper prediction method based on a rational sol Moisture content gauges (VW1∼VW6) and water
investigation. Their challenging numerical analysis pressure gauges (PW1 & PW2) were installed. Water
has been developed on finite element method, based level gauges (1∼10) were also installed. Displace-
on seepage analysis and deformation analysis. The ments are measured with laser sensor (L1 & L2) dial
numerical procedure could simulate at displacement gauges (D1 & D2) and cupper thin wall (G1∼G4 &
of slope by rain at a site of slope in Hiroshima Japan G5∼G8).
(Sasahara, Sakai, Sun & Sakajo, 2006). This is very About the used soil material, its fineness content is
unique because many researches focus on either water 3.1% and sand content is 96.9%. Particle size of sand
(Tohari, Nishigaki & Komatsu, 2007, Orence, Shi- particle is 0.211 mm of D50, with the uniformity factor
moma, Maeda & Towhata, 2004) or deformation of 1.946. The maximum dry density is 1.571 (g/cm3 )
(Sasahara, Ebihara, Tsunaki & Tsunakai, 1996). Fur- and the minimum dry density is 1.240 (g/cm3 ). Then
thermore, the numerical procedure could explain the it was compacted to be 1.34 (g/cm3 ), which is loose.
difference of deformations by the different rainfall pat- The suction pressure and water content curves on
terns (Sasahara Kurihara, Sakai, Sun & Sakajo 2007). wetting and drying processes obtained as shown in
A further necessary aspect is, of course, to know the Figure 2. Two curves in absorption and drainage of
more precise mechanism of slope collapse. Then, the water were tested, which are not much different each
authors conducted a series of model tests, which is
limited to the sandy slopes in the laboratory. The
mechanism of displacement prior to the slope collapse
was focused from the points of water and deforma-
tions. Then, two different patterns of deformations
were observed, which is quite interesting to explain the
collapse mechanisms on sandy slope prior to failure.

2 MODEL TESTING

2.1 Test device and soil material

Figure 1 shows model test container with sensors,
which were to monitor the deformations, changes
of water contents and pressures precisely in order to
investigate the mechanism of slope collapse in realiz-
ing the different conditions. Size of container is with
a height of 60 cm and a length of 200 cm. Slope angle
is 45 degrees, which is quite steep.

Figure 2. Suction and saturation curve.

Figure 1. Model test container with sensors. Figure 3. Initial water table for two cases.

908
other. The air entry suction pressure 1 to 3 kPa. The
coefficient of permeability of saturated soil is 2.0 ×
10−2 cm/sec.

2.2 Collapse procedure

Two model tests of slope collapse by rainfall were
explained. One is under rainfall and another is with
initial ground water flow prior to rainfall. The former
is Case-1 and the later is. Case-2. Case-2 had a ground
water flow and rainfall, where rainfall started after the
first 9,000 sec with ground water flow. Rainfall inten-
sity for the both cases is 150 mm/h, provided from
water pipes with small holes at the searing. Figure 2
shows the initial water tables differences at rainfall
start. The initial water flow in case-2 was made by
supplying from water tank at the top of container until
to form 10 cm high water table from the bottom of
container. And, from many preliminary trials of test-
ing, gravels were installed on the tip with two stepped
gentle slopes, to avoid the failure stopping.

2.3 Test results

Figures 4 & 5 show test results of Case-1 and Case-2,
with three sets of figures. From the top, the first figure
show the saturation degree (VW1 to VW3) and water
pressure curves (P1) on elapsed time, for the section Figure 5. Monitoring data (Case-2).
located at the lower portion of the slope. The second
figure shows the saturation degree (VW4 to VW6) and water pressure curves (P2) on elapsed time, for
the section located at the upper portion of the slope.
The last figure shows surface displacements along
the slope measured with laser sensors (L1 & L2) and
with strain gauges (D1 & D2).
<Case-1>
From the top figure in Figure 4, the saturation
degrees increased on the stepped curve with three
stages at the each depth VW1, VW2, VW3 at the lower
position of slope. VW1 at the highest position began to
increase earlier at the first saturation stage than VW3
at the lowest position. VW2 comes between them. But
this order of VW1 and VW3 became inversed at the
second and third saturation stages because of the water
penetrated from rainfall rose up to near the slope sur-
face. The water pressure PW1 increased from the third
stage of water saturation, which shows the delay of
water table formation in the ground from rainfall.
From the second figure, the water saturation also
increased on the stepped curve with three stages at the
each depth VW4, VW5, VW6 at the upper portion
of slope. The order of saturation magnitudes at VW4,
VW5 and VW6 changes on elapsed time with the same
manner of VW1, VW2 and VW3. The water pressure
PW2 increased from the third saturation stage as well
as PW1 by the same reason. The all sensors at the
lower position of slope increased earlier than those
at the upper position because penetrated water easily
Figure 4. Monitoring data (Case-1). gathers at the down slope.

909
 Meanwhile, from the third figure, the displace- Figure 6 shows the result of Case-1. It can be seen
ment on the slope surface increased corresponding to that the lateral displacements deformed like a bending
the water pressure increased gradually. All the sen- column. From this, it was accumulated increasingly
sors showed the almost same values. The first small from the bottom around the depth of 20 cm to the sur-
increase was observed at the time almost 5,800 sec face on time passing. Soil was getting much weaker at
passed. The second large increase was observed at the the surface side by the rainfall and this slope collapse
time 9,800 sec passed. Therefore, two turning points seems to be sallow slope failure. At the time 5,400
of displacement increase were observed on the curve sec passed, lateral displacements started to increase
in Case-1. Then it finally yielded a slope failure after drastically to the failure. Comparing the lower with
the second turning point, where only PW2 at the upper the upper position, G1∼G4 showed smaller displace-
position of the slope. It suggested that the propagation ments at the depth below 20 cm than G5∼G8, because
of water table to upstream caused a failure. G1∼G4 at the lower position is located near at the end
<Case-2> of slope, where meets another flat slope.
Case-2 has water flow in the slope for the first Figure 7 shows the results of Case-2. There are a big
9,000 sec. Therefore, water table was already made difference between G1∼G4 at the lower position and
before rainfall. From the top figure in Figure 5, the G5∼G8 at the upper position. In the position, the lat-
saturation degree increased on the stepped curve with eral displacements deformed like a bending column as
two stages at the each depth VW1, VW2, VW3 at the well as Case-1. However, the lateral displacement was
upper position of slope. On the contrary with Case-1,
VW3 at the lowest position began to increase earlier
at the first saturation stage than VW1 at the highest 0 0

position. The position is, the higher the higher mag-

nitude is. VW3 at the lowest position is always higher 10 10
than VW2 and VW1 at the higher positions, although
VW2 and VW1 increased to catch up VW3 gradu-

Depth H (cm)
Depth H (cm)

ally. The water pressure PW1 increased from the first 20 20

saturation stage. This is very different from Case-1.
Unfortunately, the measurement of PW1was stopped
30 30
by accident. t=1800s
From the second figure, the water saturation degree t=1800s t=3600s
also increased on the stepped curve with two stages at t=3600s t=5400s
40 40
t=5400s t=7200s
the each depth VW4, VW5, VW6 at the upper position t=7200s t=9000s
of sloe. The order of saturation magnitudes of VW4, t=9000s
50
VW5 and VW6 are same with VW1, VW2 and VW3. 50 -2 0 2 4 6 8 10
-2 0 2 4 6 8 10
The water pressure PW2 increased gradually from the Accumlated displacement d (mm)
Accumlated displacement d (mm)
upper
first stage of saturation degree as well as PW1. But (i) G1 G4(Lower) (ii) G5 G8(Upper)
PW2 showed a peak then to be decreased at the final
stage, where soil might be fully saturated and softened Figure 6. Lateral ground.
by failure.
Meanwhile, the displacements on the slope surface 0 0
started to increase corresponding at 9,000 sec passed,
when the rainfall started. Then it shortly became large 10 10
drastically to yield slope failure. This is much larger
than Case-1. By comparing these two cases, it was
Depth H (cm)

Depth H (cm)

found that the initial seepage is quite influencing to 20 20

the slope failure mechanism.
30 30
2.4 Deformation and mechanism t=9000s
t=9000s
t=9200s t=9200s
Two different slope deformation processes to collapses 40
t=9400s
40
t=9400s
were obtained with the same conditions of slope angle t=9500s t=9500s
t=9550s t=9550s
and rainfall. In order to investigate slope collapses 50 50
from the point deformation of slope, the authors mea- -15 -10 -5 0 5
Accumlated displacement d (mm)
0 5 10 15
Accumlated displacement d (mm)
sured lateral ground displacements to the depth. Figs. 6 displacement (Case-1)
and 7 show the lateral displacements at G1∼G4 at (i) G1 G4(Lower) (ii) G5 G8(Upper)
the lower position of slope and G5∼G8 at the upper
positions of slopes for Case-1 and Case-2 respectively. Figure 7. Lateral ground displacement (Case-2).

910
not like this at the upper position. It shows an inverse
trend of displacement increase at the upper position
because the water table increase from the bottom and
waterfall increase from rainfall yielded complicated
weakness of soils in the slope. Focusing the lower posi-
tion, lateral displacements seems rather small until the
time 9,000 sec passed and then it started to increase
drastically to the failure. The depth of occurrence of
lateral displacement is almost same at the depth upper
30 cm, although Case-2 had the lager slope failure than
Case-1.

3 NUMERICAL ANALYSIS

Furthermore, the authors have conducted to simulate

these results. As aforementioned, the failure mech-
anism is complicated, coupling water behavior with
deformation. The authors have proposed the new
numerical method, which could couple seepage anal-
ysis and deformation analysis on the finite element
method (Drager-Prager model considering strength
reduction with saturation). Most of numerical anal-
ysis has not used actual rain records although it is
very important to judge stability of slopes. The defor-
mation varies very much by the changes of rain fall
volume and sequence. If water content increases by
Figure 8. Computed deformation (Case-1).
rainfall and seepage, the loss of strength generates
the deformation. The used soil parameters are tabu-
lated in Table 1. The suction and saturation curve was
employed as shown in Figure 2.
The computation results of Case-1 is shown in
Figs. 8 to 10. Figure 8 shows the deformations at
the time 1,800 and 10,800 sec passed, respectively.
Figure 9 shows the saturation degree at the time 1,800
and 10,800 sec passed, respectively. Figure 10 shows
the computed lateral displacement at G1∼G4 and
G5∼G8.

Table 1. Soil parameter used in the analysis.

Parameter Values

The coefficient of
permeability, ks 0.0199 (cm/sec)
Unsaturated properties See Figure 2
Saturated volumetric water
content, θ s 0.49
Unit weight, γ 1.85 (gf/cm3 )
Specific gravity, Gs 2.73
Frictional angle, φ 35 (degree)
Void ratio, e 1.02
Cu with suction Su
(Sasahara et al., 2006) Cu (Su) = 0.2816 * Su0.433
Young’s Modulus, E 200 (kgf/cm2 )
Poisson’s ratio, ν 0.3
Figure 9. Computed saturation degree (Case-1).

911
 5 CONCLUSIONS

Two experimental model tests were conducted in a

sandy steep slope with angle of 40 degrees. Focus-
ing the saturation degree and deformation, the authors
investigated the mechanism of the slope collapse
precisely. The main conclusions were obtained as
follows.
1. If the saturation increases by rainfall, the deforma-
tion of ground of slope increases. Generally, the
surface displacement should be monitored because
-1). it is weakened more than the lower portions.
2. Saturation degrees increase with two or three
upper
(i) G1 G4(Lower) (ii) G5 G8(Upper) stepped stages prior to slope collapse. The final
stage is where saturation degree is nearly 90%. To
Figure 10. Computed lateral displacement (Case-1). find this stage must be the key to the prediction.
3. From Case-2, slope might be collapsed drastically
where water flow was made to prior to rainfall stat.
In this case, a rather large slope failure will be
occurred by softening of soils.
4. The used numerical analysis could explain the
deformation and seepage by rain in general,
although there are some limitations. It will be use-
ful to simulate the mechanism of slope collapse by
rainfall.

REFERENCES

Okayasu, T., Hadano, R., Inoue, E., Mitsuoka, M. &

Sakajo, S. 2007. Possibility of application of dis-
Figure 11. Schematic displacement and time passing. tributed agricultural information network to geotechnical
field. The 43rd Annual Conference of JGS, Nagoya,
pp.2031–2032 (in Japanese).
4 REMARKS Orense, R.P., Shimoma, S., Maeda, K. & Towhata, I. 2004.
Instrumented Model Slope Failure due to Water Seepage.
Through the results of two model tests including Journal of Natural Disaster Science, 26 (1), 15–26.
numerical analysis, the clear difference between Sakai, N., Yamakoshi, T., Kuriara, J., Sasahara, K. & Morita,
K. 2006. Modeling and the deformation characteristics
Case-1 and Case-2 can be summarized on the rela- of slope of weather granite. Annual conference of Japan
tions between displacement and rainfall elapsed time Society of Erosion Control Engineering, Wakayama,
as shown in Figure 11. pp.128–129 (in Japanese).
As seen in this figure, Case-1 has two turning points Sasahara, K., Ebihara, K. & Tsunaki, R. 1996. Experimental
to increase the surface displacements but Case-2 has study on mechanism of steep slope failure. Journal of land
only one. In Case-1, it is easy to predict slope collapse slide, 32 (4), pp.1–8 (in Japanese).
by monitoring the increase of displacement between Sasahara, K., Kurihara, J., Sakai, N., Sun, Y. & Sakajo, S.
the first and second turning points. However in Case-2, 2007. Prediction of surface displacement and collapse
it seems not easy to predict. If a water flow with a cer- critical time on slopes with rain penetration by a FEM
analytical technique. Proc. The 43rd Annual Conference
tain water table was made in the slope prior to the of JGS. Nagoya, pp.22–23 (in Japanese).
slope failure by some reasons, monitoring displace- Sasahara, K., Sakai, N., Sun, Y. & Sakajo, S. 2006. Pro-
ment for prediction of slope collapse could not work posal of slope failure analysis on 2D FEM. The 42nd
for a sudden slope failure. In such cases, the results of Annual Conference of JGS, Kagoshima, pp.2191–2192
Case-2 suggests the importance of monitoring satura- (in Japanese).
tion degree and water pressure instead of displacement Tohari, A., Nishigaki, M. & Komatsu, M. 2007.
monitoring. One of author has proposed a water con- Laboratory rainfall-induced slope failure with mois-
tent monitoring widely based on internet techniques ture content measurement. Journal of Geotechnical and
in the mountainous area (Okayasu, Hadano, Inoue, Geo-Environmental Engineering, 333 (5), pp.575–
587.
Mitsuoka & Sakajo, 2007).

912
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Calibration of a rheological model for debris flow hazard mitigation

in the Campania region

A. Scotto di Santolo & A. Evangelista

Department of Geotechnical Engineering, University of Naples Federico II, Naples, Italy

ABSTRACT: The paper reports the results of the back-analyses of the propagation of debris flows in pyroclastic
(deposits in the Campania region, Italy. 57 well-documented case histories were analyzed using the 2-3D
DAN-W) code (Hungr, 2003) with two different rheological models: the frictional model and the Voellmy
model. The latter produces the more consistent results in terms of total runout, debris spread and distribution as
well as velocity data. The results show that it may be possible to model past events reasonably accurately using
the Voellmy model. Although it is difficult to make predictions about future landslides, the calibrated model
could be used to predict their propagation if the detachment area and the morphology are known.

1 INTRODUCTION The paper reports calibration procedure of the most

suitable rheological model for the analyzed debris
Rapid long runout landslides represent a difficult chal- flows.
lenge in hazard studies because they pose a risk to areas
situated a considerable distance from the source. The
prediction of runout distance, flow velocity and depth 2 CASE HISTORIES
(hereafter referred to as dynamic parameters) are nec-
essary for designing protective measures and are a key 2.1 Geotechnical properties of the pyroclastic
requirement for the delineation of the hazard zone. The deposits
best existing prediction methods rely on empirical rela-
tionships between volume, travel distance, angle of 57 debris flows that took place in the Campania region
reach or Fahrböschung etc. (e.g. Scheidegger, 1972, between 1973 and 1998, which are well-documented in
Corominas, 1996, Scotto di Santolo, 2000; Fannin & the literature (Del Prete et al., 1998; Scotto di Santolo,
Wise 2001) but they do not make it possible to predict 2000; Di Crescenzo and Santo, 2005) were analyzed
all the dynamic parameters. (figure 1). The events occurred in the following zones:
This paper is concerned with different kinds of 1. The Phlegrean Field of which the city of Naples is
debris flows that have occurred in the pyroclastic a part;
deposits of the Campania region (southern Italy). 2. The carbonatic ridge comprising the Sorrento
According to other researchers (Di Crescenzo & Santo, peninsula (Monti Lattari) and the mountains of
2003; Scotto di Santolo, 2000), three main types of Sarno-Quindici.
landslides have been detected: un-channelled, chan-
nelled and mixed debris flows. 57 well-documented The soil concerned regards the most recent pyro-
case histories which were suitable for back-analysis clastic deposits deriving from the volcanic activity of
were selected. Each of the case histories was analyzed the Phlegrean Fields in zone 1 and from the volcanic
in 3D dimensions using DAN-W (Dynamic Analysis activity of mount Somma/Vesuvius in zone 2. The
of Landslides, Hungr, 2003) with two different rheo- substratum underlying the above-mentioned cover is
logical models: the frictional model and the Voellmy of the same volcanic nature in zone 1 and of a car-
model (1955). bonatic nature in zone 2. In granulometric terms, the
The results of each analysis were evaluated by unstable cover consists of sandy silts or slightly clayey
matching the following parameters to the values as or gravelly silty sands. The clayey part is slightly
determined from maps and on site survey: total hori- plastic though only in the Vesuvian deposits. The grav-
zontal distance (or runout) and flow velocities. Unfor- elly part mainly consists of pumices and, to a lesser
tunately, only in a few cases do we have information extent, of scoriae and lapilli. The particles are mainly
about the depth of the deposition fan. siliceous. Their structure is amorphous and porous;

913
Figure 1. Location of the study area and debris flows.

there is a double porosity system inter and intra parti- cover, not only does it increase the level of satura-
cle (not connected to the surface). The specific gravity tion of the soil, thereby causing the existing suction
of solids is 2.54. Porosity exceeds 70% for Vesuvian to diminish, but it also, and simultaneously, creates
deposits while it is lower for the Phlegrean deposits. more unfavourable conditions of water circulation in
The cover is partially saturated; the level of saturation the portion of soil closest to the surface according to
varies according to weather conditions. in situ suction monitoring (Scotto di Santolo, 2000;
Mechanical behaviour is extremely variable; the Scotto di Santolo & Evangelista, 2004; Evangelista
peak friction angle depends on stress, density and is et al., 2007).
a function of the degree of saturation. The ultimate The post failure behaviour (or mobility) of the land-
friction angle of the material varies between 35◦ and slide is conditioned by several factors such as the
37◦ ; the higher value relates to the Vesuvian deposits geometry of the slope, the mechanical properties of
(Picarelli et al., 2006). From a phenomenological per- the deposits cover (porosity, grain size distributions)
spective, it can be observed that the shear strength and the characteristic of rainfall before and after the
of non-saturated deposits, in terms of total stress, triggering. This is clearly demonstrated by the com-
gives rise to un intercepted cohesion, known as appar- parison between the channelled flows and the flows
ent cohesion, which increases with suction according on the open slope (unchannelled). In the former, the
to Fredlund and co-workers criteria (1978) (Scotto presence of the channel allows a higher concentra-
di Santolo, 2000b; Picarelli et al., 2006). Mechani- tion of water and therefore greater fluidification of
cal behaviour in conditions of partial saturation still the landslide mass.
requires more detailed research. Edometric compres- According to some authors, the triggering mech-
sion tests and direct shear tests revealed that, at a anism is attributed to the development of an
constant total stress, the reduction in suction always undrained mechanism (static liquefaction) (e.g. Sassa,
leads to structural collapse and a reduction in strength 1988). However, in the opinion of the present
(Scotto di Santolo et al., 2000). authors, there are still uncertainties regarding the
The reduction in shear strength due to the infiltra- role played by liquefaction during the triggering
tion of rainwater is considered to be the triggering phase of unsaturated deposits; it is believed that it
mechanism of the landslide phenomena considered in contributes to subsequent fluidification (Eckersley,
this paper. When rainwater seeps into the pyroclastic 1980).

914
2.2 Geometrical and dynamic characteristics
of landslides a)
An inventory of 57 debris flows was collected con-
taining geomorphological, geotechnical and dynamic
data. In table 1 some of the available informa-
tion was reported. The landslides were divided into
three distinct types: un-channelled, channelled and
mixed debris flows (Figure 2). For each flow, the
following information was evaluated according to
Figure 3:
• morphometric data: (slope angle of the crown and
the sliding and deposition zone and of active zone
αa , the extent of the area, thickness, the difference
in height between the crown and the toe of landslide
H, trigger volume V)
• dynamic data (runout L)
• unstable deposit type and bedrock
• geotechnical properties.
30% of the landslides were of the channelled type,
28% were un-channelled while 42% were of a mixed b)
typology. The angle of reach (tan (H/L)) was related
to the volume of the landslide and to the travel dis-
tance as reported in the literature. As volume increases,
so does mobility according to other observed flows
(Corominas, 1996; Finlay et al., 1999). However this

Figure 2. Main types of landslides detected: a) un-

channelled (Nocera, 2005); b) mixed (Pozzano, 1997) debris
flows.

relationship appears to be scattered. The relationship

with H was plotted in figure 4a and was selected due
to the higher degree of correlation. The angle of reach
decreases with H.
Landslides in the Phlegrean Fields are character-
ized by H less than 200 m and have a high angle of

915
 1995). It is based on shallow flow assumptions (H L,
with H flow depth normal to the base). This dynamic
model is based on the Lagrangian solution of St.
Venant’s equation. This equation can be derived by
applying the conservation of momentum to thin slices
of flowing mass which are perpendicular to the base
of the flow. These ‘‘boundary blocks’’ divide the slide
mass into n ‘‘mass elements’’ of constant volume.
The following input data were used: the trigger vol-
Figure 3. Definition of term. ume, the geometry of the slope (2D), the wide of the
channel and the constant erosion depth. The code gave
a) 60 as an output the velocity and thickness of the slid-
50 ing mass both along the slope and at a pre-specified
location along the path as functions of time.
angle of reach (˚)

40
Eight rheologies are available in DAN-W (Hungr,
30 1995). In this study, however, only two rheological
20
models were analyzed:
10 • The Friction model, where the flow resistant term
was controlled by the effective normal stress on the
0
0 200 400 600 800 1000 base of the boundary block;
H (m) • The Voellmy fluid model (1955), where the resis-
tance is a function of a friction term and a turbulent
b) 0.3
N=57 term (Hungr, 1995).
0.25
The basal flow resistance term is governed by the
frequency

0.2 rheology of the material and can be expressed thus:

0.15

0.1 ac
τ = γ Hi cos αi + · (1 − ru ) · tan ϕ (1)
0.05 g
0
30 34 38 41 45 49 52 56 60 63 for friction flow

ac v2
Figure 4. Relation between morphometric data. τ = γ Hi cos αi + μ+γ (2)
g ξ

reach. Those in the Vesuvian area have a lower angle for a Voellmy fluid
of reach where H is less than 400 m. For higher levels of where γ is the unit weight of the flowing mate-
H (the mountains of Sarno), the angle of reach is still rial, ac = vi2 /R is the centrifugal acceleration resulting
lower and decreases with H until a stationary value. from the vertical curvature of the flow path R, ru is
Therefore the debris flows with H higher than 400 m the pore-pressure coefficient (ratio of pore pressure to
displaying greater mobility depending on the fluidi- total normal stress at base of boundary block), ϕ is
fication of the mass (usually for channelled kind). In the friction angle; μ is a friction coefficient and ξ is a
figure 4b the slope of the active zone αa (crown and turbulence coefficient with dimensions of [m/s2 ].
sliding zone) were also reported. This slope is greater
than 30◦ with a maximum frequency of about 38◦ .
For each landslide the longitudinal and transver- 3.2 Calibration procedure
sal profiles were plotted (scale 1/5000 or 1/2000) and As already mentioned, the friction model and the
were then used for the numerical analyses with the Voellmy models were selected from the rheological
calculation program DAN-W (Hungr, 2003). models in the calculation program. The model was
calibrated with reference to:
3 NUMERICAL ANALYSES
• distance L, calculated horizontally between the edge
of the mound and the crown;
3.1 Introduction
• velocity of the front vf ;
DAN-W is a windows-based program used to model • and, only in certain cases, with respect to the height
the post-failure motion of rapid landslides (Hungr, of the deposit.

916
 0.7 the phenomena involved. The use of this rheologi-
I VA M all n=57
0.6 cal model made it possible to consider the dissipative
0.5 effects of the chaotic motion of water and solid grains
frequency

0.4 together.
0.3 It was also observed that with the frictional model,
0.2
velocities were significantly overestimated (Hungr,
0.1
1995). Moreover, Voellmy’s model has been widely
0
confirmed in similar contexts (Hungr & Evans, 1996;
0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.1 0.15 0.18 0.2 Fiorillo et al., 2001; Revellino et al., 2004) also using
three-dimensional geometric models (Pirulli, 2004;
Mc Dougall & Hungr, 2005) both in terms of distances
Figure 5. Results of the back analyses with Voellmy model
(ξ = 100 m/s2 ).
travelled and in terms of the velocity of the volumes
mobilized.
The interval of variation of the parameters ξ and
5000 μ already used for the landslides in Campania was
acquired from bibliographical research. The analyses
4000 were carried out by letting μ vary over an interval of
values ranging between 0.01 and 0.2 and letting ξ equal
to 100 and 200 m/s2 . Figure 5 shows the range of val-
3000 ues of the parameter μ, having set ξ at 100 m/s2 so that
LDAN (m)

LDAN was quite equal to L. It was observed that, for

2000 the channelled flows, the most frequent value of μ was
0.06, while the value for the flows on open slopes was
0.18 and was 0.03 for mixed. These values correspond
1000 I M VA to the observation of greater mobility of channelled
landslides compared to landslides on open slopes for
0 which dissipative phenomena occurred, linked to the
0 1000 2000 3000 4000 5000 widening of the transversal section along the path.
The two values chosen for parameter ξ lead to practi-
cally coincident distances travelled, but at significantly
Figure 6. Comparison between numerical runout LDAN and different velocities. The best fit μ value varies in a
measured ones L with Voellmy model (μ = 0.03 and ξ = wide range. The frequency is higher in particular for
100 m/s2 ). three values of μ: 0.03, 0.06 and 0.18. Thinking to a
possible prevision of this phenomenon through these
values for safety results the study was carried out with
With regard to velocity, it should be noted that a value of 0.03 for μ and 100 m/s2 for ξ .
there is no on site data, only estimates obtained from Figure 6 compares the observed distances and the
the back analysis of the damage caused by the flows distances calculated using these parameters for chan-
that took place in the area of Sarno-Quindici (Faella nelled, unchannelled and mixed flows. For flows on
and Nigro, 2003) or from semi-empirical expressions open slopes (modest value of L), it was observed that
(Zanchetta et al., 2004). the friction coefficient μ had a reduced influence.
In the analyses undertaken for the study, the model For channelled flows, on the other hand, a perfect
was considered valid on the basis of a comparison of coincidence was observed between the measured and
the distance (LDAN ≥ L). The second parameter for calculated results for L up to 1500 m. For greater
calculation was the maximum speed of the front which lengths, the model displays a slight tendency to over-
was set as no greater than 12 m/s, in the absence of estimate the theoretical routes (according to the results
experimental data. in Figure 4). In table 2 the synthesis of the results were
The results obtained were analysed together and reported.
the flows were then separated according to the spe- Figure 7 shows the velocity profiles of the front and
cific route (channelled flow, un channelled flows or a rear, obtained with DAN using the above-mentioned
mixture of the two). model for the three kind of path analyzed. It can be
seen that the range of the velocities reaches a maxi-
mum at around the trigger area and then decreases
almost with exponential law for the all data. Of course
3.3 Results of numerical analyses
the used model overestimates the distance for un-
Of the two models that were analysed in this study, channelled and channelled ones. For this reason it
Voellmy’s model proved to be better for interpreting could be better using the Voellmy model with three

917
Table 2. Synthesis of the results with Voellmy model (μ = runout distance, flow velocity and depth are neces-
0.03 and ξ = 100 m/s2 ). sary for planning and designing protective measures
and are a key requirement for delineating the haz-
LDAN /L ard zone. 57 well-documented case histories which
were suitable for back-analysis were selected. Each of
L (m) Mean value Dev. ST the case histories was analyzed in 3D using DAN-W
0 ÷ 500 1.18 0.14 (Dynamic Analysis of Landslides, Hungr, 2003) with
500 ÷ 1000 1.06 0.09 two different rheological models: the frictional model
1000 ÷ 4000 1.04 0.06 and the Voellmy model. The results of each analysis
were assessed by matching the following parameters
to the values as determined from maps and on site
12 600
a) survey: total horizontal distance (runout) and flow
10 slope 500
profile
velocities. The Voellmy model produces the most con-
8 400 sistent results in terms of total runout, debris spread
V (m/s)

H (m)
6 300 and distribution, as well as velocity data.
4 rear front 200
The results show that past events can be mod-
velocity velocity elled with reasonable accuracy using the Voellmy
2 100
model; however, it is still difficult to make accurate
0 0 predictions concerning the most likely runout.
0 200 400 600 800 1000
L (m)
12 300
b)
10 slope 250
profile REFERENCES
8 front 200
V (m/s)

H (m)

6 velocity 150
rear Ayotte, D. & Hungr, O. 2000. Calibration of a runout predic-
4 velocity 100 tion model for debris-flows and avalanches. Wieczorek G,
Naeser (eds) Debris flows hazard mitigation. Balkema,
2 50
Rotterdam, pp 505–514.
0 0 Corominas, J. 1996. The angle of reach as a mobility index
0 50 100 150 200 250 300 350 for small and large landslides. Can Geotech J 33:260–271.
L (m) Di Crescenzo, & Santo, A. 2005. Debris slides-rapid earth
12 500 flows in the carbonate massifs of the Campania region
c) 10 (southern Italy): morphological and morfometric data for
400
slope evaluating triggering susceptibility. Geomorphology, 66,
8 pp. 255–276.
profile
V (m/s)

300
H (m)

6 front Del Prete, M., Guadagno, F.M. & Hawkins, A.B. 1998. Pre-
velocity 200 liminary report on the landslide of 5 May 1998. Bull Eng
4
Geol Environ 57:113–129.
2 100
rear velocity Eckersley, D. 1990. Instrumented laboratory flowslides.
0 0 Geotecnique J., 40, pp. 873–885.
0 100 200 300 400 500 Evangelista, A. & Scotto di Santolo, A. 2001. Mechanical
L (m) Behaviour of unsaturated pyroclastic soil. Proc. Land-
slides: Causes, Impacts and Countermeasures, Davos-
Figure 7. Slope and velocity profile calculated with DAN: Switzerland, June 2001.
a) Channeled; b) Un-channeled; c) Mixed. Evangelista, A. & Scotto di Santolo, A. 2004. Analysis and
field monitoring of slope stability in unsaturated pyroclas-
tic soil slopes in Napoli, Italy. Proc. 5th Int. Conf. on Case
Histories in Geotechnical Engineering, New York 2004.
different values of friction coefficient according to Faella, C. & Nigro, E. 2003. Dynamic impact of debris flows
figure 5. The velocity profile is very useful for pre- on the constructions during the hydrogeological disaster
dicting the dynamic parameters of future events and in Campania 1998: failure mechanical models and evalua-
therefore for mitigating risk in similar zones. tion of the impact velocity. Proc. Int. Conf. FSM, Naples,
1, pp.
Fannin, R.J. & Wise, M.P. 2001. An empirical-statistical
model for debris flow travel distance. Can Geotech J, 38,
4 CONCLUSION pp. 982–994.
Fiorillo, F., Guadagno, F.M., Aquino, S. & De Blasio, A.
Rapid long runout landslides represent a difficult 2001. The December Cervinara landslides: further debris
challenge in hazard studies because they endanger flows in the pyroclastic deposits of Campania (southern
areas situated far from the source. Predictions of the Italy). Bull Eng Geol Environ, 60: 171–184.

918
Fredlund, D.G., Morgenstern, N.R. & Widger R.A. 1978. The Scheidegger, A.E. 1973. On the prediction of the reach
shear strength of unsaturated soils. Canadian Geotech. J., and velocity of catastrophic landslides. Rock Mech 5:
15: 313–321. 231–236.
Hungr, O. 1995. A model for the runout analysis of rapid Scotto di Santolo, A. 2000a. Analisi geotecnica dei fenomeni
flow slides, debris flows, and avalanches. Canadian franosi nelle coltri piroclastiche della provincia di Napoli.
Geotechnical Journal, 32: 610–623. Tesi di dottorato XII ciclo Consorzio Università di Napoli
Hungr, O. & Evans, S.G. 1996. Rock avalanche runout e Roma, 1/2000.
prediction using a dynamic model. Proc. Landslides, Scotto di Santolo, A. 2000b. Analysis of a steep slope in
Senneset (ed.), Rotterdam: Balkema, 233–238. unsaturated pyroclastic soils. Proc. Asian Conference
McDougall, S.D. & Hungr, O. 2005. Dynamic modelling on Unsaturated Soils, Singapore, 569–574. Rotterdam:
of entrainment in rapid Landslides. Can. Geotech. J. 42: Balkema.
1437–1448. Scotto di Santolo, A. 2002. Le colate rapide. Hevelius Ed.
Picarelli, L., Evangelista, A., Rolandi, G., Paone, A., srl, Benevento 2002, ISBN 88-86977-42-5.
Nicotera, M.V., Olivares, L., Scotto di Santolo, A., Voellmy, 1955. Uber die Zerstorungskraft von lawinen.
Lampitiello, S. & Rolandi, M. 2006. Mechanical prop- Bauzeitung, Jahrgang 73, 212–285.
erties of pyroclastic soils in Campania Region. Proc. Zanchetta, G., Sulpizio, R., Pareschi, M.T., Leoni, F.M. &
Int. Conf. on Natural soils, Singapore, 2006. Rotterdam: Santacroce, R. 2004. Characteristic of May 5–6 1998
Balkema. volcaniclastic debris flow in the Sarno Area (Campania
Revellino, P., Hungr, O., Guadagno, F.M. & Evans, S.G. Southern Italy): relationship to structural damage and haz-
2004. Velocity and runout prediction of destructive debris ard zonation. J. of Volcanology and Geothermal R. 133,
flows and debris avalanches in pyroclastic deposits, Cam- pp. 377–393.
pania region, Italy. Environmental Geology, 45: 295–311.
Sassa, K. 1988. Geotechnical model for the motion of
landslides. Proc 5th Int. Symp. on Landslides 1: 33–55.

919
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Optical fiber sensing technology used in landslide monitoring

Yan-xin Shi
Geo-Detection Laboratory, Ministry of Education of China, China University of Geosciences, Beijing, China

Qing Zhang & Xian-wei Meng

Centre for Hydrogeology and Environmental Geology, CGS, Baoding, China

ABSTRACT: Using the distributed optical fiber sensing technology in landslide monitoring, we can obtain
the main specialty of the landside and can improve the monitoring efficiency. In the paper, we introduced the
theory of FBG and BOTDR, respectively discussed their applied method, then, put forward a notion, BOTDR
combined with FBG to monitor landslide. In Canlian landslide, we laid the monitoring optical fiber on the entire
landside and used BOTDR to obtain the outline information of it. We installed FBG at the certain essential spots,
the strain fissures, to obtain their strain information. Thus we can monitor the landside from dot to line and
future to surface. Finally we obtained the completed strain information of the landslide.

1 INTRODUCTION two kind of the most representative distributed optical

fiber sensing technology.
The distributed optical fiber sensing technology is a
new technology in the domain of project survey. The
2.1 FBG
optical fiber sensor uses the light as the carrier and
transmission medium of information. It has a lot of FBG is one kind of the distributed optical fiber sensor
advantage, such as anti-electromagnetic interference, but the information it gained in the distance is not
anti-corrosive, high sensitivity, quick responds, light continuous. The refractive index of Bragg grating is
weight, small volume, variable shape, wide trans- periodically changed. If the period of FBG is not same,
port bandwidth, distributed survey and so on. It has its reflection light wave length is also different. When
widespread application in the aspect of online dynamic this kind of optical fiber with the Bragg grating is
monitoring such as the health monitoring of the high- under the stretch or compression or its temperature
rise construction, the intelligent building, the bridge, changes, its period will change, thus the reflected light
the highway and so on. We applied the distributed opti- wave length also will change. We measured the change
cal fiber sensor in the landslide monitoring in 2004 of the reflected light wave length and knew the strain
and has obtained the good effect (Han, Z.Y. & Xue, or the temperature (Chen, Y. 2003). The theory of FBG
X.Q. 2005). is shown in Figure 1.
Arranging certain FBG sensors on an optical fiber,
we used the multiplying technology (TDM, WDM
2 THE PRINCIPLE and so on) to construct the distributed sensing net-
work. Thus we can simultaneously monitor multi-
The optical fiber sensing technology measured the spots in the wide range, moreover may reduce the
change of certain parameters of the transmission light equipment number of monitor system, the length of
in fiber (for example, intensity, phase, frequency,
polarization condition and so on) to realize the mea-
surement of environment parameter. The distributed
optical fiber sensing technology has become the most
promising technology by its advantage (multiply, dis-
tribution and long distance transmission). It is the
development trend of optical fiber sensing technology
The optical fiber Bragg grating sensor (FBG) and the
Brillouin time domain reflection sensor (BOTDR) are Figure 1. The theory of FBG.

921
 Pump light

Backward Brillouin scattered light

stra BOTDR
in
ε

Power
Power
Figure 2. The distributed FBG monitoring system. VB(0) VB(ε ) Frequency

Power
VB(ε) y
DiZ1 c
sta Z2
nce VB(0) equen
Fr Z1 Z2 Distance

Figure 4. The theory of BOTDR strain measurement.

When the ambient temperature change is smaller

than or equal to 5◦ C, if longitudinal strain ε occurs
in the optical fiber, the Brillouin frequency shift VB
changes in proportion to that strain. This relation can
be expressed as:

VB (ε) = VB (0) + Cε (1)

Figure 3. The backward scattered spectrum in optical fiber.
where VB (ε) is the Brillouin frequency shift size when
longitudinal strain ε occurs in the fiber, VB (0) is the
transmission fiber and reduce the cost of monitor-
Brillouin frequency shift size when no longitudinal
ing project. The distributed FBG monitoring system
strain occurs in the fiber, C is the strain coefficient,
is shown in Figure 2.
approximately equals to 50 MHz/με. The theory of
The typical measurement accuracy of FBG Sensor
BOTDR strain measurement is shown in Fig. 4.
is high at about 0.001%, and the typical distance res-
The optical fiber which the BOTDR sensing tech-
olution is also high at about 1 cm as determined by
nology used can be embeded in the substrate structure
the grating length. In addition, measurement systems
by the random form but does not affect the perfor-
that can measure strain in real-time have been realized.
mance because it may be curving for thin and soft.
These advantages have led to the FBG being applied
Only measured the power and frequency of the Bril-
to the precise strain measurement of structures, such
louin scattered light in the sensing optical fiber, we
as bridges and tunnel health monitoring. When FBG
could obtain the distributed strain and temperature on
sensors were used in strain monitoring, mainly used at
the fiber. The biggest merit of this sensing technol-
roughly identified positions where large deformations
ogy is that the optical fiber not only is the sensing
have occurred. Based on the FBG characteristic, when
part but also is the transmission medium. The tech-
we used it in it landslide monitoring, we mainly used
nology belongs to the distributed monitor technology.
it to monitor the change of the landslide backyard or
We may utilize it to realize long distance and uninter-
the known crack in real-time.
rupted monitor and easy to build the network with the
optical fiber transmission system to realize remote-
2.2 BOTDR measurement and control of the system. So embedded
the optical fiber in the landside mass as the neural net-
The Brilliouin optical time domain reflectometer uti-
work, we could implement the real-time monitor from
lizes the characteristic that the spectrum and power of
the line to the surface of the landslide.
backward Brilliouin scattered light is correlative to the
external environment (temperature, strain and so on).
The backward scattered spectrum in optical fiber is
3 THE APPLICATION IN LANDSLIDE
shown in Fig 3.
MONITOR
In general, the Brillouin scattered light is shifted in
frequency. When fiber materials are affected by tem-
3.1 The approach to landslide monitoring
perature or strain, the Brillouin frequency shift size
by optical fiber sensing technology
will change. Therefore, we can measure the shift size
of backward Brillouin scattered light to realize the dis- The distributed optical fiber sensing technology has
tributed temperature and strain measurement (Shi B the widespread application in many domains by its
et al. 2004). unique merit. If utilize the distributed optical fiber

922
sensing technology in the landslide monitoring, we
must discuss the following questions (Wang, A.J.
et al. 2006).
The most important is how the optical fiber be
embedded in the landside mass. Only when the optical
fiber and the monitored object were coupled very well
in together, the information we obtained from the fiber Figure 5. The laying method of optical fiber.
is real effective.
Next, the distributed optical fiber sensing technol-
ogy took the optical fiber as the sensing part, so the
choice of fiber type also is a monitoring key. fiber continuously from bottom to top is arranged
Furthermore, it is also a question that how cement along the landslide body, then, continuously from left
the fiber and how build the optical fiber monitoring to right (or from right to left) is made the snake-shaped
network in order to obtain the comprehensive moni- arrangement along the horizontal direction. This type
toring information of the landside mass from the spot of network may monitor the strain changing in two
to the line and further to the surface. directions. When laying the optical fiber, we should
The natural landside mass mostly is natural rock and according to the characteristic of the work area to
soil mass which has not passed through the artificial determine the arrangement form of the optical fiber
change and mostly is covered the loose quaternary sys- network.
tem. If we want to use the optical fiber to monitor this In the actual monitor project, laying the opti-
type of landside mass, we must choose an appropriate cal fiber has two methods (shown in Figure 5): the
construction plan and craft. Perhaps we can carry on comprehensive contact type and the fixed-point
artificial processing to the landside mass, or we can contact type.
directly use the construction on the landside mass or The monitor motive of the comprehensive con-
other constructions influenced by the landside mass tact type is roundly monitoring the distortion and the
(for example, the concrete drain and stair, the slope monitor object is the entire landside mass. The mon-
control project, the road and the bridge etc.) to fix itor motive of the fixed-point contact type is mainly
and cement the optical fiber. For the landside mass monitoring the distortion of the crack, the stress con-
which has processed, we can choose the representa- centration area and so on. The monitor object is the
tive section plane on it to fix and cement the optical latent distortion point on the landslide. When laid the
fiber. Thus, we can really obtain the strain information optical fiber, we should according to the special details
of the landside mass through the strain of the optical of the landslide and the monitor motive to select the
fiber laid on it. different method.
In the landslide monitor, the optical fiber choice
is more important. We can enhance the monitor life
3.2 Combined FBG and BOTDR to monitor
by choosing the appropriate optical fiber. At present
landslide
there are two kinds of fiber usually used in monitor:
the bare fiber and the tight tube fiber. The monitor- FBG and BOTDR are two kind of optical fiber sens-
ing sensitivity of the bare fiber is very high, but its ing technology and have advantage and disadvantage
measuring range is small and easy to break off and lay- respectively. The sensitivity of FBG sensor is high that
ing the bare fiber is difficult. So the bare fiber adapts can measure the strain extremely accurately. We can
in the small strain monitoring. The tight tube optical use some FBG sensors to compose the FBG sens-
fiber is composed by the core, the envelope, the paint- ing network to realize the distributed monitor, but
ing layer and the protective tube. It has some merits: the sensing array which is used to response the sur-
the strong inoxidizability, the good waterproof perfor- rounding is set in advance, we must measure these
mance, the slightly big measuring range and not easy discrete distributed sensing spot respectively. There-
breaking off. Used the tight tube fiber in monitor is fore the monitor flexibility of FBG is low. The BOTDR
advantageous to the construction and can enhance the sensing part is the optical fiber. We may use it to real-
monitor life. Therefore we often choose the tight tube ize the long distance, the uninterrupted and distributed
fiber to monitor the strain of the landslide. monitor. But for its technical limit, the measurement
Generally arranging the optical fiber network distance resolution of BOTDR is only achieved 1 m.
has two forms: unidimensional network and two- If we combined FBG and BOTDR to monitor the
dimensional network. The former, namely, the optical landslide, we would offset the insufficiency of them.
fiber continuously from bottom to top is made the We laid the optical fiber in the entire landside mass and
snake-shaped arrangement along the landslide body. used BOTDR technology to obtain the outline infor-
This type of network is suited to monitor the strain mation of the entire landside mass. We installed the
changing in a direction. The latter, firstly, the optical FBG sensor on the essential distort spot (distortion

923
crack) in the landside mass and utilized its high mea-
surement sensitivity characteristic to obtain the strain
of the certain essential spots in landslide. Thus, unified
FBG and BOTDR to monitor landslide, we overcome
the shortcoming of BOTDR (measurement resolu-
tion is not high) and offset the insufficiency of FBG
(only realize the separate measurement). Through this
method, we may realize the landslide monitor from the
spot to the line and further to the surface and obtain
the completed strain information of landside mass.

3.3 Application example

The Canlian landslide is located in the heartland of
Wushan in Chongqing. It is river and valley slope
Figure 7. The FBG installation in field.
terrain. Although the government has carried on the
remedy against it, there are some obvious distortions in
the underside of the landslide. Based on this situation,
we have been using BOTDR technology to monitor
the landslide from August 2004. In October 2006, we
have installed the FBG strain sensor in its key distor-
tion spot.
The monitoring area is composed by two parts. On
the upside, we made grooves on the concrete trellis
which forms the control project and laid the optical
fiber in it. We arranged the fiber network with two-
dimensional form. On the underside, there is a cement Figure 8. BOTDR monitored distributed strain on the
stair passed through the landslide. We made grooves underside of Canlian landslide.
on the right of the stair and laid the optical fiber in
it. The fiber was laid with the section plane form. The
optical fiber network is shown in Figure 6. We used the plane. We found there were obvious tension and shear
optical fiber strain analyzer AQ8603 which made in fracture through the macro-survey. FBG monitor also
Japan to monitor the distributed strain of the landslide. showed that the distort was extended too. From Octo-
On 92–93 m of the stair, we installed the FBG strain ber 2006 to August 2007, the strain of the FBG
sensor. The monitoring equipment we used is FBG test point was changed from 37.65 με to 51.57 με.
demodulator which made by ourselves. Thus, we have obtained the strain distributed along the
Figure 8 showed the distributed strain along the section plane of the landside mass and strengthened the
optical fiber which laid on the underside of Cail- monitor to the essential distortion spot.
ian landslide by BOTDR monitored. Obviously, from
the figure we found 4 obvious high strain sections
along the optical fiber and the strain was symmetrical 4 CONCLUSION
(because the optical fiber was laid symmetrically). C1
was corresponded to the 92–93 m of the section plane. It’s just starts that FBG and BOTDR distributed optical
C2 was corresponded to the 142–143 m of the section fiber sensing technology were applied in the landslide
monitoring domain. We have obtained certain effect
in Canlian landslide monitor. But we still needed to
practice, summarize and perfect the approach to mon-
itoring landslide with FBG and BOTDR, such as the
fiber embedment craft, the fiber network arrangement
principle, the FBG sensor choosing and installing and
so on.
FBG and BOTDR unified to monitor landslide may
realize comprehensive monitor from the spot to the
line and further to the surface and may obtain more
complete strain information of the landside mass. The
Figure 6. The optical fiber laying network in Canlian method is one kind of perfect monitoring plan. But at
landslide. present, we still used two set of instrument systems

924
to realize FBG and BOTDR strain-measurement, the Chen, Y., 2003. The study on the measurement of strain in
FBG monitoring network and the BOTDR monitoring the buildings by using fiber grating sensors. Dissertation
network are two independent networks. We have not for the master degree in engineering of Nankai university.
really unified these two kinds of technique. It also is Shi, B, Xu, H.Z., Zhang, D., et al., 2004. Feasibility stu-
a development direction that how to unify these two dy on application of BOTDR to health monitoring for
large infrastructure engineering. Chinese journal of rock
kinds of technique in the landslide monitoring domain. mechanics and engineering, 23(3): 493–499.
Wang, A.J., Zhang, J.Y., et al., 2006. Application of BOTDR
in monitoring of Canlian landslide in WuShan. The
REFERENCES national fifth conference of geological hazard and control,
Chongqing, 299–305.
Han, Z.Y., Xue, X.Q., 2005. Status and development trend of
monitoring technology for geological hazards. The Chi-
nese journal of geological hazard and control, 16(3):
138–141.

925
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Finite element analysis of flow failure of Tailings dam and embankments

R. Singh
Department of Civil Engineering, Indian Institute of Technology, Kharagpur, India

D. Mitra
Department of Civil Engineering, National Institute of Technology, Warangal, India

D. Roy
Department of Civil Engineering, Indian Institute of Technology, Kharagpur, India

ABSTRACT: Undrained shear strength of cohesionless soils is often estimated from back analyses of dams
and embankments that have endured static undrained loading with various degrees of distress. The existing
procedures for back analysis assume isotropic material behavior. However, saturated cohesionless soils exhibit
strong inherent anisotropy during undrained loading. Consequently, the undrained shear strength of granular
soils depends on the angle between the major principal direction and the direction of deposition. Using an
anisotropic procedure for back analysis, correlations have been developed in this study between normalized
Standard Penetration Test (SPT) blow count or cone penetration resistance and anisotropic undrained shear
strength for different modes of loading (e.g. compression, simple shear and extension) from back analysis of
pre-failure geometries of 28 static flow failure case histories. To illustrate the efficacy of these correlations, finite
element analyses were carried out for three earth embankments using the shear strength parameters obtained from
the correlations proposed and the computed deformations were compared with the observations. The comparison
indicated a reasonable agreement.

1 INTRODUCTION of two embankments that underwent static undrained

loading were analyzed using shear strengths obtained
Undrained shear strength, su , of cohesionless soils from the proposed correlations.
is often estimated from back analysis of dams and
embankments that underwent various degrees of dis-
tress due to undrained loading (Olson and Stark 2 ANALYTICAL PROCEDURE
2003). Although the existing analytical procedures
are based on the assumption of inherently isotropic Undrained shear strength of cohesionless soils tends
material behavior, untrained mechanical response of to be higher when a sample is loaded in the direc-
saturated cohesionless soils is inherently anisotropic tion of deposition than that for loading in any other
(Vaid et al. 1990 and Yoshiminie 1998). A framework direction. Such behaviour is referred to as inher-
has been proposed recently accounting for inher- ent anisotropy. Inherently anisotropic undrained
ently anisotropic undrained behavior of granular soils behaviour of cohesionless soils can be approximated
(Singh and Roy 2006). In this paper, 28 case histo- testing on 84 undrained undisturbed frozen samples
ries involving static undrained response of dams and from seven sites of Canada for various types of load-
embankments have been back-analyzed using the pro- ing. The behaviour is numerically approximated by
posed procedure and to develop a set of correlations Equation 1.
between normalized cone tip resistance, qc1 , and Stan-
dard Penetration Test (SPT) blow count, (N1 )60 , and (su /σv )/(su /σv )TXC = 0.402 × cos 2θ + 0.598 (1)
peak undrained shear strength ratio, su /σv for various
modes of deformation, e.g., plane strain compres- where θ is the angle between the direction of depo-
sion and extension, and simple shear. To illustrate the sition and that of the effective major principal stress.
application of the correlations, pre-failure geometries The relationship, presented in Fig. 1, was developed

927
 Eq. [1] for until obtaining a value of (su /σv )TXC
by trial and error that gave a factor of safety
of 1.
• The above steps were repeated for other trial
surfaces until obtaining the minimum value of
(su /σv )TXC .

3 CORRELATIONS FOR su /σv

Pre-failure geometries of 28 embankments (Table 1)

that failed due to static undrained loading were back
analyzed using the procedure outlined in the preceding
section. The results of these back analyses are sum-
marized in Table 2 for anisotropic. Also included are
the results obtained from conventional isotropic back
analyses for comparison.
It appears from the results that the undrained shear
strengths from isotropic back analyses are in general
similar to the anisotropic undrained shear strengths for
simple shear loading.

Figure 1. Shear strength ratio for undisturbed sand (modi- Table 1. Case histories.
fied from Singh and Roy 2006).
Embankment qc1 , (N1 )60 Reference

Aberfan Tip No.4 2.2, 5.2 Lucia 1981

Aberfan Tip No.7 2.2, 5.2 Lucia 1981
using data from undrained monotonic laboratory tests Asele Embankment 3.8, 7.0 Konard and Watts 1995
on undisturbed (frozen) samples (Vaid et al. 1996). It Bofokeng Tailings 2.2, 7.4 Lucia 1981
is apparent from Fig. 1 that su /σv at phase transforma- Calaveras Dam 4.4, 8.0 Hazen 1920
tion in triaxial compression (denoted with subscript Fording South Spoil 1.3, 2.1 Dawson et al. 1998
‘‘TXC’’), is higher than those in simple shear (SS) and Copper Tailings Dam 2.1, 6.6 Lucia 1981
triaxial extension (TXE). It also appears that there is no Fort Peck Dam 2.6, 6.5 Casagrande 1976
systematic dependence of su /σv on the relative density Fraser River Delta 2.9, 5.3 Chillarige et al. 1997
Gypsum Tailings Dam 1.4, 4.7 Lucia 1981
of the deposit.
Greenhills Cougar 7 1.2, 2.4 Dawson et al. 1998
The following procedure was used for back analysis Helsinki Harbour 2.8, 6.0 Andresen and
to estimate the peak su /σv : Bjerrum 1968
Hoedekenskerke Dyke 4.7, 8.5 Koppejan et al. 1948
• A trial slip surface was first assumed through the Jamuna Bridge Site 3.2, 7.5 Yoshimine et al. 1999
zones of lowest penetration resistance in the pre Lake Ackerman Road 3.3, 7.0 Hryciw et al. 1990
failure configuration. Merriespruit Tailings 1.1, 3.2 Fourie and
• The undeformed section of the embankment above Papageoriou 2001
the assumed slip surface was divided into a number Mississippi River Bank 3.2, 6.8 Senour and
of vertical slices. Turnbull 1948
Nerlerk Slide1 2.6, 6.8 Sladen et al. 1985
• The mobilized undrained shear strength was
Nerlerk Slide2 1.5, 3.6 Sladen et al. 1985
assumed to be governed by the vertical effective Nerlerk Slide3 1.5, 3.6 Sladen et al. 1985
stress for the undeformed configuration. North Dyke 2.3, 4.5 Olson and Stark 2000
• Soils with qc1 ≥ 6.5 MPa or (N1 )60 ≥ 12 and Quintette Marmot 1.3, 2.8 Dawson et al. 1998
those above water table were assigned drained val- Sullivan Mine 1.8, 3.7 Davies 1999
ues of friction angle depending on their qc1 or Tar Island Dyke 1.2, 3.0 Mittal and Hardy 1977
(N1 )60 . Trondhiem Harbour 2.5, 6.0 Andresen and
• The stability analyses were performed using soft- Bjerrum 1968
ware package XSTABL Version 5.2 (1994) and Vlietepolder, Zeeland 2.8, 7.5 Koppejan et al. 1948
Western US Tailings 0.8, 3.0 Davies et al. 2002
the GLE method assuming the mobilization of
Wilheminapolder 2.8, 7.5 Koppejan et al. 1948
su at the base of the slip surface according to

928
Table 2. Pre liquefaction Su /σv .

Su /σv

Anisotropic

Embankment TXC SS TXE Isotropic

Aberfan Tip No.4 0.55 0.31 0.07 0.33

Aberfan Tip No.7 0.55 0.31 0.07 0.33
Asele Embankment 0.58 0.30 0.06 0.20
Bofokeng Tailings 0.53 0.28 0.06 0.17
Calaveras Dam 0.64 0.31 0.07 0.28
Fording South Spoil 0.34 0.19 0.04 0.19
Copper Tailings Dam 0.49 0.27 0.06 0.33
Fort Peck Dam 0.56 0.31 0.07 0.14
Fraser River Delta 0.53 0.30 0.06 0.16
Gypsum Tailings Dam 0.40 0.22 0.05 0.20
Greenhills Cougar 7 0.34 0.19 0.04 0.19
Helsinki Harbour 0.58 0.30 0.06 0.24
Hoedekenskerke Dyke 0.70 0.39 0.08 0.25
Jamuna Bridge Site 0.61 0.34 0.07 0.20
Lake Ackerman Road 0.55 0.31 0.07 0.25
Merriespriut Tailings 0.32 0.18 0.04 0.12
Mississippi R. Bank 0.52 0.29 0.06 0.27
Nerlerk Slide1 0.54 0.28 0.06 0.16
Nerlerk Slide2 0.33 0.18 0.04 0.16
Nerlerk Slide3 0.32 0.18 0.04 0.14
North Dyke 0.49 0.26 0.05 0.25
Quintette Marmot 0.39 0.21 0.05 0.19 Figure 2. Pre failure su /σv − qc1 relationships.
Sullivan Mine 0.34 0.19 0.04 0.19
Tar Island Dyke 0.41 0.23 0.05 0.20
Trondhiem Harbour 0.35 0.20 0.04 0.19
Vlietepolder, Zeeland 0.58 0.30 0.06 0.28
Western US Tailings 0.26 0.15 0.03 0.12
Wilheminapolder 0.39 0.22 0.05 0.12

The pre-failure su /σv , listed in Table 2, are plotted

in Figures 2 and 3 against qc1 and (N1 )60 , respectively.
Following correlations were developed:

(su /σv )TXC = 0.272 × qc1

0.223
(2)

(su /σv )SS = 0.189 × qc1

0.145
(3)

(su /σv )TXE = 0.033 × qc1

0.209
(4)

(su /σv )TXC = 0.123 × (N1 )0.239

60 (5)

(su /σ  v)SS = 0.116 × (N1 )0.134

60 (6)

(su /σv )TXE = 0.029 × (N1 )0.116

60 (7)

The r 2 values for Eqs. (2), (3) and (4) were 0.76,
0.72 and 0.71, respectively, and those for Eqs. (5), (6)
and (7) were 0.80, 0.77 and 0.77, respectively. Figure 3. Pre failure su /σv − (N1 )60 relationships.

929
4 FINITE ELEMENT MODELLING 4.3 Bakacka embankment, turkey
Nearly 100 m long section of the embankment fill,
In order to illustrate the efficacy of the proposed
including the northern lane of the highway, slipped
correlations, two embankments that underwent static
and spread into the valley on the north of the highway.
flow failure were analyzed using finite element pack-
The embankment was constructed nearly 40-years-
age PLAXIS software Version 8.2 (2002). The flow
ago using burrowed material from nearby cuts, and
failure case histories are described in the following
had an approximate height of 50 m with an inclina-
subsections.
tion of 2:3. The embankment failure occurred dur-
ing 12 November 1999 Du’zce earthquake due to the
intense near field ground motion. The undeformed and
4.1 Tailings Dam No. 2, South America deformed geometry of Tailings Dam No. 2 is shown
in Fig. 6. The more details of the failure are given in
This is an upstream constructed tailing dam con-
Bakir and Akis (2005).
structed in 1990s. The starter dam for this struc-
ture was constructed to a maximum height of about
10 m, of silty sand with fines contents in excess
of 30%. A filter facing on the upstream face of 4.4 Analysis procedure for flow failure cases
the starter dam, together with basal finger drains The step by step finite element procedure for the
below the starter dam (drain rock wrapped in geotex- analysis of flow failure cases are given below:
tile) were included given that the starter dam would
not function effectively as a toe drain without such • The flow failure of dams or embankments is ana-
drainage measures. The dam is founded primarily lyzed using the PLAXIS Version 8.2 software
on bedrock of relatively low hydraulic conductiv- in which deformations were calculated using the
ity, with some relatively fine-grained alluvium in Mohr’s Coulomb failure criteria.
the valley bottom. Overall, foundation conditions are • Firstly, drained analysis was carried out using
such that effective natural under-drainage is likely isotropic properties.
not present. The undeformed and deformed geom- • Soil properties were estimated using McGregor and
etry of Tailings Dam No. 2 is shown in Fig. 4. Duncan (1998).
The more details of the failure are given in Davies • The angle of the direction of vertical effective
(2002). stresses with respect to horizontal within the lique-
fiable layer were estimated from the analysis using
isotropic soil properties.
4.2 Tailings Dam No. 3, South America
This is an upstream constructed tailing dam con-
structed in 1990s. The dam is constructed across a
relatively steep-walled valley with bedrock exposed
on the valley slopes and alluvium in its base. The
starter dam was constructed of silty sand and gravel
(maximum fines content 30%), and included a zone
along its upstream face, and along its base, of Figure 5. Undeformed and deformed geometry of Dam
clean (fines content <5%) sand and gravel. The No. 3.
starter dam was constructed to a maximum height
of about 15 m. The undeformed and deformed
geometry of Tailings Dam No. 3 is shown in Fig. 5.
The more details of the failure are given in Davies
(2002).

Figure 4. Undeformed and deformed geometry of Dam Figure 6. Undeformed and deformed geometry of Bakacka
No. 2. Embankment.

930
• The liquefiable layer was divided into the various REFERENCES
vertical slices with respect to angle of the direc-
tion of vertical effective stresses with respect to Andresen, A. and Bjerrum, L. (1968). Slides in subaque-
horizontal. ousslopes in loose sand and silt, Norwegian Geotechnical
• Elastic modulus of soil layers were estimated by Institute Publication No. 81: 1–9.
assuming Poisson’s ratio 0.35 and modulus of rigid- Bakir, B.S. and Akis, E. (2005). Analysis of highway
embankment failure associated with the 1999 Duzce,
ity in between 10000 to 15000. Turkey earthquake. J. of Soil Dynamics and Earthquake
• The undrained soil properties within the liqufi- Engrg., 25: 251–260.
able soil layer were estimated below the water Casagrande, A. (1976). Liquefaction and cyclic defor-
table from propose anisotropic correlation and mation of sand: a critical review, Harvard Soil
above the water table the value of angle of inter- Mechanics Series No. 88, Harvard University
nal friction were consider in between 30 to 35 Cambridge, MA.
degree. Chillarige, A.V., Morgenstern, N.R., Robertson, P.K. and
• Using these above properties, the deformation of Christian, H.A. (1997). Seabed instability due to flow liq-
the following cases were estimated using the finite uefaction in the Fraser River delta, Canadian Geotech. J.,
34: 520–533.
element analysis. Dawson, R.F., Morgenstern, N.R. and Stokes, A.W. (1998).
Liquefaction flowslides in rocky mountain coal mine
waste dumps, Canadian Geotech. J., 35: 328–343.
4.5 Analysis procedure for seismic failure cases Davies, M.P. (1999). Peizocone technology for the geo-
The finite element procedure for seismic failure cases environmental characterization of mine tailings, Ph.D.
is same as flow failure cases except that the earthquake Thesis, University of British Columbia, Canada.
Davies, M.P., McRoberts, E.C. and Martin, T.E. (2002). A tail
time history records available at dam sites were used
of four upstream Tailings dam, Proc., Tailings Dams 2002,
in the analysis. ASDSO/USCOLD, Los Vegas.
Davies, M.P., McRoberts, E.C. and Martin, T.E. (2002).
Static liquefaction of tailings—Fundamentals and case
5 RESULTS histories, Proc., Tailings Dams 2002, ASDSO/USCOLD,
Los Vegas.
A framework proposed for undrained analysis of Fourie, A.B. and Papageoriou, G. (2001). Defining an appro-
priate steady state line for Merriespruit gold tailings,
dams or embankments, which considers inher- Canadian Geotech. J., 38: 695–706.
ently anisotropic soil behavior in an approximate Hazen, A. (1920). Hydraulic fill dams, Transactions of the
manner. In this paper, twenty-eight case histo- American Society of Civil Engineers, Paper No. 1458,
ries documenting undrained distress of dams or 1713–1821.
embankments were back analyzed using the pro- Hryciw, R.D., Vitton, S. and Thomann, T.G. (1990). Liq-
posed framework and based on the results of these uefaction and flow failure during seismic exploration,
back analyses, correlation was developed between J. Geotech. Engrg., 116: 1881–1899.
the pre liquefaction shear strength ratio, su /σv , Interactive Software Designs, Inc. (1994). XSTABL: An
and normalized penetration resistances, qc1 and integrated slope stability analysis program for personal
computers, Reference Manual.
(N1 )60 . Konard, J.M. and Watts, B.D. (1995). Undrained shear
The deformation from finite element analysis using strength for liquefaction flow failure analysis, Canadian
proposed pre failure relationships for anisotropic Geotech. J., 32: 783–794.
undrained analyses (Fig. 2) are shown in Table 3. The Koppejan, A.W., van Wamelen, B.M. and Weinberg, L.J.H.
results from finite element analyses using proposed (1948). Coastal flow slides in the Dutch province of Zee-
anisotropic undrained shear strength are showing land, Proc., 2nd Int. Conf. Of Soil Mechanics and Foun-
lesser vertical deformation than the observed vertical dation Engineering, June, 21–30, Netherlands, 89–96.
deformation. Rotterdam: Balkema.
Lucia, P.C. (1981). Review of experiences with the
flow failures of tailings dams and waste impound-
Table 3. Results of finite element analysis. ments, Ph.D. Thesis, University of California, Berkeley,
Calif.
Vertical displacement McGregor, J.A., and Duncan, J.M. (1998). Performance
and use of the standard penetration test in geotechnical
Anisotropic engineering practice, Report, Virginia Tech, Blacksburg,
Embankment Observed (FEM analysis) Virginia, USA.
Mittal, H.K. and Hardy, R.M. (1977). Geotechnical aspects
Tailings Dam No. 2 − 0.316 of a tar sand tailings dyke, Proc., Conf. On Geotechnical
Tailings Dam No. 3 − 1.16 Practice for disposal of solid waste materials, ASCE Spe-
Bakacka Embankment 7.95 7.85 cially Conf. Of the Geotechnical Engineering Division,
Vol. 1, 327–347.

931
Olson, S.M., Stark, T.D., Walton, W.H. and Castro, G. Sladen, J.A., D’Hollander, R.D., Krahn, J. and Mitchell, D.E.
(2000). Static liquefaction flow failure of the North Dike (1985). Back analysis of the Nerlerk berm liquefaction
of Wachusett Dam. J. of Geotech. and Geoenviro. Engrg., slides, Canadian Geotech. J., 22: 579–588.
126: 1184–1193. Vaid, Y.P., Chung, E.K.F. and Keurbis, R.H. (1990). Stress
Olson, S.M. and Stark, T.D. (2003). Yield strength ratio and path and steady state, Canadian Geotech. J., 27: 1–7.
liquefaction analysis of slopes and embankments. J. of Vaid, Y.P., Sivathayalan, S., Eliadorani and Uthayakumar, M.
Geotech. and Geoenviro. Engrg., 129: 727–737. (1996). Laboratory testing at University of British
PLAXIS software 2D- Version 8.2 (2002). Edited by Delft Columbia, CANLEX Report, University of Alberta,
University of Technology and PLAXIS b.v., Netherlands. Edmonton, Canada.
Singh, R. and Roy, D. (2006). Anisotropic undrained back Yoshimine, M., Ishihara, K. and Vargas, W. (1998). Effects
analysis of embankments, Proc., Int. Conf. Geoshang- of principal stress direction and intermediate principal
hai2006 ‘‘Advances in Earth Structures: Research to stress on undrained shear behavior of sand. Soils and
Practice’’, June 6–8, Shanghai, China, pp. 225–230. Foundations, 38: 179–188.
Senour, C. and Turnbull, W.J. (1948). A study of foundation Yoshimine, M., Robertson, P.K. and Wride, C.E. (1999).
failures at a river bank revetment. Proc., 2nd Int. Conf. On Undrained shear strength of clean sand to trigger flow
Soil Mechanics and Foundation Engrg., Vol. 7, 117–121. liquefaction. Canadian Geotech. J., 36: 891–906.

932
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Landslide model test to investigate the spreading range of debris according

to rainfall intensity

Y.S. Song, B.G. Chae & Y.C. Cho

Korea Institute of Geoscience & Mineral Resources (KIGAM), Daejeon, Korea

Y.S. Seo
Chung-Buk National University, Chungju, Korea

ABSTRACT: Landslide model experiments by considering rainfall intensity were performed to investigate and
predict the spreading range of debris. The model flume and the rainfall simulator were designed and manufactured
firstly, and a series of model experiments were performed. The model experiments were performed with changing
the rainfall intensity from 150 mm/hr to 250 mm/hr. In these experiments, the angle of slope inclination is 25◦ and
the relative density of slope soils is 35%. In order to measure the pore water pressure in a slope, the deformation
of slope surface, and the spreading range of debris during experiments, the instrumentation was installed in the
slope. As a result of instrumentation, the pore water pressure is increased rapidly at the time of landslides, and
the pore water pressure in a slope and the scale of landslide are increased with increasing the rainfall intensity.
In addition, the shape of spreading range of debris looks like a bulged pan, and the spreading range is increased
rapidly in its early stage and then increased gradually. The increasing velocity of spreading range is influenced
by the rainfall intensity, and the final spreading area after heavy rainfall depends on the rainfall intensity and
the rainfall duration time.

1 INTRODUCTION into debris flows as they move down slope. However,

the studies of the spreading range of debris by the
The annual rainfall of Korea ranges from 1,100 mm to landslides have not been conducted intensively.
1,400 mm, most of which concentrates in wet season In this study, therefore, landslide model experi-
beginning at June through September. The majority of ments according to the rainfall intensity are performed
landslides in Korea occurs in wet season and depends to investigate the spreading range of debris by the
highly on heavy rainfall (Park et al., 2006). The human landslides. To do this, a model flume and a rainfall
casualties due to natural disasters during 10 year in the simulator are designed and manufactured firstly, and
1990s were about 140, and 16% of them were injured monitoring instruments are installed to measure a pore
by the landslides. The most important factor causing water pressure in a slope and a deformation of the
landslides in Korea is the heavy rainfall. However, the slope surface during experiments. The model flume is
relationship between landslides and rainfall such as designed by consideration of the landslides character-
hourly rainfall, cumulative rainfall, and so on is not istics occurred in Korea as debris flows. The rainfall
yet proved clearly. simulator controls the rainfall intensity. A series of the
In case of Hong Kong, the analysis method con- model experiments according to the rainfall intensity
sidering the antecedent rainfall proposed by Lumb are performed to find out the occurrence characteris-
(1975) was used dominantly in the past, while the tics of landslides and predict the spreading range of
analysis method considering the rainfall intensity pro- debris.
posed by Brand (1985) has been used in recent years.
Kim et al. (1991) studied the occurrence mechanism
of landslides using a record of correlating rainfall with 2 LANDSLIDE MODEL EXPERIMENT
landslides in Korea. As the result of this study, land-
slides were influenced by both the cumulative rainfall 2.1 Model test equipment
and the rainfall intensity. In addition, Chae et al. The model test equipment is designed and manufac-
(2006) presented that the landslides are typical tran- tured to measure a spreading range of the run out dis-
sitional slides at the triggering position, and changed tance occurred by debris flows. The test equipment can

933
be divided into three parts: a model flume, a rainfall 2.2 Monitoring instruments
simulator, and monitoring systems, as shown schemat-
In order to measure the pore water pressure in a slope,
ically in Figure 1. The plate for measuring run out
the deformation of slope surface, and the spreading
distance is attached at the toe part of the model flume
range of debris due to heavy rainfall, an instrumenta-
to investigate the spreading process and range of run
tion system is designed and installed. The pore water
out distance.
pressure meter is PL1M model made by Senzors Co.,
The rainfall simulator consists of water sprinkling
USA with 2.5 cm diameter and 9 cm high. To measure
system, rainfall controller and water tank. It is devised
the pore water pressure in a slope, three pore water
to control the rainfall intensity by a computer. Also,
pressure meters are installed on the bottom at the slope
the monitoring systems consist of pore water pressure
part of model flume. The measured pore water pres-
meters, a data logger, marked pins and digital cameras.
sure is recorded automatically at 10 seconds’ intervals
The pore water pressure in a slope can be measured by
through the data logger. The data logger is Geolog-
the pore water pressure meter installed on the bottom
ger 515 made by Data Electronics Co., Australia, and
of the model flume.
has ten channels. The marked pins are installed on the
In Figure 1, the model flume can be divided into
center of slope surface at 20 cm intervals to measure
three parts: toe part, slope part and crest part. The
the deformation of slope. The deformation of slope is
length of each part is 0.5 m, 1.5 m and 0.3 m, respec-
measured by the photographing of marked pins using
tively, and the height is 0.5 m. The model flume is
digital and video cameras.
devised to control the slope angle ranged from 0◦ to
40◦ . The angle of slope can be controlled by use of the
height difference between crest part and toe part. The
2.3 Soil slope
model flume is made of steel frames and high strength
glass plates. The high strength glass plate is used on the The soils used in the experiment are Jumunjin standard
front side of the model flume to observe the deforma- sands, which are made and widely used for laboratory
tion of slope with the naked eye directly. The bottom tests in Korea. To make the homogeneous soil slope,
surface of the model flume is made with unevenness the standard sands are put in the flume and fallen free
in order to prevent the sliding at the interface between in the height of 75 cm using a funnel of 1 cm in dia-
soils and model flume. The run out distance plate is meter. The depth of the slope formed in this manner is
installed at the edge of toe part to observe the process 30 cm. The relative density of soil slope is 35%, and the
of spreading the debris triggered by landslides. internal friction angle obtained from the consolidated
The rainfall simulator composed of water sprinkling drained triaxial test (CD test) is 36.5◦ (Song, 2004).
device and a pedestal. The water sprinkling device is
made of PVC pipe, and its dimension is 3.1 m long
and 0.9 m wide. The water sprinkling device is con- 2.4 Experiment process
sisted of two rows of pipes, and nozzles having inner
diameters of 1.5 mm and 2.0 mm are attached to pipes The model experiment of debris flow is performed as
at equal spaces. The water sprinkling device is pro- follows;
duced for controlling the rainfall intensity ranged from 1. Control the angle of slope in model flume
100 mm/hr to 1,000 mm/hr. The pedestal having 2.5 m 2. Control the rainfall intensity of rainfall simulator
high and 1.8 m long is made of steel bar and can make 3. Install the pore water pressure meter at the bottom
a move and control the supporting height. of model flume
4. Make the model soil slope using Jumunjin standard
sand, and install the marked pins
5. Develop the debris flow due to rainfall sprinkling
6. Measure the deformation of slope surface, the pore
water pressure in a slope, and the spreading range
of debris

Table 1. Cases of the model test.

No. Rainfall intensity Slope Relative

(mm/hr) angle (◦ ) density (%)

SL-1 250 25 35
SL-2 200 25 35
Figure 1. Schematic diagram of the landslide model test SL-3 150 25 35
equipment.

934
 Through this experiment, the spreading range of
debris can be able to investigate according to the rain-
fall intensity. Table 1 shows the planning of model tests
according to rainfall intensities; three cases of model
test are performed with different rainfall intensities.
As shown in Table 1, the relative density of soil slope
is 35%, and the angle of slope is 25◦ . Three rainfall
intensities, 250 mm/hr, 200 mm/hr and 150 mm/hr, are
applied.
3 RESULTS AND DISCUSSION

3.1 Results of model experiments

To investigate the spreading range of debris caused by
Figure 4. Spreading range of debris on the plate induced by
landslides according to the rainfall intensity, the model
landslide.
experiments of landslides are performed with chang-
ing the rainfall intensities on the same condition of
soil slopes. The pore water pressure, the deformation 160
spreading width
of slope surface and the spreading range of debris are spreading length
measured during the model experiments.
120
Figures 2 to 5 show the results of model experi-
Spreading distance (cm)

ments in the case of SL-2 with the rainfall intensity

of 200 mm/hr. Figure 2 shows the variation of pore 80

0.24 40

0.20 triggering point p1 - lower part

0
Pore water pressure head (m)

0 2000 4000 6000 8000 10000

0.16 Elapsed time (sec)

0.12 Figure 5. Variation of spreading width and length according

p2 - middle part
to elapsed time.
0.08
p3 - upper part
0.04

0.00
0 2000 4000 6000 8000 10000
Elapsed time (sec) water pressures in the soil slope with elapsed time.
In Figure 2, p1, p2, and p3 indicate the pore water
Figure 2. Variation of pore water pressure on the slope pressures at the toe, middle, and top of slope parts,
according to elapsed time. respectively. The maximum pore water pressure is
occurred at the toe of slope part, while the minimum
pore water pressure is occurred at the top of slope part.
160 The pore water pressure is kept constant in its early
stage and then increased suddenly. Moriwaki et al.
Crack location from toe of slope (cm)

120 triggering point (2004) presented that the pore water pressure in slopes
is increased suddenly at the landslides occurrence
time. Therefore, the rapid increasing the pore water
80 pressure as shown in Figure 2 means the occurrence
of landslides.
40
Figure 3 shows the occurring position of tension
cracks on the slope surface triggered by rainfall. The
tension cracks are occurred firstly at the toe part of
0
0 2000 4000 6000 8000 10000
slope and then translated toward the crest part. That
Elapsed time (sec) is, the landslides are occurred firstly at the toe part,
and the scale of landslides is expanded gradually
Figure 3. Occurring position of tension crack on the slope when the cracks are translated toward the crest part.
surface according to elapsed time. It knows that the tension cracks on the slope surface are

935
 0.24 increased rapidly at the occurrence time of landslides.
The occurrence time of landslides is measured about
0.20
700 to 900 seconds. Also, the translational velocity of
Pore water pressure at p2 (m)

0.16
tension cracks toward the crest of slope is increased
250mm/hr
with increasing the rainfall intensity. It means that
0.12 the scale of landslides is increased with increasing the
200mm/hr
rainfall intensity.
0.08 Figure 7 shows the spreading area of debris caused
150mm/hr
by landslides according to the rainfall intensity. The
0.04
spreading area of debris is rapidly increased with inde-
0.00 pendent of rainfall intensity at the early period of
0 2000 4000 6000 8000 10000
Elapsed time (sec)
rainfall. But, the difference of spreading areas accord-
ing to the rainfall intensity is gradually increased after
Figure 6. Variation of pore water pressure at the middle the passage of a certain time, and the spreading area is
part of slope according to rainfall intensity. increased with increasing the rainfall intensity. There-
fore, the spreading area of debris is influenced by
not only the rainfall intensity but also the rainfall
duration time.
10000
250mm/hr
200mm/hr
8000 150mm/hr

4 CONCLUSIONS AND SUMMARY

Spreading area (cm2)

6000

In order to investigate the run out distance by the

4000 debris flow, the landslide model experiments accord-
ing to the rainfall intensity were performed. To do
2000 this, the model flume and the rainfall simulator were
designed and manufactured, and various monitoring
0 instruments were applied in the modeled slope. Based
0 2000 4000 6000 8000 10000 12000
Elapsed time (sec) on the results of model experiments, the occurrence
characteristics of landslides were analyzed and the run
Figure 7. Variation of spreading area of debris according out distance of debris was investigated. The following
to rainfall intensity. conclusions could be drawn;

1. The pore water pressure in the slope is increased

occurred when the pore water pressure are increased suddenly at the occurrence time of landslides, and
suddenly. the tension cracks on the slope surface are occurred
Figure 4 shows the photograph of the spreading at that time.
range of debris caused by landslides. This Figure indi- 2. The shape of spreading range of debris looks like
cates that the shape of the spreading range of debris bulged pan. In addition, the spreading length and
looks likes a bulged pan. Figures 5 show the spreading width is increased rapidly in its early stage and then
width and length of debris according to the elapsed increased gradually.
time. The increasing tendency of spreading width is 3. The pore water pressure in the slope and the scale of
similar to that of spreading length, while the spreading landslides are increased with increasing the rainfall
width is larger than the spreading length. The spread- intensity.
ing length and width is increased rapidly in the early 4. The increasing velocity of spreading range is influ-
stage and then increased gradually. enced by the rainfall intensity, and the final spread-
ing area after heavy rainfall depends on the rainfall
intensity and the rainfall duration time.
3.2 Comparison analysis according to the rainfall
intensity
Figure 6 shows the change of pore water pressure with
elapsed time according to various rainfall intensities. 5 ACKNOWLEDGMENTS
The data of pore water pressure are obtained from
p2 measured at the middle part of slope. The pore This research was supported by a grant (NEMA-06-
water pressure in the slope is increased with increas- NH-04) from the Natural Hazard Mitigation Research
ing the rainfall intensity, and the pore water pressure is Group, National Emergency Management Agency.

936
REFERENCES Lumb, P. 1975 Slope failure in Hong Kong. Journal of
Engineering Geology. 8: 31–65.
Brand, E.W. 1985 Predicting the performance of residual soil Moriwaki, H. Inokuchi, T. Hattanji, T. Sassa, K. Ochiai, H.
slopes. Proc. 11th Inter. Conf. on Soil Mech. and Found. & Wang, G. 2004 Failure processes in a full-scale land-
Eng. San Francisco. USA. 2541–2573. slide experiment using a rainfall simulator. Landslides. 1:
Chae, B.G. Kim, W.Y. Cho, Y.C. Kim, K.S. Lee, C.O. & 277–288.
Song, Y.S. 2006 Probabilistic prediction of debris flow on Park, D.K. Oh, J.R. Kim, T.H. & Park, J.H. 2006 Slope-
natural terrain. Proc. East Asia Landslides Symposium. related disasters and management system in Korea. Proc.
Daejeon. Korea. 144–153. East Asia Landslides Symposium. Daejeon. Korea. 35–46.
Kim, S.K. Hong, W.P. & Kim, Y.M. 1991 Prediction of Song, Y.S. 2004. Design methods of the slopes reinforced
rainfall-triggered landslides in Korea. Proc. 6th Inter. by earth retention system. Doctoral Thesis. Chung-Ang
Symp. on Landslides. Christchurch. New Zealand. 2: University. Korea.
989–994.

937
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Occurrence mechanism of rockslide at the time of the Chuetsu

earthquake in 2004 – A dynamic response analysis by using
a simple cyclic loading model

N. Tanaka & S. Abe

Okuyama Boring Co., Ltd., Yokote City, Japan

A. Wakai, H. Kawabata & M. Genda

Gunma University, Kiryu City, Japan

H.Yoshimatsu
SABO Technical Center (STC), Chiyoda-ku, Tokyo, Japan

ABSTRACT: Triggered by an earthquake that occurred in the Chuetsu area, Niigata Prefecture, Japan, in
2004, a large number of landslides occurred in the mountains where Tertiary sedimentary rock is distributed.
Three-dimensional dynamic response analysis (FEM) using a simple cyclic loading model (UW model) that takes
into account the shear strength and dynamic deformation characteristics of soil was conducted and revealed that
high horizontal acceleration occurred on the mountain tops and high shear stress occurred in the valleys. On the
basis of those analytical results, this paper discussed the mechanism of a large-scale rockslide that occurred at
Hitotsu-minesawa.

1 INTRODUCTION on the ridge topography following the Niigata Chuetsu

Earthquake in 2004, and reviews the relationship
There are a very small number of earthquake-induced between the behavior of mountain slopes and the
recurrent landslides or rockslides recorded so far in occurrence of landslides during earthquakes based
Japan except for some such events observed in parts on seismic response analysis using three-dimensional
of the Tohoku District. However, when the Niigata dynamic elasto-plastic FEM.
Chuetsu Earthquake (M = 6.8) hit in 2004, many
recurrent landslides and rockslides occurred in areas
where Tertiary formation is distributed. In particular, 2 OUTLINE OF THE LANDSLIDE
the Hitotsu-minesawa area in the mountains located
near the epicenter was subject to a large-scale rock- The Niigata Chuetsu Earthquake (hereinafter the
slide whose slip surface was 150 m deep. Abe et al. Chuetsu Earthquake) that occurred in the Chuetsu
(1997, 2005) pointed out that rockslides of this type area, Niigata Prefecture, in 2004 was a direct-hit
that occurred in conjugation with the Niigata Chuetsu earthquake that recorded a magnitude of 6.8 and a
Earthquake in 2004 and other earthquakes that have maximum acceleration of 1,500 gal (K-NET). The
occurred in the past in the area of the Tohoku Dis- rockslide investigated in our paper occurred in the
trict where Tertiary formation is distributed have many Hitotsu-minesawa area, located in the mountains about
common factors, one of which is that these land- 10 km northeast of the epicenter (Figure 1).
slides are prone to occur in ridge or questa topography. The landslide that occurred at Hitotsu-minesawa
Therefore, clarification of the occurrence mechanism transversely severed a ridge protruding west to west-
of Tertiary formation landslides triggered by inland southwest (Figures 2 and 3). The landslide scarp
earthquakes has the potential to play a pivotal role formed a collapse zone 20 to 30 m in width (b in
in hazard evaluation of future landslide disasters that Figures 2 and 3) that stretched from a ridge area
follow earthquakes. (a in Figures 2 and 3) to a paddy area until it reached
This paper focuses on the Hitotsu-minesawa land- an approximately 10 m high landslide scarp formed at
slide, which was a large-scale rockslide that occurred the northern edge of the landslide area (e in Figure 2).

939
Figure 1. Location map and location of analysis.

Figure 2. Overall view of Hitotsu-minesawa (from the side)

(Photo by Haraguchi, 2004). Figure 4. Crack occurring in a stream down the ridge (white
arrow).

The local geology is a sandstone-dominant strati-

graphic alternation of sandstone and mudstone from
the Oligocene-Neogene. The bottom layer seen in
the landslide area is presumed to be black mudstone,
which is the rock partly exposed at Hitotsu-minesawa,
the toe of the landslide area. The layer above it is silty
sandstone, the core of the moving mass of the land-
slide, which compresses thin muddy layers. Coarse
sandstone about 5 m in thickness, sandwiched by
silty sandstone, is distributed in the paddy field area.
At the topmost part of the landslide scarp is seen a
Figure 3. Entire view of Hitotsu-minesawa (from the front) more than 10 m-thick layer of black mudstone. Far-
(Photo by Haraguchi, 2004). ther above the landslide scarp is seen conglomerate
comprising coarse particles. The strike slope of the
layers is roughly N15◦ E and 0 to 5◦ W.
A crevice with an approximately 7 m level difference On both banks of a stream located at the border
is seen cracking the paddy field in the intermedi- of the paddy part and the ridge part of the landslide
ate part of the landslide area for about 80 m (c in zone (d in Figures 2 and 3) is exposed in series silty
Figures 2 and 3). A few landslide dams were also sandstone hard enough to defy crushing by hammer
formed at Hitotsu-minesawa, located at the toe of the impact. These rock masses have cracks in the north-
landslide area. The movement range of the landslide south direction at intervals of 1 to 10 m. These cracks
was about 500 m in width, about 350 m in length, and show no entry of tree roots or discoloration and are
about 150 m in the difference of elevation from the therefore judged to be fresh. Some of them are open
crown part to the front end (Figure 2). by about 5 cm (Figure 4).

940
 Hitotsu-minesawa, the toe of the landslide, has 3 SEISMIC RESPONSE ANALYSIS USING
some characteristic geologic features, including a THREE-DIMENSIONAL DYNAMIC
series of rock masses standing 3 m upright with slick- ELASTO-PLASTIC FEM
enside, a maximum 10 m high upheaval at the bottom
of the stream, and some landslide dams formed in
3.1 Elasto-plastic constitutive model
association with the upheaval (Figure 5). These are
and basic equation
formed by buckling or folding of the rock mass at
the toe following the rockslide. The slip surface has A number of reports have been issued on three-
a well-developed bedding and is presumed to linearly dimensional analyses of slope failure in the mountain-
continue from down the landslide scarp at the head to ous areas induced by seismic motion. For example,
Hitotsu-minesawa along the bedding plane due to the Mizuyama et al. (2002) reviewed the seismic motion
presence of thin layers of soft rocks such as black mud- distribution by focusing on the salient shapes of the
stone. If so, the thickness of the moving zone should slope, and Asano et al. (2006) studied the topographic
reach 100 to 150 m (Figure 6). effect of the mountain body by seismic motion and
its effect on collapse by assuming a three-dimensional
elastic and perfectly plastic model.
The elasto-plastic finite element method produces
results of a variety of characteristics depending on
the constitutive equation used in the method. For the
kind of rockslide that occurred at Hitotsu-minesawa,
the subject of this paper, it is necessary to use an
elasto-plastic constitutive model that can appropri-
ately express the kinetic properties of soil by seismic
motion.
With respect to such landslide events, the authors
have been engaged in a fundamental review based
on 3D dynamic elasto-plastic FEM to realize seis-
mic response analysis for the purpose of predicting
wide-area damage in the mountainous areas. To be
Figure 5. Rock mass with protruding slickensides at specific, our achievements include filter output of
Hitotsu-minesawa, the toe of the landslide. response acceleration waveforms (Wakai et al., 2005),
an attempt to identify wide-area ground physical prop-
erties (Tanaka et al., 2006), and development of
specifications for a program to conduct input and
output data with GIS and realize omission of compli-
cated pre- and post-processing unique to FEM (Wakai
et al., 2006). For a constitutive model used in the
3D dynamic elasto-plastic FEM, a simple cyclic load-
ing constitutive model (UW model) was used that
can consider both the dynamic deformation charac-
teristics of soil (that is, the relationship involving
shear elastic modulus G, shear strain τ , and damp-
ing ratio h-τ ) and shear strength (adhesive strength c
and internal frictional angle φ) by Wakai et al. (2004).
What characterizes this constitutive model is the use
of the hyperbolic stress strain relationship, which is
closer to the actual kinetic property of soil, as the
skeleton curve in combination with the unique hys-
teresis loop curve to solve the problem of excessive
damping ratio pointed out in a corrected HD model.
The hysteresis loop curve can express the dynamic
deformation characteristics that match the general h-τ
relationship of soil by appropriately providing two
parameters (n and bτ GO ). In our analysis, the h-τ rela-
tionship of Ishihara (1976) was used for reference.
Figure 6. Geologic longitudinal and transverse section (See As a result of the above formulation, we can realize
Photo 1 for A-A , B-B and C-C ). dynamic deformation characteristics that are closer to

941
the actual soil and reproduce the shear failure of the For the depth direction, element division was con-
ground to an excellent degree of precision, based on ducted assuming the ground to be composed of three
the Mohr-Coulomb standard. kinds of material. They are, in descending order from
The basic equation for the elasto-plastic constitutive near the top of the surface, the subsurface, weathered
model described above is expressed as follows. In a layer and fresh part. The fresh part alone is divided
dynamic elasto-plastic FEM, the following equation into two elements, and a total of four divisions were
of motion is solved by the time integration algorithm made in the depth direction. This analytical model is
based on Newmark’s β method. shown in Figure 7.
1 2 For the ground constants, the authors decided to
[M ] {ü} + [C] {u̇} + {P} = −[M ] Ü (1) choose the appropriate constants out of the ground
survey reports compiled by us for many construction
{P} is a nodal point vector equivalent to internal stress, projects conducted in the subject area. The ground
and {P} = [K]{u} in a linear elastic body. [M], [C] constants used for the analysis are tabulated in Table 1.
and [K] are respectively mass, attenuation, and initial
rigidity matrix. {u} and {U} are relative displacement
(at each position) and absolute displacement vector 3.3 Seismic wave input in the basement
(of the foundation bed), respectively. The attenua-
tion matrix was derived based on the assumption of Figure 1 shows the observation location of a seismic
Rayleigh attenuation. waveform used as a reference for our determination

[C] = α [M ] + β [K] (2)

This time two values, α and β, that lead to an damping

ratio of 3% at a frequency of 0.5 Hz and 5.0 Hz, are
assumed, which is 0.171 and 0.00174, respectively.
Refer to Wakai et al. (2004) for the details of
formulation based on an elasto-plastic constitutive
model.

3.2 Analytical model

The size of one finite element for the seismic response
analysis of the Hitotsu-minesawa landslide is set to
25 m × 25 m, and the area measuring 1.5 km × 1.5 km
in the neighborhood of the subject area, including
Hitotsu-minesawa, was analyzed. When the topogra-
phy of the subject area was converted to finite element
meshes, the raster elevation data obtained from GIS
were used. Figure 7. Finite element model for Hitotsu-minesawa.

Table 1. Ground constants used for analysis.

Ground Unit volumetric Young’s Poisson Adhesive Internal

model weight modulus ratio ν force frictional
General legend (depth in m) γ [kN/m3 ] E[kN/m2 ] C[kN/m2 ] angle ϕ[◦ ]

Sand silty layer 0 to 5 17.658 117720 0.40 9.81 30

5 to 30 18.639 1285110 0.35 98.10 35
30 to 80 19.620 5209110 0.30 981.00 40
80 to bottom layer 19.620 5209110 0.30 981.00 40
Wanazu layer 0 to 5 17.658 117720 0.40 9.81 30
and Shiraiwa layer 5 to 30 19.620 1353780 0.35 98.10 35
30 to 80 19.620 5209110 0.30 981.00 40
80 to bottom layer 19.620 5209110 0.30 981.00 40
Kawaguchi layer 0 to 5 19.620 127530 0.40 9.81 30
5 to 30 19.620 1353780 0.35 98.10 35
30 to 80 21.582 6925860 0.30 981.00 40
80 to bottom layer 21.582 6925860 0.30 981.00 40

942
of the input seismic motion (Takezawa in Yamakoshi 3.4 Analysis results
Village) and the location of the epicenter of the
The analysis results are put together in Figures 9
Chuetsu Earthquake. There are many ways of select-
through 11. They represent, respectively, the maxi-
ing the input seismic waveforms used for analy-
mum horizontal acceleration on the ground surface
sis (two components of NS and EW). In our case,
during the earthquake, residual horizontal displace-
before we determined the input seismic waveforms,
ment on the ground surface (both are the combined
we conducted FEM analysis of the area around
values of the NS and EW components) and the maxi-
Takezawa using the assumed input seismic motion
mum shear stress in the earthquake in the subsurface
and adjusted the input seismic waveforms on a
elements.
trial-and-error basis such that the response acceler-
Note that the response acceleration resulting from
ation, which is the analysis result, at Takezawa is
those dynamic elasto-plastic FEM analyses have
close to agreeing with the observed strong motion
been put through a low pass filter with a cut-off
record.
frequency of 10 Hz to remove the high-frequency
Figure 8 shows the NS and EW components of the
components.
input seismic waveforms (which were input to the loca-
As Figure 9 indicates, amplification of accelera-
tion 91 m above sea level) thus determined for our
tion is very much apparent at topographically sharp
analysis.
points or near the ridges. It is shown that accelera-
tion of over 1,000 gal particularly affected the ridge
part of the area where the landslide studied in this
2.0
paper occurred. Horizontal acceleration of over 1,000
1.0
Acceleration NS(m/s2)
gal also occurred at the ridge, shown in A-A in
Figure 2. High values of residual displacement are also
0.0

-1.0

-2.0
0 2 4 6 8 10
TIME(s)

2.0
Acceleration EW(m/s2)
1.0

0.0

-1.0

-2.0
0 2 4 6 8 10
TIME(s)

Figure 8. Input seismic waveforms (91 m above sea level).

Figure 10. Result of dynamic response analysis (residual
horizontal displacement).

Figure 9. Results of dynamic response analysis (maximum Figure 11. Results of dynamic response analysis (maximum
horizontal acceleration). shear stress).

943
 of the Hitotsu-minesawa rockslide is that the direction
of the ridge protrusion matches the direction of the
sloping direction of the ground and that the rockslide
occurred in the topography where the ridge, which has
a stream on either side of its foot, is open on both
sides. Thus, it appears that the ridge itself was severed
transversely when the land mass slipped. This judg-
ment is corroborated by the report of Abe et al. (2005)
that described the presence of a landslide very similar
to the Hitotsu-minesawa rockslide, or an earthquake-
triggered Tertiary formation landslide that occurred on
the ridge part of the mountain.
Figure 12. Schematic diagram representing the mechanism
of slope failure triggered by the earthquake.
We are currently making a stability evaluation
of mountain slopes by dynamic elasto-plastic FEM
analysis for the purpose of corroborating earthquake-
recorded at locations subjected to higher horizontal induced rockslides at the ridge from a viewpoint of
acceleration (Figure 9). On the other hand, according dynamics.
to Figure 11, a larger shear force of over 30 kN/m2
affected Hitotsu-minesawa and the mountain streams
at both ends of the A-A measuring line that pinches REFERENCES
the ridge.
Abe, S. & Takahashi, A. 1997. Landslide processes during
More specifically, a landslide occurs at a topog- earthquake in the Green Tuff area in the Tohoku district.
raphy where high horizontal acceleration occurs at Jour. Japan Soc. Eng. Geol Vol.38 No.5:265–279 (in
a ridge and high shear force affects the stream part, Japanese).
which corresponds to the dip slope of the ground Abe, S., Takahashi, A., Ogita, S. & Yoshimatsu, H. 2005.
(Figure 12). Earthquake induced landslides in the Tertiary rock area
Figure 12 Schematic diagram representing the in Japan. International Symposium Landslide hazard
mechanism of slope failure triggered by the earth- in Orogenic zone from the Himalaya to island arc in
quake. Asia:397–406.
Asano, S., Ochiai, H., Kurokawa, U., Okada, Y. 2006.
Topographic effects on earthquake motion that trigger
4 DISCUSSION AND SUMMARY landslides. Journal of the Japan Landslide Society Vol.42
No.6:457–466 (in Japanese).
Ishihara, K. 1976. The Foundation of Soil Dynamics:
The seismic response analysis using dynamic elasto- 196–206, Kajima Institute Publishing Co., Ltd.
plastic FEM revealed that high horizontal acceleration K-NET: http://www.kyoshin.bosai.go.jp/k-net/
and displacement occurred at salient topography such Mizuyama, T., Matsumura, K., Tsuchiya, S., Takahashi,
as at the head of the ridge and that, on the other hand, M. & Yang, W. 2002. Evaluation of topographic effects
the large shear force influenced the stream part. These on seismic failures of model slopes through dynamic
results agree very closely with the phenomenon by response analysis. Congress publication of Interprevent
which cracks in the north-south direction occurred at 2002 in the pacific rim Vol.1:59–66.
intervals of 1 to 10 m over the exposed bedrock of Tanaka, N., Wakai, A., Kanto, K. & Ito, H. 2006. Soil con-
stants for seismic evaluation of Mid-Niigata area and their
the stream, which is the border between the ridge part sensibility analysis. Proceeding of 45rd Conference of
and the paddy part (Figure 4). The phenomenon by Japan Landslide Society:293–296 (in Japanese).
which high horizontal acceleration occurs at a ridge Wakai, A. & Ugai, K. 2004. A simple constitutive model for
has previously been reported by Asano et al. (2006) the seismic analysis of slopes and its applications. Soils
and other researchers, but our analysis results reveal and Foundations Vol.44 No.4:83–97.
the additional finding that the mountain stream part is Wakkai, A., Watanabe, T., Kawabata, H. & Kanto, K. 2005.
subjected to a high shear force. In other words, when Fundamental study for development of mountains-area
the Chuetsu Earthquake occurred, its force severely seismic intensity prediction system used on FEM with-
shook the Hitotsu-minesawa ridge and at the same out advanced computer environment. Proc. Geo-Kanto
2005 JGS:145–148 (in Japanese).
time caused a high shear force at the stream at the Wakkai, A., Kawabata, H., Watanabe, T., Ahang, T., Kanto,
foot of the ridge, and eventually a landslide occurred K. & Tanaka, N. 2006. Mountains-area seismic intensity
that took the form of a dip slope-type slide in the prediction system based on FEM without excellent com-
direction of slope of the ground. This is how the puters. Proceeding of 45rd Conference of Japan Landslide
Hitotsu-minesawa rockslide occurred. A characteristic Society. 297–300. (in Japanese).

944
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Analysis for progressive failure of the Senise landslide based on Cosserat

continuum model

H.X. Tang
Department of Civil and Hydraulic Engineering, Dalian University of Technology, Dalian, China

ABSTRACT: Based on pressure-dependent elastoplastic Cosserat continuum model, progressive failure phe-
nomena of the Senise slope occurred in the excavation processes, characterized by strain localization due to
strain softening,are numerically simulated. Numerical results indicate the inability of classical continuum model
in simulating the whole failure progress, while the capability and performance of Cosserat continuum model in
keeping the well-posedness of the boundary value problems with strain softening behavior incorporated and in
completing simulation of the whole failure progress.
Keywords: Progressive failure, Cosserat continuum model, strain localization, strain softening.

1 INTRODUCTION Muhlhaus (1987, 1989), de Borst et al (1991, 1993),

Tejchman et al (1993, 1996), Steinmann (1994, 1999),
As strain softening constitutive behavior is incorpo- Manzari (2004), Li and Tang (2005).
rated into a computational model in the frame of de Borst (1991, 1993) formulated the von-Mises
classical plastic continuum theories, the initial and elastoplastic model within the framework of the
boundary value problem of the model will become Cosserat continuum, and further extended to the
ill-posed, resulting in pathologically mesh-dependent pressure dependent J2 flow model. Manzari (2004)
solutions. Furthermore, the energy dissipated at strain formulated a micropolar elastoplastic model for soils
softening is incorrectly predicted to be zero, and and employed a series of finite element analyses to
the finite element solutions converge to incorrect, demonstrate the use of a micropolar continuum in
physically meaningless ones as the element mesh overcoming the numerical difficulties encountered in
is refined. To correctly simulate strain localization application of finite element method in classical con-
phenomena characterized by occurrence and severe tinuum when non-associated yield criterion is incor-
development of the deformation localized into nar- porated in the constitutive model. But the strain soft-
row bands of intense irreversible straining caused by ening behavior, which may occur in the geotechnical
strain softening, it is required to introduce some type engineering, was not detected because of the four nod-
of regularization mechanism into the classical con- ded displacement-based quadrilateral isoparametric
tinuum model to preserve the well-posedness of the element interpolation approximation being used.
localization problem. Li and Tang (2005) formulated a consistent algo-
One of the radical approaches to introduce the reg- rithm of the pressure-dependent elastoplastic model
ularization mechanism into the model is to utilize in the framework of Cosserat continuum theory, i.e.
the Cosserat micro-polar continuum theory, in which the return mapping algorithm for the integration of the
high-order continuum structures are introduced. As rate constitutive equation and the closed form of the
two dimensional problems are concerned, a rotational consistent elastoplastic tangent modulus matrix.
degree of freedom with the rotation axis orthogonal to Based on Li and Tang’s pressure-dependent elasto-
the 2D plane, micro-curvatures as spatial derivatives plastic Cosserat continuum model, progressive fail-
of the rotational degree of freedom, coupled stresses ure phenomena of Senise landslide occurred in the
energetically conjugate to the micro-curvatures and the excavation processes in Italy, characterized by strain
material parameter defined as internal length scale are localization due to strain softening, are numerically
introduced in the Cosserat continuum. simulated. At the same time, the numerical results
Among the work, which utilize the Cosserat con- based on classical continuum model indicate its inabi-
tinuum model as the regularization approach to ana- lity in simulating the whole failure progress of the
lyze strain localization problems, are contributions of slope.

945
2 PRESSSURE-DEPENDENT ELASTOPLASTIC It is assumed that the strain vector ε is additively
COSSERAT CONTINUUM MODEL decomposed into both the elastic and the plastic parts,
i.e. εe and εp , and the elastic strain vector ε e is linearly
Each material point in the two dimensional Cosserat related to the stress vector,
continuum has three degrees-of-freedom, i.e. two
translational degrees-of-freedom ux , uy and one rota- σ = De εe (7)
tional degree-of-freedom ωz with the rotation axis
orthogonal to the two dimensional plane,
in which the elastic modulus matrix De for isotropic
u = [ux uy ωz ]T (1) media can be given in the form

Correspondingly, the strain and stress vectors are De =

defined as ⎡ ⎤
λ + 2G λ λ 0 0 0 0
ε = [εxx εyy εzz εxy εyx κzx lc κzy lc ] T
(2) ⎢ λ λ + 2G λ 0 0 0 0 ⎥
⎢ λ λ λ + 2G 0 0 0 0 ⎥
⎢ ⎥
⎢ 0 0 0 G + Gc G − Gc 0 0 ⎥
σ = [σxx σyy σzz σxy σyx mzx /lc mzy /lc ]T (3) ⎢ G − Gc G + Gc ⎥
⎣ 0 0 0 0 0
⎦
0 0 0 0 0 2G 0
where κzx , κzy are introduced as micro-curvatures in 0 0 0 0 0 0 2G
Cosserat theory, mzx , mzy are the couple stresses con- (8)
jugate to the curvatures κzx , κzy (Fig. 1), lc is defined
as the internal length scale.
The relation between strain components and dis- with the Lame constant λ = 2Gυ/(1–2υ), G and υ are
placement components and the equilibrium equations the shear modulus and Poisson’s ratio in the classical
can be written in matrix—vector forms as sense, while Gc is introduced as the Cosserat shear
modulus.
ε = Lu (4) In the framework of Cosserat continuum theory, a
consistent algorithm of the pressure-dependent elasto-
LT σ + f = 0 (5) plastic model, i.e. the return mapping algorithm for
the integration of the rate constitutive equation and
in which the operator matrix the closed form of the consistent elastoplastic tangent
modulus matrix can be formulated. The details of the
⎡ ∂ ∂ ⎤ model will not be recited here and the readers can
0 0 0 0 0 reference Li and Tang (2005)’s work.
⎢ ∂x ∂y ⎥
⎢ ∂ ∂ ⎥
⎢ ⎥
L =⎢ 0
T
0 0 0 0 ⎥
⎢ ∂y ∂x ⎥
⎣ ∂ ∂ ⎦ 3 INTRODUCTION OF THE SENISE
0 0 0 −1 1 lc lc LANDSLIDE
∂x ∂y
(6) According to Troncone (2005), on 26 July 1986 a
landslide of great dimensions occurred at Timpone
hill in Senise, which is a village located about 70 km
from Potenza, in southern Italy. As a result of this
yy event eight people died, and several buildings were
m yz destroyed or badly damaged. A reactivation of the
yx landslide occurred on 6 September 1986: it caused
further movement of the soil mass.
m xz This case study was analyzed by Troncone (2005),
y who provided exhaustive documentation of the land-
xx slide. After the first catastrophic event and before the
x xy reactivation of September 1986, a site investigation
consisting of boreholes, penetration tests and conven-
tional laboratory tests was performed. On the basis of
the available data, the main features of the landslide
mechanism were highlighted in the aforesaid studies:
Figure 1. Stress and couple-stress in a two-dimensional the location of the slip surface was detected, and the
Cosserat continuum. laboratory tests showed that the soils involved in the

946
Figure 2. Schematic geological section of Senise hill.

Figure 3. Clayey silt level within sand formation.

movement were characterized by a pronounced strain-

softening behavior. A summary of these investigated
aspects will be reported below. Figure 4. Experimental and simulated results of triaxial
A schematics geological section of Senise hill is tests: (a) sand; (b) clayey silt.
shown in Fig. 2. The subsoil is essentially constituted
of the Plio-Pleistocene Aliano formation, consisting
of yellowish sand with varying grain size, interbedded cemented sand formation. Both these materials exhi-
by clayey silt levels. These levels, dipping downslope, bited a pronounced strain-softening behavior during
have thickness ranging from some centimeters to sev- the laboratory tests.
eral decimeters and an inclination of about 18◦ , as can
be seen from Fig. 3. The sand is very dense, and char-
acterized by a significant degree of diagenesis due to 4 STABILITY ANALYSES FOR THE SENISE
calcareous cementation. Fig. 2 also reveals the pres- LANDSLIDE
ence of a basic formation consisting of blue-grey clay.
A water table was found at a depth of about 23 m from Very different interpretations of the landslide were
the ground level, at the slope toe. It therefore was well provided on the basis of stability analyses conducted
below the soil mass involved in the landslide occurred using the limit equilibrium method. Nevertheless, all
after a period of very scarce rain. the authors agree on the fact that the event was trig-
The strength properties of the soils under considera- gered, or at least influenced, by the deep excavations
tion were obtained from consolidated-drained triaxial carried out at the toe of the slope before the construc-
tests and direct shear tests. As shown in Fig. 4, one tion of some of the damaged buildings. The cuts were
peculiar aspect of these soils is their mechanical insta- essentially vertical, and were sustained by reinforced
bility, pointed out by a pronounced strain-softening concrete retaining walls.
behavior after the peak is reached. It revealed that Using the traditional limit equilibrium approach,
both the main and the secondary sliding surfaces were Troncone (2005) performed the stability analysis of the
located in thin clayey silt layers present in the sand Senise slope. By considering a potential slip surface
formation by exploration, ant the kinematics of the close to that which occurred (Fig. 5), the safety fac-
Senise landslide was essentially a translational slid- tor SF, computed using Sarma’s method in conjunction
ing: the main failure surface developed largely along with the peak strength parameters, is 1.73. By contrast,
a thin clayey silt layer interbedded within a slightly using the residual strength, the slope is unstable with a

947
safety factor of about 0.6. This should imply that a pro- Based on Li and Tang’s pressure-dependent elasto-
gressive failure really occurred, with the operational plastic Cosserat continuum model, progressive failure
strength along the sliding surface varying between the phenomena of Senise landslide occurred in the excava-
peak and the residual values. tion processes in Italy, characterized by strain localiza-
The application of the finite element method in tion due to strain softening are numerically simulated.
capturing the development of the yield zone and the At the same time, the numerical results based on
effect of progressive failure in strain-softening soils classical continuum model indicate its inability in
has been illustrated by many authors. One of these simulating the whole failure progress of the slope.
studies employed an elasto-plastic constitutive model According to Troncone’s analyses, the soil para-
associated with a law of Mohr-Coulomb type, in which meters used in the analyses are reported in Table 1.
the softening behavior of the material is accounted for The soil dilatancy angle has been assumed to be nil,
by the progressive reduction of the strength param- implying a non-associated flow rule with zero volume
p
eters with the accumulated deviatoric plastic strains. change during yield. kshear is set to zero to account for
The solution based on this approach may be affected the brittle behavior experienced by the soils during the
by a lack of convergence and may depend strongly on laboratory tests. The parameters Gc and lc , which are
the mesh adopted, in terms of both the size and the related to Cosserat continuum model, are chosen to
orientation of the elements. Therefore an analysis of correspond with the general rule. In order to take into
this type requires particular care, especially when the account the strain-softening behavior of the soil, the
location and propagation of the shear zones cannot be stability analysis of the slope has been conducted using
predetermined on the basis of some evidence. the FEM. Fig. 6 shows the mesh adopted in the cal-
In order to overcome the numerical difficulties culation, and it consists of the eight nodded quadratic
of the classical continuum when the strain-softening quadrangle elements.
behavior incorporated, Troncone (2005) used the The initial stress state within the slope, before
elasto-viscoplastic model with strain-softening to ana- the excavation, has been reproduced by progressively
lyze the stability of the slope. It was found that increasing the gravity acceleration up to the value
this model was very effective in capturing the failure of 9.81 m/s2 , under the assumption that all the soils
process that occurred. exhibit elastic perfectly plastic behavior with the

Figure 6. Mesh and boundary conditions used in finite

Figure 5. Failure surface considered in stability analysis element analysis.
performed using limit equilibrium method.

Table 1. Soil parameters used in the analysis.

p∗
γ E cp ϕp cr ϕr ψ kshear r
kshear Gc lc

kN/m3 kPa ν kPa ◦ kPa ◦ ◦ % % kPa m

Yellow sand 20 70000 0.25 37 43 0 35 0 0 4 14000 0.15

Clayey silt level 20 25000 0.25 15 30 0 12 0 0 4 5000 0.15
Blue-grey clay 20 70000 0.25 150 31 150 31 0 0 0 14000 0.15

∗ p
‘‘p’’ means peak, ‘‘r’’ means residual. For example, kshear means the peak shear strain.

948
Drucker-Prager failure criterion. At the end of the
gravity loading, the corresponding displacements and
strains have been reset to zero. A linear-elastic element
simulating the retaining wall has been inserted at the
slope toe, without affecting the stress state within the
slope. The excavation process is simulated by remov-
ing the elements at the same level in 9 increments
from the excavation zone of the initial finite element
mesh, and Mana’s method (1981) is used to calculate
the excavation loading. (a)
At first, analysis has been performed assum-
ing an elasto-plastic strain-softening model with the
Drucker-Prager plastic law for all the soils involved.
Fig. 7 shows the evolution of the effective plastic
strain occurring during the last excavation steps. It
can be seen that the plastic strain field develops with a
clear localization within the clayey silt level. However,
the whole development of the failure surface can not
be completed. As the excavation process continues,
the classical finite element numerical solution faces (b)
significant difficulties with the number of negative
eigenvalues in the system stiffness matrix increasing
and the numerical calculation can not be carried out
any more. Therefore the numerical analysis based on
this elasto-plastic model with strain softening is not
fully effective in capturing the failure process that
occurs in the present study.

(c)

Figure 8. Effective plastic strain distribution in the slope

during the excavation process given by Cosserat continuum
model: (a) after the removal of the 7th layer; (b) after the
removal of the 8th layer; (c) after the removal of the 9th layer.

The calculations have been repeated using the

(a) elasto-plastic Cosserat continuum model with strain-
softening. Fig. 8 shows the evolution of the more
significant effective plastic strain occurring during the
last excavation steps. As can be seen that the plastic
strain are concentrated in the clayey silt level, and
as the excavation process continues they propagate
up the slope extending also to the sand formation
and, finally cause the collapse of the slope when
the excavation is completed. Fig. 9 shows the cor-
responding deformation configuration at the end of
excavation. The position of the failure surface at both
(b) the top and the toe of the slope corresponds with that
of Troncone’s analyses and that actually observed.
Figure 7. Effective plastic strain distribution in the slope Therefore the numerical analysis based on Cosserat
during the excavation process given by classical continuum elasto-plastic model with strain softening is effec-
model: (a) after the removal of the 7th layer; (b) after the tive in capturing the failure process that occurs in the
removal of the 8th layer. present study.

949
 REFERENCES

de Borst R., Sluys L.J., 1991. Localization in a Cosserat

continuum under static and dynamic loading conditions.
Comput. Methods Appl. Mech. Engrg., 90: 805–827.
de Borst R., 1991. Simulation of srtain localization:a reap-
praisal of the cosserat continuum. Engineering computa-
tions, (8): 317–332.
de Borst R., 1993. A generalization of J2-flow theory for
polar continue. Computer methods in applied mechanics
and engineering, 103: 347–362.
Li Xikui, Tang Hongxiang, 2005. A consistent return
mapping algorithm for pressure-dependent elastoplastic
Cosserat continua and modeling of strain localization.
Computers & Structures, 83 (1): 1–10.
Figure 9. Deformation configuration of the slope after the
Mana A.I., Clough G.W., 1981. Prediction of movements for
removal of the 9th layer with Cosserat continuum.
braced cuts in clay. Journal of Geotechnical Engineering
Division, 107: 759–778.
Manzari M.T., 2004. Application of micropolar plasticity to
5 CONCLUSIONS post failure analysis in geomechanics, Int. J. Numer. Anal.
Meth. Geomech., 28: 1011–1032.
Based on pressure-dependent elastoplastic Cosserat Muhlhaus H.B., Vardoulakis I., 1987. The thickness of
continuum model, progressive failure phenomena of shear band in granular materials. Geotechnique, 37 (3):
the Senise slope occurred in the excavation pro- 271–283.
cesses, characterized by strain localization due to Muhlhaus H.B., 1989. Application of Cosserat theory in
numerical solution of limit load problems. Ing. Arch., 59:
a pronounced strain softening, are numerically sim-
124–137.
ulated. Numerical results indicate the inability of Steinmann Paul, 1994. A micropolar theory of finite defor-
classical continuum model in simulating the whole mation and finite rotation multiplicative elastoplasticity.
failure progress, while the capability and performance Int. J. Solids Structures, 31: 1063–1084.
of Cosserat continuum model in keeping the well- Steinmann Paul, 1999. Theory and numerics of ductile
posedness of the boundary value problems with strain micropolar elastoplastic damage. Int. J. Numer. Methods
softening behavior incorporated and in completing Eng., 38: 583–606.
simulation of the whole failure progress. Tejchman J., Wu W., 1993. Numerical study on patterning
of shear bands in a Cosserat continuum. Acta Mechanica,
99: 61–74.
Tejchman J., Bauer E., 1996. Numerical simulation of
ACKNOWLEDGEMENTS shear band formation with a polar hypoplastic constitutive
model. Computers and geotechnics, 19: 221–244.
The author is pleased to acknowledge the support of Troncone A., 2005. Numerical analysis of a landslide in
this work by the National Natural Science Foundation soils with strain-softening behaviour. Geotechnique, 55
of China through contract/grant numbers 50278012, (8): 585–596.
10672033 and the National Key Basic Research and
Development Program (973 Program) through con-
tract number 2002CB412709.

950
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Large-scale deformation of the La Clapière landslide and its numerical

modelling (Saint-Etienne de Tinée, France)

E. Tric, T. Lebourg & H. Jomard

University of Nice, Sophia Antipolis, Géosciences Azur, UMR 6526, Valbonne, France

ABSTRACT: The large-scale deformation of high mountain slopes finds its origin in many phenomena from
which time-constants are very different. Gravitational effect, tectonics forces, water infiltration are generally the
principal causes. However, it is always very difficult to distinguish which cause is dominating and which are their
effect respective. The numerical approach offers the possibility of testing some of these causes, in particular the
gravitational effect. Then, a two-dimensional numerical experimentation with ADELI code was carried out to
determine the effect of gravitational force on mechanical behaviour of the ‘‘la Clapière’’ area. The results show
that gravitational instability is possible for values of cohesion and angle of internal friction respectively lower
than à 2,00 ± 0,01 Mpa and 27,69◦ ± 0,14◦ . These values are compatible with the geomechanical parameters
proposed by Gunzburger (2001) from measurements directly done on the site. The numerical results show also
a good agreement between the calculated deformation, the actual morphology of the site and geophysical data
obtained by resistivity investigation. The deformation leads to destabilisation of the massif by a regressive
evolution of the landslide from the bottom up to 1800 m, which is actually the top of Ma Clapière landslide.
This progression of the deformation concerns only a depth of around 150 ± 50 m, which could be correlated to
the sliding surface like suggested by resistivity investigations (Lebourg et al. 2005; Jomard et al. 2007).

1 INTRODUCTION 2001, Merrien-Soukatchoff et al. 2001, Eberhardt

et al. 2004).
For the study of the deformation or the stability In this contribution, we present an application of
analysis of high mountain slopes, knowledge of the the numerical approach to the ‘‘La Clapière’’ Land-
large-scale mechanical properties is important. Differ- slide (Alpes-Maritimes, France). Indeed, if numerous
ent approaches exist to estimate those parameters like studies have been realised on this site (hydrologi-
seismic measurements (Eberhart-Philips et al., 1995; cal, geologic, tectonic, topographic, . . .) (Follacci,
Brueckl and Parotudis, 2001), geotechnical mea- 1987, Ivaldi et al. 1991, Compagnon et al. 1997;
surements like ‘Schmidt hammer tests’ (Bieniawski, Guglielmi et al. 2000, 2002, 2003, Gunzburger &
1989; Gunzburger, 2001; Gunzburger and Merrien- Laumonier, 2002), very few numerical studies have
Soukatchoff, 2002) or mechanical test in laboratory been done (Quenot, 2000; Merrien-Soukatchoff et al.
(Lebourg et al. 2004). However, these different obser- 2001, Gunzburger et al. 2002, Cappa et al. 2004) and,
vational analysis have not the potential to describe the to our knowledge, no study has been devoted on the
processes involved on the generation and development localisation of the deformation and its temporal evo-
of a creeping rock mass out of an originally compact lution as a function of the geomechanical parameters
rock. A promising supplement could be the geome- and their relation with the gravitational destabilisation
chanical modelling of the structures of the creeping of the slope.
or sliding rock masses. Thus, geomechanical parame-
ters may be determined by fitting the geomechanical
model to the structures as determined from geophy- 2 THE ‘‘LA CLAPIÈRE’’ LANDSLIDE
sical exploration (seismic or electrical). This approach
rapidly evolving due to the increasing computer power ‘‘La Clapière’’ landslide is a large unstable slope
allows also to define stress-strain state of the slope located in the South-eastern part of France, in the Alps,
with use of the relevant non-linear mathematical mod- about 80 km North of Nice city. This landslide which
els of the soils and rigorous consideration of all acting mobilizes a huge volume (55 × 106 m3 ) of metamor-
forces as well as the variation of mechanical properties phic slope in the Mercantour massif (Follacci et al.
with depth (Savage et al. 2000; Brueckl & Parotudis, 1988) is developed on the left bank of Tinee Valley and

951
 Ténibres
valley Rabuons
valley

Tinée valley
(altitude: 1100 m)

Figure 1. The ‘‘La Clapière’’ landslide in 1998 (photo from

CETE de Nice).

affect a slope that culminate at 3000 meters, between

1100 and 1800 meters of altitude (figure 1).
It is bordered on its northwestern side by the Teni-
bres river and to its southeastern sibe by the Rabuons
river, flowing into the Tinée river. A large rupture has
been identified since the beginning of the century: in
1936 the wrenching at the top of the landslide was Figure 2. Geomorphological context of La Clapière land-
already quite visible. In the seventies, the movements slide from Cappa et al. 2004.
became more continuous and a monitoring of the site
is working since 1982 (Follaci, 1987, 1988). For a few
years, a distance monitoring has been combined with
Laumonier, 2002, Delteil et al. 2003) or gravitational
other types of investigations: hydrogeological stud-
(Follacci, 1987, 1988). However, the numerical
ies (Compagnon et al. 1997; Guglielmi et al. 2000,
approach done by Merrien-Soukatchoff et al. (2001)
Cappa et al. 2004), remote sensing (Casson et al. 2003)
shows that gravitational toppling proposed by Follacci
and subsurface geophysical investigations (Lebourg
(1987) led to be reconsidered. The ‘‘La Clapière’’
et al. 2003, 2005, Jomard et al. 2007a, 2007b).
slope itself is affected by a lot of tectonic discontinu-
Different types of geometric, mechanic or hydrome-
ities. A complete description can be read in Guglielmi
chanic models have been derived from these data sets
et al. (2000), Gunzburger and Laumonier (2002) and
(Merrien-Soukatchoff et al. 2001; Cappa et al. 2004)
Cappa et al. (2004).
but none of them visualise the morphology of the main
Cappa et al. (2004) have tested numerically the
slip surface geometry. However, Jomard et al. (2007b)
influence of the location and the amount of water
have recently investigated by electrical method the ver-
infiltration on the hydromechanical behaviour of La
tical structure of the upper part of the NE lobe of the La
Clapière slope. Their results showed that most desta-
Clapière landslide. Their results shows a very strong
bilizing area is located in the middle part of the slope.
decreasing of the resistivity to a depth of approxi-
However, due to the internal complexity of the slope
mately 90 m which could be associated to the main
hydrogeology, especially the existing perched aquifer
shear surface of this landslide for this altitude.
in the decompression toppled zone directly connected
In this area, the basement unit is composed
to the landslide, the most dangerous infiltration inflow
of migmatitic paragneisses (Anelle Formation) and
is not necessarily that on the landslide area.
orthogneisses (Iglière Formation) bearing a strong
Hercynian foliation (Figure 2). This foliation is nor-
mally oriented 115◦ E, 70◦ NE but, in the la Clapière
area, it is progressively rotated to a subhorizontal atti- 3 MECHANICAL MODELLING
tude by a structure called the Clapière fold (Fabri and
Cappa, 2001; Gunzburger and Laumonier, 2002). The The 2D numerical modelling of the ‘‘La Clapière’’
axis of the fold is 120◦ E, 15◦ NW; the axial plane is landslide has been done with the computer code
140◦ E, 40◦ SW. The Clapière landslide occurs in the ADELI developed by R. Hassani and J. Chery (Has-
upper/short limb of the fold (Figure 2). sani, 1994, Hassani et al. 1997). This code is based
Two origins of the La Clapière fold are actu- on a finite element method for space discretization
ally proposed. It is either tectonic (Gunzburger and associated with a dynamic relaxation method for time

952
discretization. This method is described in Hassani great number of fractures of various scales and ori-
et al. 1997. It leads to a set of nonlinear equations: gins, which play a major role in rock mass strength
deterioration and deformability increase. To take them
M ü = Fint (u, u̇, t) + Fext (u, t) + Fc (u, u̇, t) into account, a method consist to homogenise the rock
mass by using the RMR methodology (Bieniawski,
where the vectors Fint and Fext are the internal and 1989; Gunzburger & Merrien-Soukatchoff, 2002) to
external nodal forces, Fc is the vector of contact reac- bring to an Equivalent Continuous Medium (ECM).
tion, u, u̇ and ü are the vectors of nodal displacements, The mechanical parameters of the ECM obtained
nodal velocities, and nodal accelerations, respectively. by this approach for ‘‘La Clapière’’ site are: Young’s
M is a fictitious mass matrix. The quasi-static solu- Modulus, E = 6,4 Mpa , cohesion, c = 210 kPa, angle
tion is reached when the inertial regularising term of internal friction, φ = 29◦ , Poisson’s ratio, v = 0,
Mü is negligible compared to the external forces. The 3 and density, ρ = 2400 kg/m3 (Gunzburger, 2001;
numerical divergence of the result means that rupture Gunzburger & Merrien-Soukatchoff, 2002):
is appeared and involves a destabilising of the solid These parameters are considered in our study as
mass. the reference parameters. In our numerical approach,
For each numerical experiment presented in this we model the mechanical behaviour using an elasto-
work, several mesh have been tested (6000, 10000, plastic pressure dependent law for which the failure
15000 and 20000 elements) and 5 × 104 to 1 × 106 criterion is the Drucker-Prager one (Desai & Siriwar-
times steps were used for a total duration of 100 or dane, 1984). We consider in this study a homogeneous
10000 years. The lengths of the time steps are then 52 rock mass.
minutes to 17 hours for the duration of 100 years, and From the five parameters given previously, two have
87 hours to 72 days for the duration of 10000 years. been tested systematically: the cohesion (c) and the
friction angle (φ). The others were kept constant.
Three tests have been performed.
3.1 Geometry and boundary conditions
The geometry used is two-dimensional. The initial – Test 1: We tested the couple (c, φ) from the refer-
state topography is taken from the most ancient avail- ence couple (210 kPa, 29◦ ) in order to determine the
able map (dated from 1933) in order to eliminate as critical couple for which the destabilisation appears.
far as possible recent slope deformations. The veloc- – Test 2: The cohesion (c) is constant and we research
ity boundary conditions are shown in figure 3. The the critical value of φ.
gravity is the only force. – Test 3: The friction angle is constant and we research
the critical value of c.

3.2 Mechanical parameters

The ‘‘La Clapière’’ slope is constituted of migmatitic 4 NUMERICAL RESULTS
paragneisses and migmatitic orthogneisses. Those
rocks bear a strong Hercyninan foliation, and are The results of the test 1 are given in table 1. They are
therefore highly anisotropic. However, numerical identical whatever the used mesh and the applied dura-
modelling has first been undertaken with isotropic rhe- tion. They show a critical couple (c, φ) equal to 95,0%
ological properties. The rocks are also affected by a of the reference couple, i.e. (1,995.105 Pa, 27,55◦ )
from which the numerical solution does not converge.
All the converged solutions, which correspond con-
3000 m
sequently to a static solution, are characterised by three
plastic deformation zones. The first one is located at
u, v the base of the slope up to 1400 meters of altitude and
g

u=0 3000 m
v Table 1. Numerical results for the test 1.
Rock mass

u=0
c (kPa), φ % of the reference Numerical
v couple solution

210, 29.0◦ 100 Convergence

204, 28.1◦ 97 Convergence
201, 27.8◦ 96 Convergence
Figure 3. Geometry of the model and boundary conditions. 199, 27.5◦ 95 Divergence
u et v are the horizontal and vertical component of the velocity 197, 27.2◦ 94 Divergence
respectively. Only the gravity (g) is present.

953
on a depth about 200 m. The second one is located Table 2. Numerical results obtained for different tests.
along the slope between 1400 m and 1800 m of alti-
tude and below the surface (about 100 m). The third Numerical solution
zone is at the top of the slop (figure 4a).
Test Constant Variable Convergence Divergence
When the couple (c, φ) is equal or lower to the
critical couple, the divergence of the numerical exper- 1 (c, φ) ≥ 96 % ≤ 95%
iment is due to a strong plastic deformation located 2 c φ ≥ 96 % ≤ 95 %
at the bottom of the slope (figure 4b). This plastic 3 c φ ≥ 65 % ≤ 62 %
deformation leads to the destabilisation of the mas-
sif by a regressive evolution of the landslide from
the bottom to the top of the slope. This deforma-
tion propagates up to 1800 m, which is actually the
top of the ‘‘La Clapière’’ landslide. This progression The whole of the results, obtained in the different
of the deformation concerns only a depth of around tests, is resumed in the Table 2. We observe that the
100–130 m. It is interesting to note that the two alti- instability seems to be mainly controlled by the fric-
tudes given previously (1400, 1800 m) correspond to a tion angle. Indeed, when the cohesion is kept constant
change of the slope in the topography of 1933. Do these and equal to 210 kPa, the divergence of the calcula-
results suggest that the distribution of the deformation tion is observed from a value of internal friction angle
is controlled by these changes of the slope? The ques- identical to those obtained when the variable is the
tion remains open and deserves to be to study more couple (c, φ). Nevertheless, whatever the used vari-
attentively. able, the observed deformation and its temporal and
spatial evolution is the same.

5 CONCLUSION

Different mechanical numerical models have been

performed to better locate the deformation in the rock
mass and determine critical geomechanical parame-
ters from which the landslide occurs in the particular
case of ‘‘La Clapière’’. Although the solid mass
was regarded as mechanically homogeneous, the first
conclusions of this study are encouraging and are:

– A destabilisation of the rock mass is possible with

the geomechanical parameters obtained by Y. Gun-
zburger (2001). Thus, the combination of these
mechanical parameters, the topography and the
gravity force is completely sufficient to cause a
destabilisation of the solid mass.
– Our numerical experiments allow to locate the
deformation and his evolution. We observe system-
atically a good correlation between the calculated
deformation and the current morphology of the
slope. Indeed, in both cases, the disorders are for an
altitude ranging between 1100 m and 1800 m and
with a maximum deformation located at the bottom
of slope.
– The depth of the deformation is between 80 and
200 m according to the altitude, which is in good
Figure 4. Cumulated plastic deformation obtained for two
agreement with the geophysical results obtained by
simulations: (a) the couple (c, φ) corresponds to the reference electrical prospecting in previous studies (Lebourg
couple (100%). The solution is converged, (b) The couple et al. 2005; Jomard et al. 2007)
corresponds to 94 % of the reference couple. The solution is – Our results suggest an important role of the changes
not converged. of the slope in the evolution of the deformation.

954
REFERENCES Gunzburger Y. 2001. Apports de l’analyse de la fractura-
tion et de la modélisation numérique à l’étude du versant
Bieniawski Z.T. 1989. Engineering Rock Mass Classifica- instable de La Clapière (Saint-Etienne-de-Tinée, Alpes-
tions, J. Wiley. Maritimes’’), Mémoire de DEA PAE3S, LAEGO, Ecole
Brueckl E. & Parotidis M. 2001. Estimation of large-scale des Mines de Nancy.
mechanical properties of a large landslide on the basis Gunzburger Y., Merrien-Soukatchoff V. & Guglielmi, Y.
of seismic results. Int. J. Rock Mech. Sci. & Mining Sci. Mechanical influence of the last deglaciation on the initia-
1365–1369. tion of the ‘‘La Clapière’’ slope instability (southern french
Cappa F., Guglielmi Y., Soukatchoff V. M., Mudry J., alps), 5th European conference on numerical methods in
Bertrand, C. & Charmoille, A. 2004. Hydromechanical geotechnical engineering. Paris, France, 4–6.
modelling of a large moving rock slope inferred from Gunzburger Y. & Laumonier B. 2002. origine tectonique du
slope levelling coupled to spring long-term hydrochemical pli supportant le glissement de terrain de La Clapière
monitoring: example of the La Clapière landslide (Nord-Ouest du massif de l’Argentera—Mercantour,
(Southern Alps, France). Journal of Hydrology 291: Alpes du Sud, France) d’après l’analyse de la fracturation.
67–90. C. R. Géosciences 334: 415–422.
Casson B., Delacourt C., Baratoux D. & Allemand P. 2003. Hassani R. 1994. Modélisation numérique de la déforma-
Seventeen years of the La Clapière landslide evolu- tion des systèmes geologiques’’ Thèse de l’Université de
tion analysed from ortho-rectified aerials photographs. Montpellier II.
Engineering Geology, 68, 123–139. Hassani R., Jongmans D. & Chéry J. 1997. Study of plate
Compagnon F., Guglielmi, Y., Mudry, J. & Follacci, J-P. & deformation and stress in subduction processes using two-
Ivaldi, J-P. 1997. Chemical and isotopic natural tracing of dimensional numerical models. J.G.R. 102: 17, 951–17,
seepage waters in an important landslide: example from 965.
La Clapière landslide (Alpes-Maritimes, France). C. R. Ivaldi J-P., Guardia P. Follacci J-P. & Terramorsi, S. 1991.
Acad. Sci. Paris, no. 325: 565–570. Plis de couverture en échelon et failles de second ordre
Delteil J., Stephan J-F. & Attal, M. 2003. Control of permian associés à un décrochement dextre de socle sur le bord
and triassic faults on Alpine basement deformation in the nord-ouest de l’Argentera (Alpes-Maritimes, France).
Argentera massif (external southern French Alps). Bull. C.R. Acad. Sci. Paris, série II, 313: 361–368.
Soc. Géol. Fr. (174) 55–70. Jomard H., Lebourg T. & Tric E. 2007a. Identification of the
Desai C.S. & Siriwardane H.J. 1984. Constitutive laws gravitational discontinuity in weathered gneiss by geo-
for engineering materials, with emphasis on geologic physical survey: La Clapière Landslide (France). Applied
materials, 457 pp., Prentice-Hall, Englewood Cliffs N.J. Geophysics, 62: 47–57.
Eberhardt-Philips D. Stanley W.D. Rodriguez B.D. & Lut- Jomard J., Lebourg T., Binet S., Tric E. & Hernandez M.
ter W. J. 1995. Surface seismic and electrical methods to 2007b. Characterisation of an internal slope movement
detect fluids related to faulting. JGR 100: 12, 919–12, structure by hydrogeophysical surveying. Terra Nova, 19
936. (1): 48–57.
Fabri O. & Cappa F. 2001. Apport de l’analyse structurale à la Lebourg T., Tric E., Guglielmi Y., Cappa F., Charmoille A.
compréhension de la dégradation du glissement de la La & Bouissou S. 2003. Geophysical survey to understand
Clapière, Massif du Mercantour, Alpes Maritimes, S. Spé. failure mechanisms involved on Deep Seated Landslides.
Soc., Géol. Fr., 13–14. EGS, Nice.
Follacci J-P. 1987. Les mouvements du versant de la Clapière Lebourg T., Binet S., Tric E., Jomard H. & El Bedoui, S. 2005.
à Saint-Etienne de Tinée (Alpes—Maritimes). Bull. Liais. Geophysical survey to estimate the 3D sliding surface and
Labo. P. et Cha. 150–151. the 4D evolution of the water pressure on part of a Deep
Follacci J.P., Guardia, P. & Ivaldi J.P. 1988. Le glissement de la Seated Landslide. Terra Nova, 17, 399–406.
Clapière (Alpes Maritimes, France) dans son cadre géody- Merrien-Soukatchoff V., Quenot Y. & Guglielmi Y. 2001.
namique. Compte rendu du 5ème symposium international Modélisation par éléments distincts du phénomène de
sur les glissements de terrain, p1323–1327. fauchage gravitaire. Application au glissement de La
Guglielmi Y. & Bertrand C. Compagnon F. Follacci, J.P. Clapière (Saint-Etienne de Tinée, Alpes-Maritimes)
Mudry, J. 2000. Acquisition of water chemistry in a mobile 95/96, 133–142.
fissured basement massif: its role in the hydrogeological Quenot X. 2000. Etude du glissement de La Clapière. Modéli-
knowledge of the La Clapière Landslide (Mercantour mas- sation du phénomène de rupture, Mémoire de DEA, DEA
sif, southern alps, France). Journal of hydrology, 229, pp PAES3S, Ecole Doctorale PROMEMA, Institut national
138–148. polytechnique de Lorraine, Nancy.
Guglielmi Y. et al. Hydrogeochemistry: an investigation Savage W.Z., Baum R.L., Morrissey M.M. & Arndt B.P.
tool to evaluate infitration into large moving rock masses 2000. Finite-element analysis of the Woodway landslide,
(Case study of the La Clapière and Séchilienne alpine Washington, U. S. Geological Survey Bulletin 2180, 1–9.
landslides). Bull Eng Geol Env, vol. 61, pp 311–324,
2002.
Guglielmi., Y. Vengeon, J.M. Bertrand, C. Mudry, J. Follacci
J.P. & Giraud, A. 2003. Hydrogeochemistry: an inves-
tigation tool to evaluate infiltration into large moving
rock masses (case study of La Clapière and Séchilienne
alpine landslides). Bulletin of Engineering Geology and
the Environment, no. 61, pp.311–324.

955
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

A novel complex valued neuron model for landslide assessment

Kanishka Tyagi, Vaibhav Jindal & Vipunj Kumar

Govind Ballabh Pant University of Agriculture and Technology, Pantnagar, India

ABSTRACT: Landslides are widely spread natural calamity on Earth and is a general term used to describe
the down-slope of soil, rock and organic material under the influence of gravity. In this paper we propose a novel
method for landslide assessment with the help of input variables that have direct physical significance, using
a Novel Neuron Model Approach that works in complex domain and is an extension of the self-look up table
approach based Counterpropagation neuron model. Many researchers have investigated landslide assessment
problem using conventional neuron models but they suffer from there own limitations of extensive computation
and dependency on the pre conditioning of the data being used. The novel part of the paper is that the computation
land investigation data is done by converting it into its equivalent complex number form. Architecture of
CVCPN, Data validation, Processing of land slide investigation data, and the result of network with various
land investigation data have been discussed.

1 INTRODUCTION already in use real valued counterpropagation network

has been used as a means of assessing potential slope
There are a number of potentially damaging natu- failure hazards on the southern slopes of ENE trend-
ral phenomenon, which are termed ‘geo-hazards’ and ing ridge near Malpa village situated 43 Km. from
slope instability is one of them. Engineers often speak Dharchula in Himalayan region of India.
of landslides when the mechanics of movement may Till now complex valued neural network was
be any one of the manifestations of slope instability. actively used in backpropagation algorithm with com-
In geological term landslide is defined as the displace- plex weights and complex valued neuron-activation
ment of rock masses and the soil along the sloping functions, though this network suffers with many lim-
surface from their normal position under the influence itations like it is slow with a large set of data, lack
of when structural stability of rock mass is disturbed. of bounded and analytic complex nonlinear activa-
This downward movement of the consolidated and tion functions in complex plane (Silverman, 1975).
unconsolidated rock matter is termed as landslide. Several approaches have been suggested to process
Excess quantity of water contents makes the rock weak the complex data using backpropagation algorithm
and more mobile. The intensity and magnitude of these (Yadav, 2005). Due to strong power of generalization
movements varies greatly depending upon the extent and simplicity in calculations, Counterpropagation
and amount of slope, and the type of soil. network has an advantage over backpropagation net-
These are frequently responsible for considerable work. Here the complex number theory is successfully
losses and they are subsequently considered among applied on forward only Counterpropagation network.
the most serious geologic hazards, which plague many Many algorithms have been developed in recent years
parts of the world. Landslides are among the major that work on neural computations techniques with
hydro-geological hazards that affect large parts of complex values (Hirose, 2003). Many of them have
India, especially the Himalayas, the North-eastern hill applied complex values on backpropagation algorithm
ranges, the Western Ghats, the Nilgiris, the Eastern using multilayer or multiplicative neurons (Yadav,
Ghats and the Vindhyas. 2005). Computation of various Geological parameters
Landslide hazard assessment of existing landslide by using complex numbers theory and complex val-
areas, in which the most probable patterns and mech- ued Counterpropagation network is another plausible
anisms of future movements are the recurrence of past approach, which we have explored in this paper.
patterns and mechanisms, is often seen as requiring the Section 2 describes the types of failure, Section 3
recognition of those patterns. Areas of different hazard describes the network architecture of the Complex-
intensities are mapped according to past activities. Valued Counterpropagation Network (CVCPN).
In this study a novel complex valued counterprop- Section 4 shows the evaluation methodology used to
agation neural networks which is an extension of process the geological parameters. Section 5 describes

957
the experimental results obtained from the network as can be as high as more than 1 m/sec if the rate is
well as the possible future work. slow (less than 10 m/day) it can be termed as mudslide.
The mudslide generally takes place on slope steeper
than 5◦ , but a mudflow can take place on a gentler
slope. In mudflow the water/soil ration is high and the
2 TYPES OF FAILURE
rate of motion is usually frequent and sudden. It can
take place after heavy downpour or after collapse of
The type of movement of the slope forming mate-
steep hill slope.
rial depends upon the inherent characteristics of the
Slide: Earth slide involves loose and consolidated
material, geometry of slope and on factors causing
rock matters however there is more significant and
movements. These can be broadly described below:
rapid movement. Sometimes, it is even be very sudden
Rock fall: When the movement of slope material
and violent causing enormous damage.
is vertically downward due to gravity and not along
Slump slides: These are rotational failures with lit-
any plane it is termed a fall. It can be a rock fall if
tle vertical displacement in thick overburden material
a rock block gets detached from the main rock mass
or in weathered and soft rocks. Slump failures are rare
due to presence of at least three sets of adversely ori-
in hard rocks. However, in certain special setup guided
ented joints and here is undercutting of slope by some
by joints, step like movement planes are developed as
eroding agency. On steep rocky slope at high altitude
a result of stress releases and these cracks are called
having vertical or steeply dipping open joints, whose
glide cracks. The head region of a slump is character-
strike is parallel or subparallel to slope large block of
ized by steep escarpments, where tension cracks are
rock may fill due to frost wedging as a result of freez-
developed which are generally concentric and paral-
ing of water filling the gaping joints. These filled rock
lel to the main scrap. The foot region of the slump
masses may ultimately get detached from the main
slide is marked by a zone of tension and uplift, and a
rock mass and come down as rock fall.
bulge is formed in this region. Due to different rate of
Debris fall: When debris is accumulated on a slope
movement in the central portion and due to resistance
it stands for sometime but when more material is added
created by the firm strata on the periphery, en-echelon
or if the shear resistance between the debris/rock con-
cracks are developed in the periphery region on the
tacts is reduced due to uplift pressure caused by rain
sides of the slide.
the debris starts moving along this place. Some debris
Block slide: These are very common in loose
may pose perpetual danger to the roads.
deposits overlying glacial till or where very low dip-
Creep: A creep is slow (0.01–1 m/year) movement
ping rock beds have bedding clay seams which are
in over burden material or weathered rock on gen-
plastic in nature and the overlying rock is dissected
tle slope without any defined surface of failure. The
by steeply dipping joints which dissect the rock into
movement is parallel to slope and the reduced with
blocks. The movement in this case is transnational.
depth. The creep is caused when the slope material is
Planar failure and Wedge failure: In jointed rocks
cohesive and there is variation in diurnal temperature.
the stability of slopes depends upon the nature of
The rate of movement depends on rainfall, and the
joints, their altitude and shear friction along them. If
motion is assisted by cyclic swelling and shrinkage.
the joints are adversely oriented they offer planes for
Mudflow: As against creep the rate of movement in
the movement of overlying rocks. If it is failure along a
case of low is very fast. In mudflow the movement
single plane, it is called a ‘planar failure’. If however,
two or more set of joints form a wedge and the dip of
the line of contact of two planes is towards the valley
at angle more than the angle of shear friction along
those planes it may result as wedge failure.
Internal causes: This includes such causes which
tend to reduce the shearing strength of the rock. The
steeping of the slope, water content of the mass, its
mineralogical composition and structural features are
important factors which will define the stability or
otherwise of a given landmass.

3 NETWORK ARCHITECTURE

Figure 1. Showing rock slide due to failure at foliation In its simplest version, Counterpropagation network
planes. (CPN) is able to perform vector to vector mapping

958
 After updating they become wr (t + 1) and wi (t + 1)
respectively. Then,

wr (t + 1) = wr (t) + αt (xr − wr (t)) (1)

wi (t + 1) = wi (t) + αt (xi − wi (t)) (2)

The network is trained until there is no significant

change in the updated weight and the old weight i.e.

wr (t + 1) = wr (t) (3)

wi (t + 1) = wi (t) (4)

In that case the real and imaginary value of the input

vector is copied on to its corresponding weight vector.
When this situation arises the training is said to be
Figure 2. Complex valued counterpropagation network. completed and this is the stopping condition for the
network. The weight vectors have now settled near
the centroid of each cluster. These clusters are different
for real and imaginary parts.
similar to heteroassociative memory networks. The Euclidean distance method or the nearest neighbor
complex numbers are treated as vector real and imagi- method is found to be most suitable. The reason for
nary parts are given equal significance thus following a taking the minimum distance (and not the maximum)
real-valued approach [Tyagi, 2007]. The weights used is that the distance which is minimum is closest to
in the network are also taken in complex form. a particular cluster; we then update the weights in
The CVCPN functions in the recall mode as a near- such a manner so as to bring it more closely to the
est match look-up table. The difference from the usual cluster thus moving that value to the centroid of the
table look-up is that the weight vectors are obtained by cluster.
the training algorithm, rather than in an adhoc man- We know that if x1 and x2 are two complex
ner. The input vector xr + ixi finds the weight vector vectors, then the Euclidean distance between them is
wrm + iwim that is its closest match among k vectors given by
available in the hidden layer. Weights vrm + ivim which 
are fanning out from the winning mth Kohonen’s neu- | x1 − x2 | = (x1r − x2r )2 + (xii − x2i )2 (5)
ron are updated. The output given by the network is the
statistical averages of the similar kinds of input given
If n be the total input neurons and k be the total
in the training period.
number of hidden layer neurons, then let x1 , x2 . . . xn
The input layer contain n cells for distributing
are the complex input given in the Instar layer. wr11 ,
training vector x, Kohonen layer with k cells pro-
duces output neth. This neth being a real number is wr12 . . . wr1k are the real and wi11 , wi12 . . . wi1k be the
compared with other k-1 cells and minimum neth is imaginary components of the weights connecting the
declare as winner. A fan-in structure, from all input input x1 and the k’th neurons of Kohonen layer,
cells is thus created to the winning cell in Koho- similarly other weights are connecting input neuron
nen layer. During training stage each complex value and the hidden layer. The Euclidean distance between
is fed into Kohonen network for the self organizing x1 and w11 then is:
classification. The classification is done separately
on real and imaginary part of input values by a pre- x1 = xr1 + ixi1 (6)
defined algorithm which has been investigated by
Tyagi et al. w11 = wr11 + iwi11 (7)

x1 − w11 = (xr1 − wr11 ) + i(xi1 − wi11 ) (8)

3.1 Training of input vector in Kohonen layer

We assume that wr (t) and wi (t) are the real and imag-
|x1 − w11 | = (x1r − w1r )2 + (x1i − w1i )2 (9)
inary components of Kohonen layer weight vector.

959
 Equation 9 is the Euclidean distance between any in cluster 2) is into this cluster is same is the case with
two points in a complex plane. For a particular set cluster 4 where vector C wrongly directing towards
of values of input vectors x1 , x2 , x3 . . . xu . . . xn we cluster 4. This is a practical situation and is the cause
calculate the following real valued quantity. of error and misclassification. This can be removed
by either with more training or proper selection of α
n 
 and β. The clusters are spread out and are not confined
neth1 = [(xur − wu1
r 2
) + (xui − wu1
i 2
) ] (10) in a single quadrant.
u=1

3.2 Training the outstar layer

Similarly
Once the training in Kohonen layer is completed and
n 
 all the input vector have been clustered (both real and
neth2 = [(xur − wu1
r 2
) + (xui − wu1
i 2
) ] (11) imaginary) the vector x now makes a fan out connec-
u=1 tion from Kohonen layer to the Grossberg layer and
only weights in the Grossberg layer are updated which
Generalizing it we get; are connected to the winning neuron. After the train-
ing stage, outputs of the CVCPN are separate in real
n 
 and imaginary parts. They are again combined after
nethk = [(xur − wu1
r 2
) + (xui − wu1
i 2
) ] (12) the output layer to give the complex output. If each
u=1 input vector in a cluster maps to a different output
vector, then the outstar learning procedure will enable
The minimum of all neth (neth1 , neth2 . . . nethk ) is the outstar to reproduce the average of those output
calculated and is considered as winner for that set of vectors when any member of the class is presented to
values of x. the inputs of the CVCPN. A stuck vector problem is
Figure 3 gives a pictorial representation of how the seen if we generate the weights in the range of [−1 1].
clustering process takes place after the complete train- Also to avoid the condition of orthogonality of ran-
ing of network. For the sake of simplicity we have domly generated weight vectors we have used weight
shown clustering only for the real part of input vec- vectors in the range of [0 1].
tors, similar clustering takes place for the imaginary
part also. Here A, B . . . E are the input vectors (real
numbers) and WrA , WrB . . . WrE are the real compo- 4 EVALUATION METHODOLOGY
nents of weight vectors. It is only after the training
that these weight vectors are oriented at the center of 4.1 Study area
the circle. This is what we mean when we say that the
weights are adjusted to the centroid of the cluster. The For assessing the various soil features, southern slope
bold line from origin to each cluster is the weight vec- of ENE trending ridge near Malpa village (Darchula,
tor and is the mean of all values of the vectors coming Uttaranchal, India) in the main range of the central
in that circle. Observe that cluster 3 is not properly crystalline at the latitude of 30◦ 01 55 and longitude
clustered as orientation of vector D (which should be 80◦ 45 07 is considered. The rock fall lies between
two major tectonic plates, i.e. Vaikrita thrust and trans
Himadri fault, both trending NW-SE. The slided mass
consisted of massive quartzite interbedded with thin
bands of garnet bearing scricite schist dipping homo-
clinally at a step angle of 60◦ –70◦ . The failure surface
of the north east slided portion shows a series of paral-
lel foliation planes. Near vertical joints, perpendicular
to the foliation plane, constitute the face of the haz-
ardous area. The precipitation in the preceding days
increase the total water content in the open fractures
and decrease the cohesion and shearing resistance in
the rocks, thus leading to rock slide. The falling rocks
reached the base elevation of approx. 2200 m from
elevation of 3000 m with a high velocity.

4.2 Data processing and conditioning

In land slide assessment the data is typically a series
Figure 3. Clustering process in CVCPN. of analog values that has to be kept within certain

960
Figure 4. Encircled area is under consideration.
Figure 5. Scattering diagram for aspect.

ranges. The data considered are the various geological

parameters (soil aspect, slope, precipitation, elevation)
that contribute towards Land sliding. Complex Val-
ued Counterpropagation Network (CVCPN) is trained (in meter)
to recognize optimal values for such data, and make
necessary adjustments to control valves, whenever
necessary. The data used is for the upper Himalayan
region of Malpa. The data conditioning is an important
step while using CVCPN. To convert the data in com-
plex number form we first normalize the data (which
are real numbers) and then arrange them in such way
that the complex part is the next geological parame-
ter from the real part in that complex number. i.e. if
a + ib is the complex number then b is the next geo-
logical parameter after a. Similarly other parameters
is also converted on this form and then feed into the
complex CPN.
After training is done in this fashion and during the
testing phase the data is again converted into the real Figure 6. Scattering diagram for elevation.
number form and de-normalized.

5 RESULTS AND DISCUSSIONS (in mm)

For each problem, Intel (M) (Celeron Mobile),

1.5 GHz, CPU with 256 MB RAM is used for sim-
ulation work with MATLAB® 7.0 as the simulation
software. Training and testing data were normalized
wherever necessary. It is observed that the proposed
model exhibits a more efficient learning in each case.
It is due to the reason that in the Kohonen Layer
of CVCPN unsupervised and in Grossberg Layer
supervised learning is taking place. It is actually the
combination of two independent layer of different
learning rule that makes up the CPN and this basic
structure is preserved in our proposed model also.
Figure shows 4 different graphs obtained between
the actual and the computed value of various geologi-
cal parameters. Figure 7. Scattering diagram for precipitation.

961
 CONCLUSIONS
(in degrees)
By reducing the number of testing set and still main-
taining the required accuracy, reducing the number
of hidden layers thus reducing the complexity from
hardware point of view and circuit realization are
some of the future areas which are still needed to be
explored with forecasting using complex valued neural
networks.

REFERENCES

Akira Hirose. 2003. Complex-Valued Neural Network: The-

ories and Applications, World Scientific.
Figure 8. Scattering diagram for slope. Cai F. & Ugai K. 2004. Numerical Analysis of Rain-
fall Effects on Slope Stability, International Journal of
Geomechanics ASCE.
Table 1. Error values for various geological parameters. Punamia B.C, Jain & Jain., 1998. Soil mechanics and
foundations Laxmi publication.
S. no. Geological parameter Error Silverman H. 1975. Complex Variables, Houghton, Newark,
USA.
1 Aspect 0.0079 Subramanya K. 1984. Engineering Hydrology, Tata McGraw
2 Elevation 0.0487 Hills Publication.
3 Precipitation 0.0684 Thandaveswara B.S. & Sajikumar N. 2000, Classification of
4 Slope 0.0361 River Basins using Artificial Neural Network, Journal of
hydrologic engineering: 290–297.
Tyagi K., Mishra D. & Kalra P.K. 2007. A Novel Complex
Valued Counterpropagation Network, IEEE Proceed-
ings on Computational Intelligence and Data Mining,
From the various graphs obtained it is observed pp. 81–87.
that the minimum scattering is obtained with the Yadav A., Mishra D., Ray S. & Kalra P.K. 2005. Represen-
aspect, thereby concluding that it is the most important tation of Complex-Valued Neural Networks: A Real-
geological parameter in landslide assessment. Total Valued Approach, IEEE Proceedings on intelligent sens-
number of hidden neurons are taken as 80. Training ing and information processing, pp. 331–335.
and testing time is approximately 40 seconds, num-
ber of parameters are 28. Learning rate α varies from
0.9–0.1 and β varies from 0.1 to 0.5. The variations
of these learning parameters are exponential in nature
contrary to linear change that is more common while
using CPN network with real numbers.

962
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Prediction of slope behavior for deforming railway embankments

V.V. Vinogradov, Yu.K. Frolovsky, A. Al. Zaitsev & I.V. Ivanchenko

Track and Track Facilities Department, Moscow State University of Railway Engineering, Moscow, Russia

ABSTRACT: The results of the experimental and theoretical researches of slope behavior for deforming
railway embankments are presented in the paper. The criteria of similarity for modeling of the serviceability of
railway embankments on geotechnical centrifuges under the vibro-dynamical loads from a train are given. In the
paper the basic ways of reinforcement and stabilization of potentially-dangerous and deforming embankments
of railways are shown, requirements to which are developed on the basis of results of centrifugal modeling of
slope stability of a range of embankments, both on real prototypes, and on the basis of development of the
generalized parameters of embankments according to the statistical data of the railway network of the Russian
Federation. One of the most perspective and economically attractive technical decisions is use for these purposes
MSE walls, gabion walls which have received the big distribution abroad and the beginnings 90th years began
to be applied widely in Russia in civil and transport construction. For this purpose modern methods of physical
modeling and full scale monitoring on centrifuge are used. Among the last the method of centrifuge modeling
which allows providing similarity of the processes occurring on small models of soil constructions, to changes
of conditions of natural prototypes of a subgrade has the special importance.

1 THEORETICAL INVESTIGATION The tasks of research of slope stability of the sub-

OF SLOPE BEHAVIOR grade (as well as other geotechnical constructions) can
be divided into two groups: the problems which aim at
The most significant deformations of separate objects studying the physical nature of processes of infringe-
of the railway subgrade are failures of stability of their ment of stability of slopes (they can be carried out to
slopes. check the hypotheses or various analytical decisions)
Any object of the subgrade, including an intensively and tasks of applied character when a definite con-
operated railway embankment together with external, struction with a definite goal to estimate the slope
natural influences and internal loads at studying stabil- stability is studied.
ity of slopes, is regarded as an open dynamic system. At centrifugal modeling all physical nature of the
If to treat a parameter of stability of such an embank- process of infringement of stability of slopes remains
ment in a general way as interdependence between the unchanged and even in most cases the geometrical sim-
factors holding a slope and the factors, leading to its ilarity of blocks of displacement is provided. Therefore
failures, it can be written down as: the experimental estimation on the models of slope sta-
bility can supplement or replace in very many cases
Ks (t) = f [τo (t), γd (t), U1 (t), U2 (t), U3 (t), t] (1) theoretical methods of estimation. Appropriate theo-
retical preparation is necessary for providing of the
where Ks = coefficient of stability; t0 = generalized experiment. The development of criteria equations of
shear strength of soil; γd = unit weight of dry soil; the process and their analysis, as well as the subse-
U1 (t) = the function of external loads (external loads quent analysis of the results of experiments with the
on an embankment from constructions, the weight help of received criteria.
of the track upper structure and the loads from the In analysis of slope stability of railway subgrade in
train); U2 (t) = the function of natural factors (atmo- Russia the well-known method of calculation of sta-
spheric precipitations, temperature, wind, flooding); bility by G.M. Shakhunyants is usually used according
U3 (t) = the function of internal factors (pressure from to which the stability of the slope is estimated by the
the weight of the soil, moisture, pore pressure). coefficient of stability.

963
 The necessary criterion of similarity for modeling of dynamic pressure, elastic fluctuations and vibra-
spontaneous deformation of a slope in a period of time, tions; the size and an orientation of these influences
using the differential equation of deformability of the is various in space and time;
kind is calculated as: – Owing to fluctuations and vibrations arise iner-
tial and dispersive forces, the size and direction of
∂z ∂ 2z ∂z which is also various in space and time;
= R 2 + μy + μz (2)
∂t ∂y ∂y – Influence of dynamic pressure on the ground layer
is marked in the appearance of irreversible and con-
where y and z = coordinates of points of a slope; vertible processes; these processes are stipulated at
R = the factor of deformability dependent on vis- vibrations by infringement of structural reactions
cous properties of the ground; m = the coefficient of between particles and their units (for embankments
proportionality between the speed and the parameter these reactions are restored to a certain extent after
of displacement of a particle of the soil. deliberate condensation during construction); there
Having lowered the intermediate calculations, we is also a transformation of the cohesive water and
obtain the general criterion of similarity of the process clearing of immobilized water that reduces the resis-
in question as: tance of the soil to a shift, as physically cohesive
water possesses ability to resist shifting, free pore
π Rt 2 water, on the contrary, possesses ability of greas-
π= = idem (3) ing; at vibrations and significant elastic settlements
L2
there is also the general time dilatation of the soil
where L = general linear size; t = time. and as a consequence of it:
The received criteria can be used for planning exper- – Decrease of strength characteristics; all these phe-
iments and the analysis of results of modeling and their nomena and processes are shown non-uniformly in
transferring on natural objects. space and time;
For practical use the following criteria of similarity – Duration of continuous influence of dynamic pres-
are recommended: sure is of great importance, as thus irreversible pro-
cesses start to prevail above convertible in view
Pu s · P · W of decrease in bearing capacity of the ground and
π1 = = idem;
L 3 · γd · τ0 resistance to shift that leads to its more intensive de-
  formability; in this connection it is necessary to take
W W
π2 = Pus
2
· P2 · I II into account the character and duration of the actions
L 3 · γd · τ0 L 3 · γd · τ0 of dynamic influence on the model of the basic plat-
= idem; form of subgrade, their cinematic communication
with the model of the prototype.
2
Pus P·W – Influence of dynamic external factors should be
π3 = · = idem
L4 (γd · ϕ · c)n (γd · ϕ · c)c such as to answer natural conditions of influence
at which many factors, previously not taken into
R·t
π4 = = idem (4) account or partially used by static calculations
L2 reveal themselves, factors: various microstructures
of the soil, their natural physical properties—
where Pus and P = external factors from the weight of
friction, cohesion, shear strength, molecular links
the upper structure and the influence of a train; W =
of particles of soil and water, chemical cementation
moisture of soil; γd = density of dry soil; t0 = the
of structure, etc.;
generalized resistance to the shift of the soil.
– External dynamic influence is transferred to the
In the second equation (4) index I designate char-
subgrade through a ballast layer, the length of this
acteristics of a subgrade, index II for foundation. In
transfer is such, that the problem can be considered
the third equation of the system (4) index n character-
as a 2D problem.
izes properties of draining ground of ballast deepening
and loops of an exploited embankment, index c =
All the specifics should be whenever possible taken
cohesive soil embankments.
into account at the creation of dynamic model of train
The distinctive feature of modeling of stability of
pressure and definition of operating conditions with it.
slopes of embankments of railways is the necessity to
At modeling of dynamic influence of train pressure
take dynamic influence of train pressure into account.
on the subgrade three variants of its cinematic commu-
The specifics of dynamic influence of train pressure
nication with model of the subgrade can be realized:
on the subgrade are expressed in the following:
the model of pressure can transfer dynamic influences
– Influence of external dynamic factors from passing through models of rails, at a level of a ballast layer and
trains on the ground layer is shown in occurrence at a level of the basic platform of the subgrade.

964
 It is by process of elimination received; that it is the For the creation on the level of the model of the basic
most expedient to carry out the application of dynamic platform of subgrade of the dynamic pressure, close
model of train pressure at a level of the basic platform to real, the special transmitting device was created.
of model of the subgrade since modeling of the ele- This dynamic model is executed on MIIT centrifuge
ments of the upper structure of a track is practically as a set of elastic beams of variable section from the
impossible because of their small sizes on the model. material with the corresponding module of elasticity
The first condition is the preservation on the model e.g. textolite (Figure 1).
in the same points and in the same moments of time of Force Pm (see Figure1) equal from 0 up to 16
sizes of natural dynamic pressure, i.e. (i = 1, 2, 3 . . . ). kgf can be transferred to every beam, meanwhile the
This condition will be executed, if at a level of the basic dynamic pressures are created in the sections under
platform of a subgrade the stresses for the model and forces Pm and vibrations can be transferred onto the
the prototype are equal: subgrade under frequency from 0 to 200 Hz.
Criteria of similarity which are included in the basic The efficiency of dynamic pressure has been tested
criterion of similarity of N.A.Nasedkin are used: in experiments where the settlements of the basic
platform of model under certain pressure were cal-
gl culated—Figure 2.
= idem (5)
ω2
where ω = speed of influence on the soil; g = earth
gravity; l = the linear size.
10 20 30 40 50 60 70 80 90 100 t ,min
The second criterion is similar to the criterion of 0
10
5 7 6 12
9 Y
Frude, at centrifugal modeling: 1 2 11
loading
0
unload
2

An effective way of research slope stability of rail- 10

Z
way embankments is the method of physical modeling 2 6

11
on geotechnical centrifuges (Yakovleva, T.G. Ivanov, 3
D.I. 1980). 5
4
It has been proved (Shakhunyants, G.M. 1953), S,m
that the only and necessary condition, providing equal
characteristics of soil strength of nature and models Figure 2. Settlements of the embankment model with and
in the same points, in the same moments of time, is without dynamic loading.
Am = An and νm = νn .

Ðm
ì Ðm
ì
2 2

Vertical
Stresses

Displacements

model

Figure 3. Settlements of the embankment model with and

Figure 1. Scheme of model load and construction view. without dynamic loading.

965
 For the estimation of efficiency of centrifugal mod- track objects (subgrade) are occupy a big role that time
eling stability of objects of railway embankments, in Russian railways
experiments on modeling embankments of different The full-scale monitoring and physical modeling
height have been executed. on the geotechnical centrifuge are main directions of
It has been experimentally established, that at researches conducted in the Moscow State University
steeper slopes of model of an embankment, the weight of Railway Engineering.
of the soil in failure mass on slopes is more that Monitoring of the object of the subgrade on 103 km
corresponds to the natural data. of the Moscow-Aleksandrov line executed: for the
Also it has been tested, whether the cylindrical evaluation of changing a condition of geotechnical
form of the surface of failure is reproduced when structures in the exploitation time; for the qualification
modeling. The studying of formation of the failure of terms of changing of its condition; for the forecast
surface on models has shown, that is reproduced— of the behavior and determinations of life-time.
Figure 3. The several cycles of studies was realized on
When modeling the dependence of influence of the object, during which was supporting-geodetic
moisture w and the height of embankments h on the network, executed of the observation and geodetic
value of destroying dynamic pressure of a train. measurements, winnowed functional diagnostic of the
subgrade—a determination vibro-dynamic parame-
ters of soil vibrations from train loading.
2 PRACTICE OF SLOPE BEHAVIOR For the realization of supervision the displacement
PREDICTION FOR DEFORMING RAILWAY and revealing the change tense deformed conditions
EMBANKMENTS AND REINFORSING at different parts of the wall (gabion structure) were
CONSTRUCTIONS provided the bookmark an anchors and winnowed
geodetic measurements (Figure 5).
On the basis of the results of centrifugal modeling of Vertical and horizontal displacement of gabion
stability and behavior of slopes of some embankments structures defined comparatively pawned of the
both on real prototypes, and on the basis of develop- anchors and with the supporting geodetic network.
ment of the generalized parameters of embankments Practically the geodetic removal gabion walls com-
on the statistical data of Russian railways the num- prise of it:
ber of effective ways of reinforcement and stabilization – Utter measurement of the gabion levels height on
of potentially-dangerous and deforming embankments top for each row with step 1 meter.
of railways and a number of normative documents on – Measurement of the height for anchors (accessories
reinforcement of the subgrade of railways has been by diameter 16 mm and length 1 meter) with card
developed (Zaitsev A.A., Frolovsky Y.K. 2000). face of gabion wall.
In the Figure 4 the photo of one of the embank-
ments with reinforced by the gabion walls the design Analysis of gabion wall deformations executed on
of which has been developed on the basis of technical results of measurements settlements of the anchors and
instructions is presented. measurements, executed as of measurements of the
The MSE walls and gravitation gabion structures at level surface orientations.
geotechnical practice of reconstruction of the railway Executed analysis of the measurements of the
anchors displacement mortgaged with the card face
gabion walls is indicate of the happened settlements

Level A Level A
h.168.5 h.169.0 Level B
Level E
Level B1
Level G

Figure 4. Reinforcing of the subgrade on line Aleksandrov- Figure 5. Scheme of the levels for left and right side of
Balakirevo (Nothern Railway). gabion walls.

966
Table 1. Results of the utter measurements of the gabion
levels height (Right side).

1 3 7 17 29
Points∗ mm mm mm mm mm

Level A 10 24 13
Level B −18 12
Level B1 10 19
Level G 10 11 10
∗ Points in meters from the construction beginnings.

Table 2. Results of the utter measurements of the gabion

Figure 6. Correlation of the design and the actial positions
levels height (Left side).
of the wall. Settlements on right side.
0 1 2 4 10
Points∗∗ mm mm mm mm mm
their physical modeling on geotechnical centrifuge of
Level A 31 23 32 20 MIIT’s was spent.
Level E 32 36 21 According to the preliminary specifications on
∗∗
modelling solved following problems:
Points in meters from the construction beginnings.
– Researches of reasons and conditions of a strain-
ing of subgrade (based on the analysis of the
of the construction (walls) with following maximum information on plant of a subgrade);
values: – Estimations of working capacity of construction (at
a stage of construction and maintenance);
– Maximum values a settlements for construction of – Predictions of a deformation property of a subgrade;
the reinforcement on the right slope of embankment – Estimations of reliability of use of the accepted
and for anchor, which establishing on the level ‘‘A’’ computational models;
(in the cross-section 25 meters from the beginnings), – The comparative analysis of outcomes of modelling
has formed 9 mms; average values the deformation and numerical calculations of the is tense-deformed
do not exceed 3 mms (settlements) condition of a subgrade;
– Maximum values a settlements for the constructions – Development of recommendations on perfecting
of the reinforcement of the left slope of embank- designs of amplification and modes of their con-
ments and for anchor, which establishing on the struction.
level ‘‘D’’, has formed 6 mms; average values the
deformation do not exceed 4 mms (settlements) Application of a method of centrifugal modeling
has allowed to research and validate a deformation
The results of the utter measurements of the gabion property of various plants of a subgrade and designs
surface are shown of the lumpy settlements of the con- of their reinforcing on the small models fulfilled in
structions. The maximum values of the settlements are scale from 1: 25 till 1:50.
given in the Table 1 and Table 2. For reinforcing of embankment with height up to
Analysis of the actual position of the gabion walls 9.0 m the two-lane adapted to the use of electric
for design mark has shown certain updating actual power section Moscow-Aleksandrov Moscow Rail-
height in the construction period. ways have been developed technical instuctions with
So for the right side the maximum excess of the use of gabion gravitational walls.
actual mark on the design height has formed, for The embankment intersects river Gorelyi Krest with
example for the level ‘‘A’’ excess is 96 mms the constant stream flow. There is a water-absorbent
The results of full scale monitoring have shown pipe here an arch type combined of concrete blocks on
small precipitation of separate elements of the wall a cement slurry, with an orifice a breadth of 3,2 m and
construction and non-uniform changing a position in 4,5 m. In the location of a water-absorbent building
different parts of gabion walls. However revealed dis- (a southern leg) are observed by altitude a slope of
placements are not critical for the operation of the embankment of a bank.
construction. In a upper part the embankment is strengthened
For an estimation of operational working capacity by metal columns and the armoring. The breadth of
of designs on stabilization and reinforcing of a main- the basic platform is made narrower. The ballast is
tained railway subgrade with use of gabion structures, showered on the slope. Embankment slopes have the

967
overestimated steepness. From a southern leg of a bank The results of full scale monitoring have shown
the silted section of locality is arranged. The marsh small precipitation of separate elements of the wall
has the mixed feed—atmospheric, superficial and sub- construction and non-uniform changing a position in
soil waters. Average depth varies within the limits of different parts of gabion walls. However revealed dis-
0,3—0,6 m. Exist 5 ringlets which waters unload in placements are not critical for the operation of the
a marsh. The water condition has the mixed type of a construction.
feed which develops due to atmospheric precipitation Results of physical modeling and its analysis have
and subsoil waters. The basic quantity of flow of the allowed to evaluate operational working capacity of the
river (about 60%) happens due to atmospheric precip- offered the designs with use gabion and MSE walls. It
itation. Because of small depth a river valley the soil is necessary to note necessity of the account of possible
feed has smaller value. The hydrological condition of change of state of a reinforced embankment under con-
the river is tightly connected with climatic singularities struction and the subsequent maintenance as the time
of territory. It represents a strongly pronounced spring of consolidation of the waterlogged clayey grounds
high water on which share 60–70% of an annual flow of the foundation can exceed considerably a time of
are necessary. A bottom river Gorelyi Krest generally construction of designs of its reinforcing gravitational
it is silty, covered by vegetation. gabion walls.
Hence, it is possible to state, that outcomes of mod- For an estimation of a deformation property of nat-
eling and their calculated analysis has shown, that ural railway subgrade objects after their reinforcing
origin of not stabilized condition of an embankment by gabion structures was developed the program of
if to the foundation combined by soils with low fil- geotechnical and geodesic monitoring.
tration ability, exterior loading will be fast enclosed On an instance of embankment of 103 km Moscow-
is possible. As a result of its operation owing to for- Alexandrov Moscow Railway, reinforced by gravita-
mation of a gauge pressure in void water the angle tional gabion wall, are performed works on monitoring
of an interior friction of the soils, largely determin- of an condition of the gabion structures and defini-
ing a load-carrying capacity of usual soils of the tion vibro-dynamic parameters of soils under loadings
foundation, is not realized. Thus, the load-carrying from a rolling-stock in a cycle 2006–2007 years.
capacity of the foundation will be defined only by For improvement of quality and effectiveness of
values specific cohesion its soils, and it can appear realization of researches by a method of centrifugal
insufficient. Therefore, emersion of strains of an modeling on geotechnical centrifuge of MIIT it was
embankment, because of deficiency of its overall equipped by the modern control equipment and the
stability is possible. automated laboratory complex for definition of physic
In this connection, it is necessary to note neces- mechanical properties of soils. There are begun works
sity of the account of possible change of state of a on working off of conditions of use of control and
reinforced embankment under construction and the measuring system at modeling subgrade objects of
subsequent maintenance as the time of consolidation railways.
of the waterlogged clayey grounds of the foundation In the further researches we project to continue
can exceed considerably a time of construction of works on monitoring and modeling of subgrade
designs of its reinforcing gravitational gabion walls. objects of railways both under static and under
dynamic loadings, reinforced by gabion structures;
to carry out researches with application of the auto-
3 CONCLUSIONS mated laboratory complex for compiling certificates
of soils which will be used at numerical calculations
The carried out experimental and theoretical investi- of embankment.
gations of prediction of slope behavior for deforming
railway embankments of have shown that the forms of
surfaces of slope displacement of subgrade depend on REFERENCES
the dynamic pressure of the train on the basic of the
subgrade. Yakovleva, T.G. & Ivanov, D.I. 1980. Modeling of toughness
The specially created for such purposes device and stability of railway subgrade. Moscow: Transport.
placed on the centrifuge facility has allowed to etimate Shakhunyants, G.M. 1953. Subgrade of railways. M.: Tran-
the levels of destroying pressures of a train. sjeldorizdat. 827.
For check of working capacity and an estimation Pokrovsky G. Y. & Fedorov I. S. 1975. Centrifugal testing in
the construction industry. Vols. 1 & 2. English translation
of a use reliability of new technology with use of by Building Research Establishment Library Translation
gabion structures for stabilization and reinforcing of Service of monographs originally published in Russian,
subgrade of railways have been worked out experi- Watford.
mental researches: full scale monitoring and physical Vinogradov V.V., Yakovleva T.G., Frolovsky Y.K. &
modeling on the geotechnical centrifuge of MIIT. Zaitsev A.Al. 2002. Centrifugal modeling of the railway

968
 embankments with reinforcement by the various rein- Ministry of Railway Transport. 1998. Technical instructions
forced earth constructions. Proceedings of the Interna- on reinforcement of embankments with using an gabions.
tional Conference on Physical Modeling in Geotechnics, 140p
St.John’s, Newfoundland, Canada, 10–12 July: 987–991. Yakovleva, T.G. 1992. Principle bases of using reinforced soil
Vinogradov V.V., Yakovleva T.G., Frolovsky Y.K. & Zait- for reinforcement of subgrade. Transactions of Moscow
sev A.Al. 2005. Evaluation of slope stability of railway Railway Engineering Institute. 844: 45–58.
embankments. Proc. of the 16th International Conference Yakovleva, T.G. & Vinogradov, V.V. & Frolovsky, Y.K.
on Soil Mechanics and Geotechnical Engineering, Osaka, 1997. Methods of stabilization of the embankments by
Japan. the constructions from the reinforced soils. Put i putevoe
Vinogradov V.V., Yakovleva T.G., Frolovsky Y.K. & Zait- hoziaystvo. 1: 7–11.
sev A.Al. 2006 Physical modeling of railway embank- Frolovsky, Y.K & Zaitsev A.A. 2000. Database ‘‘Infor-
ments on peat foundations. Proc. of the 6th International mation card’’ for analogues of railway embankments.
Conference on Physical Modelling in Geotechnics—6th Proc. of the 3rd International Conference on Advances of
ICPMGE’06, Hong Kong. Computer Methods in Geotechnical and Geoenviromental
Zaitsev A.A., Mnushkin M.G. & Vlasov A.N. 2001. Object- Engineering, Moscow: Balkema.
oriented programming in application to engineering meth- Yoo, N.J. & Ko, H.Y. 1991. Centrifuge modeling of rein-
ods for slope stability analysis. Proceedings of the 8th forced earth retaining walls. Proc. of the International
International Conference on Enhancement and Promotion Conference Centrifuge, Boulder: Balkema.
of Computational Methods in Engineering and Science
(EPMESC’VIII), Shanghai, China, 9p
Ministry of Railway Transport, 1991. Technical instructions
on reinforcement and stabilizations an embankments on
strong foundation of the soil reinforced constructions.
101p

969
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Finite element simulation for the collapse of a dip slope during 2004 Mid
Niigata Prefecture earthquake in Japan

A. Wakai & K. Ugai

Gunma University, Japan

A. Onoue
Nagaoka National College of Technology, Japan

K. Higuchi
Kuroiwa Survey and Design Office Co., Ltd., Japan

S. Kuroda
Institute for Rural Engineering, JAPAN

ABSTRACT: Many landslides in mountain area occurred during 2004 Mid Niigata Prefecture Earthquake in
Japan. In this paper, numerical simulations for the collapse of a dip slope by the 2-D dynamic elasto-plastic finite
element method is reported. In the case, an upper mudstone has slid along the bedding plane. To simulate such
a catastrophic failure, i.e., a long distance traveling failure, it is very important to consider the strain-softening
characteristics of the slip surface precisely. In the analysis, the material parameters for the thin sand seam put
along the bedding plane were determined by the cyclic direct shear tests of undisturbed block samples. As a
result, the observed phenomena could be simulated by the analysis appropriately.

1 INTRODUCTION finite element method is reported. In the case, an upper

mudstone has slid along the bedding plane. To simu-
In engineering point of view, it should be empha- late such a catastrophic failure, i.e., a long distance
sized that the following two kinds of slope failure traveling failure, it is very important to consider the
can be distinguished definitely. The fist one is a large strain-softening characteristics of the slip surface pre-
deformation, where the sliding mass stops after the cisely. In the analysis, a newly proposed constitutive
earthquake. In such a case, the degree of the deforma- model that is extended from the previous model (Wakai
tion often attracts attention for the seismic design of & Ugai, 2004) is adopted to consider such a behavior.
adjacent structures. The second one is a catastrophic The material parameters for the thin sand seam put
failure that is often called as a collapse. In such a along the bedding plane are determined by the cyclic
case, the sliding mass continues moving even after direct shear tests of undis-turbed block samples. As a
the earthquake as far as there are no obstacles on result, the observed phe-nomena are simulated by the
the way. It means that the sliding mass will not be analysis.
supported after the earthquake, although it was sup-
ported statically before the earthquake. The reduction
of the shear resistance along the slip surface during
2 ANALYTICAL MODEL
the earthquake may be one of the main causes for
the induction of such a long distance traveling fail-
2.1 Yokowatashi landslide
ure. In most of previous studies based on the dynamic
elasto-plastic FEM, the slope failure of the latter type Many landslides in mountain area occurred during
has not been treated, because of the difficulty of the 2004 Mid Niigata Prefecture Earthquake in Japan. In
modeling. this paper, one dip slope failure along the Shinano
In this paper, numerical simulations for the col- River, Yokowatashi landslide, is simulated numeri-
lapse of a dip slope by the 2-D dynamic elasto-plastic cally as an example of natural slope failure. In the

971
analysis, the dynamic elasto-plastic FEM is applied
for reproducing the long distance slide taking strain
softening and decrease in strength due to cyclic shear
loading into consideration.
As seen in the geological map shown in Figure 1,
the Shinano River joins the Uono River at Kawagu-
chi town near the epicenter of the earthquake and flows
toward almost north from Shiroiwa, Nagaoka city after
scraping the right coast at Yokowatashi, Ojiya city. The
river encroaches on the attacking rock slopes ranging
from Shiroiwa to Myoken, Na-gaoka city along the
right coast. Neogene deposits, such as Shiroiwa layer,
and Uonuma layer sedi-mented by early Diluvium dis-
tribute as hills and ter-races from Kawaguchi town to
Nagaoka city along the Shinano in this area. These
layers fold several times forming the Higashiyama hill
areas and the axis directions of these anticlines and
synclines are almost South-North in direction. The
Shinano locates to the west of the west end anticline
axis. The slopes inclining toward west and facing to
the river are dip slopes from geological structure view
point.
Figure 2 (a) is a photograph taken from a north-west
direction. As seen in the figure, a part of the upper
Shiroiwa layer and the surface earth with high trees
remain as they were on the bedding plane at far end

Figure 2. Yokowatashi landslide.

Nagaoka

Alluvium
Terrace
deposit
Uonuma of this picture. The remaining upper Shiroiwa layer of
layer soft silt rock exposes its side face. The other part of
New the upper Shiroiwa layer which made up the opposite
tertiary
side of the slid area is visible on site. The portion of
Ojiya Syncline the upper Shiroiwa layer between them had covered
axis
city Anticline the planer tectonic dip surface which is clearly seen in
axis the picture, and it has slid more than 72 m to the west
Slant of
strike toward the Shinano River. The inclination of the bed-
T he Failure ding plane facing to almost west is approximately 22◦ .
Shinano area The thickness of the slid Shiroiwa block at the south
R iver end is about 4 m and those of earth on the block ranges
Yokowatashi
from 20 cm to 1 m. The height of upper Shiroiwa
layer remaining at the north side is about 2.5 m with
earth cover of 60 cm thick near the ridge of the slope.
Figure 2(b) is a close picture showing soil sampling
for the laboratory tests. A thin seam layer of 5–10 mm
Kawagu- thick was sandwiched between the upper and lower
chi town Shiroiwa layers. The material of the sand seam is tuff
sand. Both Shiroiwa layers were gray, weathered, and
changed their color to brown up to about 8 cm inside
from the boundary of the sand seam. The details of
Yokowatashi landslide has already been reported by
Figure 1. Geological map around Ojiya city. Onoue et al. (2006).

972
 Extracted cross section

Moving direction Slip surface along

the bedding plane
Bedding plane (10mm thick) 20.7m

Sand and rock layer Point A

Shiroiwa layer 63m
30m
22˚

Figure 3. Two dimensional finite element meshes for the simulation.

Table 1. Input parameters.

Sand
Sand and
Layer Shiriowa seam gravel

Basic Young’s 100000 30000 30000

parameters modulus
E(kN/m2 )
Poisson’s 0.3 0.3 0.3
ratio v
Cohension – 24 0
c(kN/m2 )
Internal – 30.9 35
friction
angle
φ(deg) Figure 4. Test specimen consisting of upper and lower
Dilatancy – 0 0 Shiroiwa layers with sand seam in between.
angle
ψ(deg)
b · γG0 – 8.0 18 analysis was focused on the cross section of the slid
n – 1.40 1.35 slope with its medium thickness. The finite element
Unit 20.0 18.0 18.0 mesh consisting of eight nodes per each element were
weight shown in Figure 3. The time history of the response
γ (kN/m3 ) at Point A in the figure will be mentioned later. The
Strain Residual – 0.30 – upper and lower soft rock layers were assumed to be
Softening strength elastic material and the sandwiched tuff sand layer
parameters τπ /τf 0
was assumed to be elasto-plastic material having a
A – 4.0 –
thickness of 10 mm taking strain softening into con-
sideration. The surface soil at the foot of the slope was
assumed to be sand and gravel spreading down to the
Shinano.
2.2 Analytical model
The basic concept of the elasto-plastic model used
As aforementioned, the thickness of the slid soft rock here is the same as the cyclic loading model originally
plus earth with trees was thin at the north end and it was proposed by Wakai and Ugai (2004). The undrained
thick at the south end. A two dimensional numerical strength parameters, c and φ, which specify the upper

973
 100 100

80 80

60 60

(kN/m )
(kN/m2)

2
40 40

20 20

0 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0 0.1 0.2 0.3 0.4 0.5 0.6
-20 -20

-40 -40

(a) Initial normal stress, —=40kN/m2

100 100
80 80
60 (kN/m ) 60
(kN/m2)

40 40
20 20

0 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0 0.1 0.2 0.3 0.4 0.5 0.6
-20 -20

-40 -40
(Shear displacement ) / (Thickness : 10mm)

(b) Initial normal stress, —=80kN/m2

100 100
80 80
60 60
(kN/m )
(kN/m2)

40 40
20 20

0 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0 0.1 0.2 0.3 0.4 0.5 0.6
-20 -20

-40 -40
(Shear displacement ) / (Thickness : 10mm)

(c) Initial normal stress, —=120kN/m2

Figure 5. Comparison between tested (left column) and simulated (right column) hysteresis loops.

asymptotic line of the hyperbolic skeleton curve of strain softening characteristics (Wakai et al. 2005) and
their model was modified as the decreasing functions used here. The shear strength during earthquake is
of accumulated plastic strain γ p to incorporate the given as,

974
 τfr − τf 0 p
τf = τf 0 + γ (1) 10

Horizontal acceleration (m/s )

2
A + γp Input motion
5
where the initial strength is denoted as,
0
 0 10 20 30 40 50
σ 1 + σ3 -5
τf 0 = c · cos φ + × sin φ (2)
2 initial
-10
Time (s)
The shear stiffness ratio, G0 , is also assumed to
decrease in proportion to the decrease of shear
strength. The cyclic loading model disregarding strain Figure 6. Acceleration record at Takezawa (EW).
softening was used for the sand and gravel layer.
The constants of Rayleigh damping were assumed to 1
be basically α = 0.171 and β = 0.00174 which Point A

Horizontal displacement (m)

are equivalent to a damping ratio of about 3% for a 0.5

vibration period of 0.2 through 2.0 s. The material 0

properties used in the analysis were summarized in 0 10 20 30 40 50
-0.5
Table 1.
Figure 4 shows an intact sample consisting of the -1 Input acceleration amplitude
upper and lower Shiroiwa soft rocks and the tuff sand Full
-1.5
seam in between. The sample was subjected to the Half
cyclic direct shear test under the constant volume con- -2
dition. Figure 5 compares the simulated hysteresis loop Time (s)
and the tested loop of each specimen during cyclic
loading. The axis of abscissas is written in strain. In the 10
figures for the tests, the strain is defined as the horizon- 0
Horizontal displacement (m)

tal displacement divided by 10 mm, which corresponds -10 0 10 20 30 40 50

to the thickness of the sandwiched layer. Although they -20
don’t perfectly coincide, they are roughly similar to -30
each other for the respective consolidation pressure. -40
-50
-60 Full
-70 Half
3 ANALYTICAL RESULTS
-80

3.1 Time histories of response Time (s)

The acceleration record in EW direction observed at Figure 7. Time history of horizontal displacement of upper
Takezawa (Figure 6) was used in analyses. Two cases Shiroiwa layer.
of analysis were conducted to examine the influence
of seismic intensity on the inducement of sliding. One
was the analysis for which the observed acceleration
sliding on the bedding plane started at an elapsed time
record was input as it was at the base of the analysis
of about 35 s.
area, and the other was the one for which the accelera-
Figure 8 shows the relationship between the mean
tion amplitude was compressed to one half that of the
shear stress and the mean accumulated shear strain of
observed wave was input.
the sand seam, both of which were averaged in all over
Figure 7 shows the time histories of horizontal dis-
the sandwiched layer. The accumulated plastic strain
placement at the foot of the slope, namely Point A,
converged at γ = 1.5 and τ = 474 kN/m2 in the case
in Figure 3. As seen in this figure, the slope does
of half acceleration amplitude. On the contrary, the
not fail in the case where the acceleration amplitude
shear stress decreases continuously with increasing γ
is compressed to one half that of the actual wave re-
for the actual wave case.
cord. Contrarily the large-scale slope failure occurs in
the case of actual acceleration amplitude. The sliding
amount in horizontal direction is 20 m at t = 40 sec
3.2 Deformation increasing endlessly
and almost 65 m at t = 50 sec. Since the shear strength
became smaller than the shear stress induced only by The residual displacement at 0, 10, 20, 30, 40 and
the self weight of upper Shiroiwa layer, continuous 50 seconds after the beginning of the seismic motion,

975
 70 0s
Initial stress

60

50
Averaged shear stress (kN/m 2 )

Solid line converged here.

40 10 s

30

20
Input acceleration amplitude
20 s
10 Full
Half
0
0 1 2 3 4 5 6 7 8 9 10
Averaged shear strain (-)

30 s
Figure 8. Relationship between shear stress and shear strain
of the sand layer during earthquake.

in the case where the observed acceleration record is

input, was shown in Figure 9. The long distance slid-
ing of the upper Shiroiwa layer along the bedding plane 40 s
can be seen discontinuously at the sand seam in this
cross section. The large-scale slide occurred on site
was thus reproduced quantitatively through the present
analysis.

3.3 Newly proposed index for seismic stability 50 s

At the time of the collapse of the slope, the total value

of maximum shear resistance mobilized on the sliding
plane Ts begins to be less than the total value of the
static sliding force corresponding to the gravity of the Figure 9. Movement of the sliding mass on the bedding
sliding block Rf . An appropriate stability index for plane during the earthquake.
descriptions of this phenomenon can be denoted as the
following equation.
 the sandwiched material on the bedding plane and
Rf
Fd =  (3) its long distance sliding was appropriately simulated
Ts by the dynamic elasto-plastic FEM. The following
con-clusions were obtained from the present work;
The time histories of the proposed index obtained in
the analyses are shown in Figure 10. The moment when 1. There exists a tuff sand seam at the bedding plane
the value of the index becomes 1.0 accords with a of Yokowatashi dip slope structure.
moment of the collapse. This index will be extremely 2. The shear strength and stiffness of the sand decrease
convenient for explanation. markedly with increasing number of cycles and the
sand seam finally lose its shear strength enough to
support just the self weight of soils above it.
4 CONCLUSIONS 3. The long sliding distance of the slope failure
was reproduced through the elasto-plastic dynamic
The cause of the slope failure at Yokowatashi was finite element analysis taking the cyclic shear
examined based on the cyclic shear properties of properties of the sand into consideration.

976
 1.8 REFERENCES
1.6 Collapse Griffiths, D.V. & Prevost, J.N. 1988. Two- and three-
1.4 dimensional dynamic finite element analyses of the Long
Fd (=SRf/ STs)

1.2 Valley Dam, Geotechnique, Vol. 38, No. 3, pp.367–388.

Iai, S., Ichii, K., Sato, Y. & Kuwazima, R. 1999. Earthquake
1 response analysis of a high embankment on an existing hill
0.8 slope. 2nd International Conference on Earthquake Geo-
technical Engineering, pp. 697–702, Lisboa, Portugal.
0.6 Onoue, A., Wakai, A., Ugai, K., Higuchi, K., Fukutake, K.,
Full
0.4 Hotta, H. & Kuroda, S. 2006. Slope failures at Yokowa-
Half tashi and Nagaoka College of Technology due to the 2004
0.2 Niigata-ken Chuetsu Earthquake and their analytical con-
0 siderations. Soils and Foundations, Vol. 46, No. 6, pp.
751–764.
0 10 20 30 40 Toki, K., Miura, F. & Oguni, Y. 1985. Dynamic slope sta-
Time (s) bility analyses with a non-linear finite element method,
Earthquake Engineering and Structural Dynamics, Vol.
13, pp. 151–171.
Figure 10. Time histories of the proposed index. Ugai, K., Wakai, A. & Ida, H. 1996. Static and dynamic anal-
yses of slopes by the 3-D elasto-plastic FEM, Proc. of 7th
International Symposium on Landslides, pp.1413–1416,
Trondheim, Norway.
4. A simple stability index for descriptions of the col- Uzuoka, R. 2000. Analytical study on the mechanical behav-
lapse of the slope was proposed. It will be extremely ior and prediction of soil liquefaction and flow. A doc-
convenient for explanation. toral dissertation of Gifu University, Japan (in Japanese),
pp.161–170.
Woodward, P.K. & Griffiths, D.V. 1994. Non-linear dynamic
analysis of the Long Valley Dam. Computer Methods and
Advances in Geomechanics, Balkema, pp.1005–1010.
ACKNOWLEDGEMENTS Wakai, A. & Ugai, K. 2004. A simple constitutive model for
the seismic analysis of slopes and its applications. Soils
and Foundations, Vol. 44, No. 4, pp. 83–97.
This study was conducted in ‘‘Research project for uti- Wakai, A., Kamai, T. & Ugai, K. 2005. Finite element
lizing advanced technologies in agriculture, for-estry simulation of a landfill collapse in Takamachi Housing
and fisheries’’ supported by the Ministry of Agricul- complex (in Japanese). Proc. simpojium on Safeness and
ture, Forestry and Fisheries of Japan. The au-thors performance evaluation of ground for hausing, JGS, pp.
wish to thank for the great support. 25–30.

977
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Sensitivity of stability parameters for soil slopes: An analysis

based on the shear strength reduction method

Ren Wang, Xin-zhi Wang, Qing-shan Meng & Bo Hu

State Key Laboratory of Geomechanics and Geotechnical Engineering, Institute of Rock and Soil Mechanics,
Chinese Academy of Sciences, Wuhan, China

ABSTRACT: The stability of soil slope is analyzed through explicit Lagrangian finite-difference method
based on the strength reduction method. The predominance of this method is discussed. It is a fast method.
Dilatation angle, cohesion and internal friction angle are taken into consideration in analyzing of soil slope. It is
concluded that the safety factor will become bigger as the dilatation angle, the cohesion and the internal friction
angle, while the plastic area and the slip plane will have fluctuation. The results can be used in practical slope
design.

1 INTRODUCTION plane is showed in figure 1. This example has been

widely accepted to checkout the validity of analy-
There are many reasons causing slope failure such sis program. The FLAC program based on Strength
as strength reduction of rock and soil mass, load- Reduction Method reduces the shear strength of soil
ing, slope excavation and so on. But for a natural through dividing the cohesion and internal friction
slope, the key factor is the strength reduction of
rock and soil mass. Therefore, the Strength Reduc-
tion Method can be used in slope analyzing work.
Explicit Lagrangian Finite-Difference Method is pop-
ularly used in geomechanical analyzing work. Because
it has nonlinear constitutive functions and can be appli- possible slip plane
cable for any complex boundary condition (Hnang
Run-qiu & XU Qiang 1995.). Failure surface must be
assumed in analyzing slope stability using Limiting
Equilibrium Method. However, the FLAC based on
Strength Reduction Method does not need assumption
of failure surface, and can educe factor of safety, and Figure 1. Shape of slope and the possible slip plane (m).
display the contour and location of failure surface. So
this method is widespread in geoengineering.
Slope stability comes down to many parameters Table 1. Mechanical parameter of slope.
such as gravity of soil, dilatation angle, cohesion and
internal friction angle. The single parameter is ana- c (KPa) φ(◦ ) γ (kN/m3 ) E (kPa) μ
lyzed to reveal the most important influence to the
3.0 19.6 20.0 1.0e4 0.25
slope stability. The results can be used in slope design.

2 CALCULATION MODEL

The analysis model is an examine originating from

Australia Computer Society designed by B. Donald
and P. Giam in 1987. The slope consists of homoge-
neous soil. The shape and parameters of the soil slope
are displayed in figure 1 and table 1. The safety fac-
tor is recommended for 1.0. The most possible slip Figure 2. Slope model and lattice division.

979
angle by the same discount coefficient (Chi Shi- 4 INFLUENCE TO STABILITY OF SLOPE
chun & Guan Li-jun 2004) (formula 1∼3).
4.1 Dilatation angle

c For dense sand and over consolidated clay soil, dilata-

c = (1) tion is observed in shear process. There is explicit
Fsr peak strength on the stress-strain curve. Firstly, stress
tan φ and stain increases simultaneously to the peak value,
φ  = tan−1 (2) and then strain continues to increase but stress dimin-
Fsr
ishes to residual strength. Dilatation occurs before
tan ψ the peak strength and terminates to material fail-
ψ  = tan−1 (3) ure. Soil becomes loose and shear strength reduces
Fsr
because of dilatation. So the dilatation angle should
not be ignored in the slope stability calculation. Adopt-
Fsr —discount coefficient ing non-associated flow rule which Ignores dilatation
The slope is analyzed using the FLAC program will underestimate the shear strength of soil in the
to validate the adequacy of the method of strength stability analysis of slope based on elastic-plastic
reduction. model (Zienkiewicz O.C., Humpheson C. & Lewis
R.W. 1975). But the bearing capacity of soil will be
over-evaluated adopting associated flow rule (Zhang
3 BOUNDARY CONDITION AND RESULTS Peiwen & Chen Zu-yu 2004). Therefore, the influence
of dilatation angle in FLAC analyses based on Strength
In the FLAC2D analysis model, the bottom is fixed, Reduction Method should be studied particularly.
side surface has normal restraint and sloping sur- In order to study the influence of dilatation to slope
face is free. Material properties are given in table 1. stability, series of numerical analysis have been done
The elastic-plastic constitutive equation and Mohr- which dilatation angle ψ varies from 0◦ to 17◦ and
Coulomb model are employed. Large deformation other parameters remain the same. The results display
is considered to adjust the grid coordination. The on table 4 and figure 3 that the safety factor has a linear
safety factor of the slope is 0.99 comparing to
some traditional limiting equilibrium method listed on
Table 4. Dilatation angle and safety factor.
table 2. The results of judging program are showed on
table 3. Dilatation angle (degree) ψ Factor of safety F
The result of using strength reduction method
is quite close to true value. Plastic and tensioned 0 0.96
area distribution, shear strain rate and displacement 3 0.97
vectograph are almost the same to the judging pro- 5 0.98
gram. It can be considered reliable in analyzing slope 8 0.99
stability. 10 0.99
12 1.00
15 1.01
17 1.01
Table 2. Results of calculation cases.

Safety factor F
1.02
Standard y = 0.003x + 0.9622
Analytic method Mean deviation Min.F Max.F 1.01 R2 = 0.9765

ALL 0.991 0.031 0.94 1.08 1.00

Factor of safety

Bishop 0.993 0.015 0.96 1.03

Janbu 0.978 0.041 0.94 1.04 0.99

0.98

Table 3. Results of judging program. 0.97

0.96
SSA STAB
0 5 10 15 20
Donald (Baker) (Chen) GWEDGEM EMU Fredlud
Dilatation angle (degree)
1.00 1.00 0.991 1.00 1.00 0.99
Figure 3. Safety factor vs dilatation angle.

980
relationship with dilatation angle. The formula can be
described as follows:

F = 0.003ψ + 0.9622 (4)

Safety factor and dilatation angle have good lin-

ear correlation (figure 3) which R2 = 0.9765. Safety Figure 4. Shear strain rate and displacement vectograph
factor slightly increases as the dilatation angle but (c = 0 KPa).
not significantly. When dilatation angle increases, the
plastic area spreads from slope crest to bottom, dis-
placement increases at the bottom of the slope and the
location of slip surface almost remain stable.

4.2 Cohesion
Cohesion which is one of the strength parameters rep-
resents the cementation of soil. The cohesion of sand is Figure 5. Plastic and tensioned area distribution (c = 0 KPa).
considered zero. We consider the slope consists of sand
soil which its cohesion is zero. The slope is unstable. 3
Safety factor is only 0.72. Plastic area and failure con- y = 0.0467x + 0.8474
centrates on slope surface. If diminish the angle of 2.5 R2 = 0.9871
slope, the stability of slope will be improved. We let
2
the cohesion value varies from 0 to 35 KPa, analyses
factor of safety F

the stability of the slope and find that the safety fac- 1.5
tor has a linear relationship with cohesion value. The
formula can be described as follows: 1

0.5
F = 0.0467c + 0.8474 (5)
0
0 10 20 30 40
The safety factor increases linearly with cohesion, cohesion c (kPa)
gliding mass enlarges, failure surface shifts down and
becomes concave-down from a plane, its radian grad- Figure 6. Safety factor vs dilatation angle.
ually enlarges and drops back on the slope crest, slip
band extends to deep soil layer. Plastic area distributes
around the slip band. Most of the slope crest has tensile
stress.

Table 5. Calculation parameters and results.

c (KPa) φ(◦ ) ψ(◦ ) γ (kN/m3 ) F

Figure 7. Plastic and tensioned area distribution (c =
case1 0 19.6 10 2000 0.72 12 KPa).
case2 1 19.6 10 2000 0.84
case3 2 19.6 10 2000 0.92
case4 3 19.6 10 2000 0.99 4.3 Internal friction angle
case5 6 19.6 10 2000 1.18
case6 9 19.6 10 2000 1.33 The shear strength of soil is related to internal friction
case7 12 19.6 10 2000 1.47 angle which influences the stability of soil slope. The
case8 15 19.6 10 2000 1.61 relation between internal friction angle and safety fac-
case9 18 19.6 10 2000 1.74 tor of soil slope has been studied in order to make out
case10 21 19.6 10 2000 1.86
the regularity. The analyses model has been presented
case11 25 19.6 10 2000 2.02
case12 30 19.6 10 2000 2.21 on figure 1 & figure 2. Internal friction angle varies
case13 35 19.6 10 2000 2.4 from 0◦ to 25◦ and the other parameters are fixed. The
parameters and results are displayed on table 6.

981
 The results show that failure occurs when φ = 0◦ great importance to stability of soil slope. The impact
(figures 10∼11). Slip band is concave-down deep index is 0.454.
at the bottom of the slope. When internal friction
angle increases slip band rises close to the slope sur- F = 0.454φ + 0.1114 (6)
face and plastic zone concentrates to gliding surface
(figures 12∼15). Safety factor increases linearly with
the internal friction angle. The formula 6 and figure 9
given below suggest that internal friction angle has a

Figure 11. Plastic and tensioned area distribution of slope

(φ = 0◦ ).

Figure 8. Shear strain rate and displacement vectograph

(c = 12 KPa).

Table 6. Calculation parameters and results.

C (KPa) (◦ ) (◦ ) γ (kN/m3 ) F

case1 3 0 10 2000 0.09

case2 3 5 10 2000 0.36 Figure 12. Shear strain rate and displacement vectograph
case3 3 10 10 2000 0.58 (φ = 10◦ ).
case4 3 15 10 2000 0.79
case5 3 20 10 2000 1.01
case6 3 25 10 2000 1.24

1.6
y = 0.0454x + 0.1114
1.4 R2 = 0.9991
1.2
Figure 13. Plastic and tensioned area distribution of slope
1
(φ = 10◦ ).
factor of safety

0.8
0.6
0.4
0.2
0
0 10 20 30
internal friction angle (degree)

Figure 14. Shear strain rate and displacement vectograph

Figure 9. Safety factor vs internal friction angle. (φ = 30◦ ).

Figure 10. Shear strain rate and displacement vectograph Figure 15. Plastic and tensioned area distribution of slope
(φ = 0◦ ). (φ = 30◦ ).

982
5 CONCLUSIONS of slip band according to cohesion and internal friction
of soil so that can select proper length of anti-slide
The stability of soil slope is one of the hot prob- piles or inject grout to enhance the stability of slope.
lems because of the harmfulness and its complexity.
We analyses the stability of the given soil slope
through Lagrangian Finite-Diference Method based ACKNOWLEDGEMENTS
on Strength Reduction Method which is proved to be a
good method. Through analysis some suggestions can This work was funded by the National Science
be made that might be helpful to slope design. Foundation of China (No. 50639010), the Chinese
Dilatation slightly influences the stability of soil Ministry of Science & Technology projects (No.
slope that the safety factor slightly increases with 2006BAB19B03), the Knowledge Innovation Program
it. When dilatation angle increases, the plastic area of the Chinese Academy of Sciences (O712041Q01) to
spreads from slope crest to bottom, displacement Wang. Authors would like to thanks State Key Labora-
increases at the bottom of the slope and the location of tory of Geomechanics and Geotechnical Engineering.
slip surface almost remain stable. Variation of dilata- We thank the anonymous reviewer and Chairman
tion does not impact the location of slip plane. The Organizing Committee of the 10th ISL Prof. Zuyu
safety factor increases linearly with cohesion, glid- Chen for constructive comments and suggestions.
ing mass enlarges, failure surface shifts down and
becomes concave-down from a plane, its radian grad-
ually enlarges and drops back on the slope crest, slip REFERENCES
band extends to deep soil layer. Plastic area distributes
around the slip band. Most of the slope crest has tensile Chi Shi-chun & Guan Li-jun. 2004. Slope stability anal-
stress. ysis by Lagrangian diference method based on shear
When the internal friction angle is close to zero, soil strength reduction[J]. Chinese Journal of Geotechnical
slope must be unstable. Slip band is concave-down Engineering. 26(1): 42–46.
deep at the bottom of the slope. When internal fric- Huang Run-qiu & Xu Qiang. 1995. Application of explicit
tion angle increases slip band rises close to the slope lagrangian finite—difference methodin rock slope engi-
surface and plastic zone concentrates to gliding sur- neering. Chinese Journal of Rock Mechanics and Engi-
face. Safety factor increases linearly with the internal neering, 14(4): 346–354.
Zhang Pei-wen & Chen Zu-yu. 2004. Finite element method
friction angle. The formula 6 and figure 9 suggest for solving safety factor of slope stabilily[J]. Rock and
that internal friction angle has a great importance to Soil Mechanics, 25(11): 1757–1760.
stability of soil slope. The impact index is 0.454. Zienkiewicz O.C., Humpheson C. & Lewis R.W. 1975. Asso-
In slope prevention project designers can estimate ciated and non-associated visco-plasticity and plasticity in
the location of slip plane, the thickness and dimensions soil mechanics[J]. Geotechnique, 25(4): 671–689.

983
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Back analysis of unsaturated parameters and numerical seepage simulation

of the Shuping landslide in Three Gorges reservoir area

Shimei Wang, Huawei Zhang, Yeming Zhang & Jun Zheng

Key Laboratory on Geological Hazards of Three Gorges Reservoir Area, Ministry of Education of China,
Three Gorges University, Yichang, Hubei, China

ABSTRACT: Obvious ground deformation and displacement occurred in Shuping Landslide in Three Gorges
reservoir in June 2003, when the reservoir level reached to EL.135 m. Rainfall and the change of groundwater
seepage caused by raise of reservoir water level are the main reasons of landslide deformation. Groundwater
movement in landslide mass is a kind of unsaturated and unstable seepage flow, which should be analyzed by
the theory of unsaturated unstable seepage. Before the analysis, the determination of parameters is the primary
problem. Based on the observed rainfall and water level of drill holes in the landslide area in July 2005, orthogonal
optimizing method was used to conduct back analysis for getting the seepage parameters. After that, numerical
simulation of groundwater seepage in Shuping landslide mass was carried out by considering real situation of
reservoir water level changes. The variation of groundwater seepage in landslide area with the reservoir water
level changes was analyzed. The research results will provide useful information for predicting landslide stability
under the action of reservoir water level.

1 INTRODUCTION holes in July 2005. After that, numerical simulation of

groundwater seepage in Shuping landslide mass was
The impoundment of Three Gorges reservoir has conducted according to the real situations of reservoir
brought a great impact on deformation and stabil- water level changes. The variation of seepage field
ity of reservoir bank. In June 2003, after one month in landslide mass with the changes of reservoir water
of the reservoir water level reached to EL.135 m, a level was got from the analysis.
large scale high-speed landslide with the volume of
24 million cubic meters occurred at Qianjiangping,
Shazhenxi town, Zigui County in the reservoir area[1]. 2 BACKGROUND OF SHUPING LANDSLIDE
At the same time, obvious deformation and displace-
ment was found in Shuping landslide area, where it Shuping landslide area is located on the south bank
is only 3 km away from Qianjiangping landslide area. of the main stream of Yangze River, reservoir area of
It can be predicted that a great numbers of landslides Three Gorges Project, Shazhenxi town, Zigui County,
areas which already have certain deformation will be Hubei province. It is 3 km away from the upstream
unstable, and many new potential unstable slopes will Qianjiangping Landslide area and 47 km away from
appear. This situation will bring a great threat on the the downstream Three Gorges Dam (Figure 1).
lives and properties of residents in the reservoir area.
The variation of groundwater seepage due to the
change of reservoir water level and rainfall are the
essential reasons of slope deformation. For simu-
lating the variation of groundwater seepage under Shuping Landslide

the condition of rainfall and reservoir water level

changes, numerical analysis by using the software of Chang Jiang River
TGP Dam

SEEP/W of Geo-slope from Canada, which is based

on the non-saturated seepage theory were conducted.
In the analysis, the orthogonal optimizing method was
Qianjingping Landslide
employed to design the testing schemes. Back analysis
was carried out to get the seepage parameters based on
the observed rainfall and groundwater table from drill Figure 1. The location of Shuping landslide.

985
 500
NE0 and the stratum structure can be roughly classified into
(m)

400
5
T2b following layers: 1) plough horizon, 2) slope wash,
3) landslide accumulation; 4) sliding zone; 5) bedrock.
300
4
The typical geological profile is shown in Figure 2.
T2b
200
The water level of Three Gorges Reservoir was
135m
1
2
3
raised to 135 m in June, 2003. From October to
100
4

5 T2b
3 November in 2003, deformation was occurred at
2
T2b many places of Shuping landslide area. The observ-
100 200 300 400 500 600 700 800 900 1000 able deformations were mainly represented by ground
cracks and house cracking. New shear movement was
Figure 2. The typical geological profile of Shuping land- found at the side edge of the landslide mass (Figure 3).
slide. From February to March in 2004, large area of muddy
water appeared in the river near the shearing exit of
the landslide mass (Figure 4).

3 LAYOUT OF TEXT

Consider the obvious deformation occurred in Shup-

ing landslide area after the Three Gorges reservoir
impoundment, for observing the development of
landslide deformation and understanding the rela-
tionship between landslide deformation with rainfall,
reservoir water level and groundwater table as well
as the relationship between groundwater table with
rainfall, reservoir water level, Yichang Institute of
Geology and Minerals Resources under China Geo-
logical Investigation Bureau has conducted a series
Figure 3. Shear zone in the lateral edge of the slope. of monitoring works, including surface deformation,
deep displacement, ground temperature, groundwa-
ter table, rainfall and reservoir level. The following
is a brief introduction of groundwater and rain-
fall monitoring and the relationship between them[2] .
Then the monitoring data of groundwater table and
rainfall were used to conduct back analysis of the
unsaturated seepage parameters of Shuping landslide
mass.

3.1 Monitoring instruments and allocation

Groundwater level was measured by WS-1040, an
automatic dynamic groundwater monitoring device
made by the Institute of Hydrogeology and Engi-
neering Geology Technology under China Geologi-
cal Investigation Bureau. In order to understand the
impacts of rainfall on ground water table, a hydro-
Figure 4. Muddy river water at the exit of the landslide. logical observation hole was drilled at EL. 180 m in
the leading edge of the landslide. The observation
is automatic and the frequency of data acquisition is
The Shuping landslide mass is 700∼800 m long 2 hours.
and 700∼900 m wide. The elevation of landslide The rainfall observation is conducted by using N-
exit is 65∼68 m and the elevation of trailing edge is 68 automatic pluviometer, which is made in Japan. It
450∼500 m. Its thickness is 40∼70 m, the area is was installed on the roof of residents outside the land-
525,000 m2 and the volume is 26 million m3 . It slide area. From the observed data, the peak of rainfall
is a multi-period massive landslide with complicate all occurred between May to August. At the same
material composition. According to geological inves- period, the reservoir water level was kept at EL.135 m.
tigation and borehole data, the material composition So the observed data in July 2005 were selected for

986
 the parameters of unsaturated seepage can be obtained
by back analysis.[3–8]
precipitation(mm)

20 158

borehole water
15 156 4.1 Brief introduction of V-G model

level(m)
10 154 Based on measured soil water characteristic curve,
5 152 Van Genuchten presented the V-G model in 1980. The
formula of the model is:
0 150
7-1
7-5
7-9
7-13
7-17
7-21
7-25
7-29
θs − θr
θw = θr +   n m (1)
time(month-day) 1 + ψα
Figure 5. Groundwater table and rainfall in July, 2005.
Where, θ is the volumetric water content, θs is sat-
urated water content, θr is the residual water content,
analyzing the relationship between groundwater table and ψ = μa − μw is matric suction, α, n and m
and rainfall. are respectively nonlinear regression coefficients, here
m = 1–1/n.
From formula (1), based on the relationship
between unsaturated permeability function and the
3.2 The relationship between rainfall and soil water characteristic curve, Van Genuchten gave
groundwater table expression of unsaturated permeability as:
According to the data of rainfall and the groundwater
table from July 1 to July 31, 2005, the curve of ground-
ks {1−[α(μα −μw )]n−1 [1 + [α(μα −μw )n ]−m ]}2
water table and rainfall can be drawn as Figure 5. From k=
the figure, a continuous rainfall was occurred from [1+[α(μα −μw )]n ]m/2
July 7∼12. Two days after the rainfall, groundwater (2)
level in the drill hole was raised from El.151 m to
El.157 m. Then, the water level became stable. The Where, K is the permeability function of unsaturated
fact has shown that the rainfall has a significant impact soil, κs is the permeability coefficient of saturated soil.
on groundwater table in Shuping landslide area. Great The unsaturated parameters α and n will be gotten
changes of groundwater table will be happened under through back analysis.
the condition of concentrated rainfall in flood sea-
son. But it will lag behind the rainfall. Normally, the
ground water table will be raised two or three days Table 1. The permeability of saturated soil and rock mass in
after concentrated rainfall. Shuping landslide.

Sliding Sliding Sliding

Material mass 1 mass 2 zone Bedrock
4 BACK ANALYSIS OF THE UNSATURATED
SEEPAGE PARAMETER OF SHUPING Saturated 2.75 6.0 0.05 impermeable
LANDSLIDE permeability
(10−5 cm/s)
The coefficient of saturated and unsaturated perme-
ability of landslide mass is the necessary parameters
for conducting saturated and unsaturated unstable Table 2. Factor and level programs.
seepage analysis. The permeability of saturate soil can
be gotten from the conventional laboratory permeabil- Level Level Level
ity test (the results are shown in table 1). But due to Level program program 1 program 2 program 3
the complexity of experimental conditions, the unsat-
urated permeability of soil is rather difficult to obtain. Gliding mass 1 n1 2 2.5 3
Normally, the parameters are estimated through soil- α1 0.01 0.015 0.03
water characteristic curve. Based on the data of rainfall Gliding mass 2 n2 7 10 12
α2 0.02 0.025 0.03
and groundwater level of the observation drill hole, by Sliding zone n3 1 1.5 2
using the soil water characteristic curve V-G model and α3 0.005 0.01 0.015
the corresponding unsaturated permeability function,

987
 Table 3. Orthogonal experiment parameters.

No. n1 (A) α1 (B) n2 (C) α2 (D) n3 (E) α3 (F) /

1 1 (2.0) 1 (0.01) 1 (7) 1 (0.02) 1 (1) 1 (0.005) 1

2 1 2 (0.015) 2 (10) 2 (0.025) 2 (1.5) 2 (0.01) 2
3 1 3 (0.03) 3 (12) 3 (0.03) 3 (2) 3 (0.015) 3
4 2 (2.5) 1 1 2 2 3 3
5 2 2 2 3 3 1 1
6 2 3 3 1 1 2 2
7 3 (3.0) 1 2 1 3 2 3
8 3 2 3 2 1 3 1
9 3 3 1 3 2 1 2
10 1 1 3 3 2 2 1
11 1 2 1 1 3 3 2
12 1 3 2 2 1 1 3
13 2 1 2 3 1 3 2
14 2 2 3 1 2 1 3
15 2 3 1 2 3 2 1
16 3 1 3 2 3 1 2
17 3 2 1 3 1 2 3
18 3 3 2 1 2 3 1

4.2 Orthogonal optimizing test scheme

According to Ven-Genuchten model, the unsaturated
properties of each material can be represented by
the two parameters α and n. Considering the there
kinds of materials in Shuping Landslide, e.g. sliding
mass, sliding zone and bedrock, there are 6 para-
meters. By using the rainfall and groundwater table
data in July 2005, orthogonal optimizing method was
employed to design the test schemes, and the unsatu-
rated seepage computation was conducted for getting
the parameters[9–10] . By taking the reference of the Figure 6. Mesh of seepage analysis model.
empirical value of the similar landslides, the level pro-
gram for each group of parameters is shown in table 2.
157
According to the requirement of permutation and com- borehole water level
from monitoring(m)

156
bination of orthogonal table, only 18 tests are needed. simulation 2

The test arrangement is shown in Table 3. 155

borehole water level(m)

simulation 6

154
simulation 11

153
simulation 12

4.3 Back analysis of seepage 152

simulation 15

4.3.1 Model and boundary condition 151

simulation 17

The geological profile of the computation section is 150

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
shown in Figure 2. There were three kinds of material days

in seepage analysis model, e.g. gliding mass, sliding

zone and bedrock. The whole section was divided into Figure 7. Observed groundwater table and calculated
3,855 elements and 4,010 nodes (Figure 6). groundwater table.
For the numerical model, the boundary condition of
the left and right edge is water head boundary. The left
side water head takes the top elevation of the slope and 4.3.2 Computation results and the analysis
the right side water head is 135 m, which is the water According to the 18 groups of parameter listed in
level of Three Gorges reservoir. The upper boundary is Tab. 3, the variation of groundwater level of the
a flux boundary, which takes the real values of the rain- observation point in the landslide mass along with
fall infiltration during July 5 to July 25. The bottom the infiltration of rainfall was analyzed respectively.
of the model is an impermeable boundary. The computation results and the observed changes of

988
Table 4. Parameters from back analysis.

Material Gliding mass 1 Gliding mass 2 Sliding zone

Parameter n1 α1 n2 α 2 n3 α 3 135
Value 2.0 0.015 7 0.02 2 0.015 5

4
3
2
1
0 77
groundwater level are shown in Figure 7. By compar-
ing the computation results and the observed values,
the parameter of the 11th group was the best one. It
can be considered as the parameters from the back
analysis, which is shown in Tab. 4. Figure 8. Changes of groundwater table when reservoir
water level rose form EL.77 m to EL.135 m.
(Note: 0, 1, 2, 3, 4, 5 represent the groundwater table at the
initial time and the time after 12, 27, 40, 100, 200 days).
5 NUMERICAL SIMULATION OF SEEPAGE
FIELD UNDER DIFFERENT RESERVOIR

5.1 Model, boundary condition and initial

condition
The profile of the model was the same as Figure 6. The
left boundary is the water head with the elevation of 175
top of the slope. The right boundary is the actual reser- 7 5

2
3
voir water level. The bottom is impermeable boundary. 145

1
0
The reservoir level and observed groundwater table
of landslide mass before reservoir impoundment were
taken as the initial condition of the analysis.

5.2 The variation of the Three Gorges Reservoir

water level
Figure 9. Changes of groundwater table when reservoir
The actual variations of the reservoir water level were water level rose form EL.145 m to EL.175 m.
as follows: (Note: 0, 1, 2, 3, 5, 7 represent the groundwater table at the
initial time and the time after 20, 40, 60, 100, 200 days).
1. Initial impoundment period: From April, 2003, the
reservoir level was raised from 68 m to 135 m. From
EL.68 m to EL.77 m, it was natural impoundment. three typical conditions: First, reservoir water level
From 77 m to 135 m, it was forced impoundment, was raised from El.77 m to EL.135 m within 27 days,
when reservoir water level was rapidly raised in and then kept at EL.135 m until the stabilization of
27 days. After that, the reservoir water level was groundwater level. Second, the reservoir water level
kept at 135 m, and then slowly raised to EL.145 m. was raised from EL.145 m to EL.175 m within 40
2. Normal operation period (water level fluctuated days, and then kept at EL.175 m until the stabiliza-
between EL.145 m to EL.175 m): Every year, reser- tion of groundwater level. Third, reservoir water level
voir water level will be raised from El.145 m in was fallen from EL.175 m to EL.145 m within 110
June. During middle July to the end of August, the days, and then kept at 175 m until the stabilization of
reservoir water level fluctuate between EL.160 m to groundwater level.
EL.145 m by the situations of flood. In September,
the water level is kept stable at EL.145 m. From
October to early November, the water level was 5.3 Results of numerical simulation
raised to EL.175 m, and then kept this level until
beginning of next year. During January to the end According to the above three typical conditions,
of May of the next year, the water level falls slowly numerical simulations of groundwater seepage in land-
to EL.145 m again. slide area were carried out. The results were shown in
Figure8∼Figure10.
To simplify the calculation, the changes of the Three From the computation results, the groundwater
Gorges reservoir water level were generalized into table changes with the fluctuation of the water level.

989
 fluctuation of Three Gorges Reservoir water level,
it can be found that: the groundwater table varies
with the changes of the reservoir water level. But
the changes of groundwater table lag behind the
175 changes of reservoir water level. The faster of
the reservoir water level changes, the longer of the
7 5

2
3 time lag, vice versa.
145

1
0
Based on the above characteristics, it can be con-
cluded that: during the initial period of rapid rising
of reservoir water level, the rise of groundwater table
lag behind, water flow from reservoir into landslide
mass and the seepage force toward inside of the slope,
which will be in favor of the stability of landslide mass.
Figure 10. Changes of groundwater table when reservoir Whereas, during the initial period of reservoir level
water level rose form EL.175 m to EL.145 m. quickly drop down, as the water inside the slope has
(Note: 0, 1, 2, 3, 5 represent the groundwater table at the no time to drain, high values of hydraulic gradient will
initial time and the time after 50, 110, 150, 100, 200 days). be occurred in landslide mass, which will be unfavor-
able to the stability of the slope. That is main reason for
most of landslides failure occurred during the initial
period of reservoir water level drop down.
But the changes of groundwater table are much lag
behind for the changes of reservoir water level. The
faster of the reservoir water level changes, the longer REFERENCES
of the time for ground water table lag behind. Figure 8
and Figure 9 show that when reservoir water level Fa-Wu Wang, Ye-Ming Zhang & Zhi-Tao Huo. The July 14,
was raised to EL.77 m and EL.135 m, groundwater 2003 Qianjiangping landslide, Three Gorges Reservoir,
China. Landslides (2004) 1:157–162.
table are still far from the stable level. At this time,
Zhang Huawei. Real-time monitoring and stability forecast-
groundwater table is lower than reservoir water level. ing of Shuping Landslide in the Three Gorges Reservoir.
The seepage direction is from reservoir to the land- Three Gorges University master’s thesis, 2005.
slide mass (as shown in Figure 8). When reservoir Wu Mengxi & Gao Lianshi. Saturated-unsaturated soil
water level falls, the groundwater table in the land- unstable seepage numerical analysis. Water Conservancy
slide mass will also drop down. As the fall down speed Journal, 1999, (12): 38–42.
of reservoir water level is very slow, the change of Ye Weimin, Qian Lixin, Baiyun & Chen Bao. Forecasting
groundwater table in landslide mass is almost synchro- permeability coefficient of Shanghai saturated soft soil
nized with the reservoir water level. Only a little time by soil-water characteristic curve. Journal of Geotechnical
Engineering, 2005, 27 (11): 1262–1265.
lag occurred.
Zhang Hua, Chen Shanxiong & Chen Shouyi. Numer-
ical simulation of unsaturated soil infiltration. Rock
Mechanics, 2003, 24 (5): 715–718.
6 CONCLUSIONS Qi Guoqin & Huang Yunqiu, China. The general mathemat-
ical model of soil-water characteristic curve. Journal of
1. Based on the monitoring groundwater table with the Engineering Geology, 2004, 12(2) :182–186.
rainfall in July, 2005, by applying Van-Genuchten Zhang Yingke, Zan Huiping & Huang Yi. Unsaturated soil
model and using SEEP/W of GEO-SLOPE, back seepage function equation. Xi’an Science and Technology
analysis was conducted for getting the unsaturated University Journal, 2001, 17 (2): 174–177.
Song Xuejun, Chen Yu & Shi Jiying. The application of opti-
seepage parameters of Shuping landslide mass. The
mizing design on the super-network division. Journal of
parameters have provided sound support for the Harbin Industrial University, 2002, 34 (2): 265–269.
numerical simulation of groundwater seepage in Xue Maogen, Gu Mingtong & Lin Jieren. The fuzzy analy-
Shuping landslide area. sis of multi-objective optimization orthogonal design of
2. From the numerical simulation of the seepage field the ship main parameters. Boating Engineering, 1994,
of Suping landslide area under the condition of 4:20–27.

990
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Slope failure criterion: A modification based on strength

reduction technique

Y.G. Wang & R. Jing

Gansu Provincial Communications Planning, Survey & Design Institute, Lanzhou, China

W.Z. Ren
Key Laboratory of Rock and Soil Mechanics, Chinese Academy of Sciences, Wuhan, China

Z.C. Wang
Department of Civil Engineering, University of Calgary, Alberta, Canada

ABSTRACT: The FLAC method based on finite difference technique can simulate the behaviour of slope
geomaterials which undergo plastic flow when their yield limits are reached. Thus it can forecast the deformation
after slope failure, so it is widely used in slope stability analysis. This paper explored the exiting criteria of slope
failure and relevant limitations. Considering the fact that plastic shear strain catastrophe occurs when the points
in the critical slice change from a limit equilibrium state to plastic flow after yielding, this paper focused on the
changing rate of plastic shear strain of crucial points in the critical slice with the strength reduction coefficient
and made a modification for the slope failure criterion: when the catastrophe of the changing rate of plastic shear
strain occurs, the corresponding state is a critical instability state and the relevant strength reduction coefficient
is a critical factor of safety of the slope. The case analysis in this paper validated that the modified slope failure
criterion is rational and feasible.

1 INTRODUCTION the judgment of the expansion of plastic area and

shear deformation area according to these physical
The FLAC method based on finite difference tech- parameters is not so rational and accurate; the cru-
nique can simulate the behaviour of slope geomaterials cial point displacement method takes the changing
which undergo plastic flow when their yield limits are curve of strength reduction coefficients with crucial
reached, and thus forecast the deformation after slope points as analyzing object and takes strength reduc-
failure. Besides, the slope can have several yield sur- tion coefficient in the turning point of displacement
faces at the same time (Cala M., 2003), so it is widely changing rate as the indication of the beginning of
used in slope stability analysis. At present, there are slope failure (Ge, 2003). However, there are two prob-
three kinds of methods to judge slope failure in critical lems need to be solved: ① the displacement changing
slice: the first one is to judge slope failure accord- rate of the crucial points with strength reduction coef-
ing to deformation characteristics, for example, shear ficient may increase at any time, so how much change
strain method in broad sense, crucial point displace- of displacement can serve as a criterion is uncertain;
ment method, etc. (Lian, 2001, Zheng, 2002), the ② the changing curve of displacement and strength
second one is to judge according to slope stress dis- reduction coefficient can display different changing
tributing condition, for example, whether existing the effect due to the difference of greatest displacement,
run-through of plastic zone (Luan, 2003); the third one so how much reduction coefficient is proper in cal-
is to judge slope failure according to convergence con- culation is not clear; the run-through of plastic zone
dition in numerical calculation (Griffiths D.V. 1999, in slope is the necessary but not the sufficient con-
Zheng, 2002). Different researchers adopt different dition for slope failure. Besides, whether large and
methods and until now no unified criterion comes into infinite expanding deformation and displacement in
being. plastic zone occurs should also be taken into con-
There exist some problems more or less in the above sideration (Zhao, 2005). The geomaterials can still
three methods. In practice, broad sense shear strain have certain load-carrying capacity when it comes into
includes not only plastic strain, but elastic strain, so plastic condition, so simply taking the run-through of

991
plastic zone as sole judging criterion is not rational; Mohr-Coulomb strength criterion can be presented as
Taking the non-convergence condition in numerical follows:
calculation as judging criterion of critical instability
state of the slope is closely related with numerical cal- τf = c + σ tan φ (1)
culation technique, the quality of the model and the
convergence criterion, so the finite element calcula- So the factor of safety of this point on this prear-
tion may sometimes get into the control of minimum ranged shear plane can be presented as:
factor of safety in part of the slope. And the criterion
on the calculation of convergence also needs further τf c + σ tan φ
discussion. Fs = = (2)
τ τ
Considering the close relation between rock and soil
plastic failure and the occurrence, expansion and dis- Supposing that no shear failure occurs at this time,
tribution of plastic zone, and the ability of plastic shear the shear stress is equal to shear strength brought into
strain to memorize and describe the developing and play in practice, namely:
failure evolving process of plastic zone, this paper puts
forward the modified slope failure criterion to estimate τf c + σ tan φ
slope stability with finite difference numerical method τ= = = cl + σ tan φl (3)
based on strength reduction and taking changing rate of Fs Fs
plastic shear strain with strength reduction coefficient
Therefore, shear strength brought into play in prac-
in the run-through of plastic zone as research object.
tice equals to shear strength index after reduction.
Shear strength indexes after reduction are as follows:
2 THE DEFINITION OF SAFETY 
c tan φ
COEFFICIENT BASED ON STRENGTH cl = , φl = arctan (4)
REDUCTION Fs Fs

Strength reduction technique is a preferable way to In this sense, Fs can be considered as strength reduc-
solve the factor of safety of the slope with numerical tion coefficient, namely slope whole stable factor of
calculation technique and is more and more widely safety or strength reserve factor of safety.
applied at present. In 1975, Zienkiewicz and his fel-
lows first brought forward the definition of shear
strength reduction coefficient in geotechnical ela- 3 IMPROVED SLOPE FAILURE CRITERION
plastic finite element numerical analysis and hereby AND ITS ADVANTAGES
determined strength reserve factor of safety is consis-
tent with stable safety coefficient in definition which Based on the analysis of existing criterion of slope
is presented by Bishop in limit equilibrium methods failure, this paper combines shear strain methods and
(Zienkiewicz, 1975). Dawson’s research show that the stress condition methods and puts forward the mod-
factor of safety derived from numerical method based ified slope failure criterion: firstly, judging whether
on strength reduction technique is very close with the there exists the run-through of plastic zone, because
factor of safety from limit equilibrium methods which the run-through of the plastic zone is necessary condi-
more strictly meets the need of equilibrium condition tion of slope failure. Slope failure necessarily needs a
(Dawson, 1999∼2000). Moreover, Zheng (2004) fur- continuous sliding plane, and the run-through of plas-
ther reveals that the critical slices based on these two tic zone is the minimum shear strength area within
methods are very similar. the slope which provides conditions for the movement
Shear strength reduction coefficient can be defined of sliding mass. So this is the first concern when
as follows: under the condition of consistent outer judging whether the slope is in critical state; Sec-
load, the coefficient is the ratio of maximum shear ondly, taking the changing rate of plastic shear strain
strength inside slope to practical shear stress pro- of the crucial points on sliding plane with strength
duced by outer load within slope. The practical shear reduction coefficient as research object, when the
stress produced by outer load should be equal to the catastrophe of the changing rate of plastic shear strain
shear strength after reduction according to practical occurs, the corresponding state is critical instability
strength index. In finite difference numerical calcula- state of the slope and the relevant strength reduc-
tion based on strength reduction, for a certain point tion coefficient is the critical factor of safety of the
on the shear plane, supposing normal and shear stress slope.
are σ and τ , so according to the definition of fac- Suppose: (x , y ) are the points below and near the
tor of safety given by Bishop in limit equilibrium sliding plane; (x, y) are the monitoring points on the
method and considering the shear strength at this point, sliding plane; ux , uy , ux , uy are the displacements

992
of monitoring points respectively; t is the distance of crucial points as research object can fully con-
between the two points and meets the condition: sider anti-sliding potential after the run-through of
plastic zone.
|tx | ≤ 1, |ty | ≤ 1; 3. The selection of critical points on sliding plane
avoids some blindness. Because taking changing
Making |k| = | uy |/| ux |, ux , uy are the ratio of plastic shear strain after the rull-through of
increment of displacement respectively. plastic zone as research object, according to practi-
According to the definition of shear strain: cal condition of the run-through of plastic zone, one
or more critical points can be selected to monitor
 
∂ux ∂uy ux − ux uy − uy and judge.
εxy = + ≈ + 4. The existing method with the run-through of plastic
∂y ∂x ty tx zone as judging criteria is a specific example of this
ux − ux uy − uy ux uy paper. Namely, the condition of instant catastro-
= + = + phe of plastic shear strain of critical points after the
ty tx ty tx
 run-through of plastic zone, also the strength reduc-
1 k tion coefficient of the run-through of plastic zone
≈ + ux ≥ (1+k) ux (5) happens to be the corresponding reduction coeffi-
ty tx
cient when the changing ratio of plastic shear strain
appears turning point, hence the strength reduction
Because (x , y ) are the points below and near the slid-
coefficient at this time is factor of safety, and the
ing plane respectively, ux , uy are relatively small
run-through of plastic zone at this time means the
and can be neglected. From the equation (5), we can
appearance of critical state of slope failure.
clearly see that after the run-through of plastic zone,
considering the catastrophe of shear strain changing
rate of crucial points on sliding plane as the crite-
rion of slope failure is better than considering the 4 PROJECT ANALYSIS
catastrophe of displacement changing rate of crucial
points as the criterion, for the former catastrophe value 4.1 The creation of numerical calculation model
increases 1 + k times, which is more obvious and can This paper takes the example of HanJiaYa landslide in
be effectively avoid the gradual influence of displace- Xiangshi Expressway to analyze slope stability in nat-
ment changing rate as well. Therefore, the improved ural condition and its strain characteristics with finite
slope failure criterion is more practical. difference numerical method based on strength reduc-
Which needs explanation is that for general whole tion combined with above mentioned improved slope
slip, critical points on sliding plane are usually selected failure criterion.
on shear outlet of main section in landslide, and shear According to survey report, main section (I-I)
strain should be calculated from the beginning of is selected for analysis. In FLAC3D , plane strain
run-through of plastic zone. Compared with above model is created through controlling deformation of
mentioned slope failure criteria, the improved ones two-directions in front and back of the model. The
possess the following characteristics: calculation adopts Mohr-Coulomb yield criterion and
1. The improved slope failure criterion fully consider non-associated flow law. The angle of the dilatancy
the fact that there exists catastrophe of shear strain is regarded as zero. Numerical calculation model is
from limit equilibrium to shear flow after plastic presented by figure 1∼2.
yielding which greatly reduces the uncertainty of The landslide has all together 9322 units and
criterion. And there is no need to consider the 3188 nodes. The horizontal direction is presented as
proper reduced coefficient. With the increase of x, which is 420 m long, and the vertical direction is
reduction coefficient, as long as there is an obvi- presented as y, and the strike along slope is presented
ous turning of plastic shear strain changing ratio as z. The bottom of the model is fixed boundary, and
of crucial points, certain judgment can be made; other boundary is restricted by displacement of rele-
otherwise, in the process of incessant expansion vant directions. Physical and mechanical parameters
of plastic zone, the plastic strain and displacement are presented in table 1.
changing ratio usually appear several inflexion
points, which results in too much subjectivity in
4.2 Slope stability analysis under natural state
determining the factor of safety.
2. Taking the run-through of plastic zone as pre- According to the idea of ‘‘improved slope failure cri-
condition of judgment can meet the basic condi- terion’’, when the run-through of plastic zone appears
tions of slope failure; Meanwhile, judging critical in landslide, 1#:2# points on sliding plane round shear
state with the changing ratio of plastic shear strain outlet in rock mass 5 are selected as crucial points on

993
 Rock mass 1

Rock mass 2

Rock mass 3
Rock mass 6
1# monitoring point
Upper sliding plane 2# monitoring point
Rock mass 5
Potential sliding plane

Rock mass 7 Rock mass 4

Figure 1. Terrane distribution of main section I-I.

Figure 2. Terrane distribution of main section I-I.

Table 1. Mechanical parameters of rock layers.

Weight Cohesion Frictional angle Elasticity modulus Poisson ratio

Rock mass KN /m3 c/Kpa ϕ/◦ E/Gpa μ

Rock mass 1 21.2 18 9 0.006 0.2

Rock mass 2 23.2 70 20 2.4 0.245
Rock mass 3 29.3 100 22 6.87 0.261
Rock mass 4 29.7 150 25 17.6 0.283
Rock mass 5 22 50 20 3.32 0.175
Rock mass 6 25 70 21 8.35 0.185
Rock mass 7 27.3 100 22 12.2 0.21

sliding plane and the changing ratio of plastic shear plastic shear strain catastrophe will mean the instabil-
strain with strength reduction coefficient hereafter is ity of anti-sliding section at last and the occurrence of
analyzed to determine the relevant critical instability slope failure of greater scope. So these two points are
state and the factor of safety. Selecting the above two selected as crucial points in study.
points as monitoring points is based on the following Using shear strength parameters which correspond
consideration: Looking from the distribution of plastic to a series of strength reduction coefficients (the space
zone, nephogram of shear strain ratio and displace- between strength reduction coefficient is 0.01) to sim-
ment vector, the shear outlet is in front of landslide ulate landslide can obtain slope stress, strain and
and these two critical points are all on sliding plane, deformation results. Figure 3 is the vector diagram
especially 2# monitoring point. The main anti-sliding of displacement distribution which is not strength
section is also round the shear outlet of the slope. The reduced, namely in natural state.

994
 The figure shows that there is no large scale run- mass 5 in front of slope toe of steep bedding rock
through plastic zone in natural state. The displacement structure. At trailing edge forms tension crack and
between rock mass 7 in landslide trailing edge and the double-layer landslide creep slides down slowly along
area around interface of double-layer landslide is rel- with sliding plane. Under residual thrusting force of
atively large, so slow creep deformation occurs in the double-layer landslide, creep deformation of reverse
double-layer landslide which exerts certain thrusting rock-layer continuously occurs in rock mass 5, which
force to front rock mass 5 and yielding elements come results in continuous yielding failure. The slope keeps
into being in part of the area around the interface of on sliding down until the whole landslide forms large
front rock mass 5. This part of slope in front of the scale run-through of plastic zone, And the sliding
landslide become main anti-sliding area of the whole plane in rock mass 5 expands to full run-through.
landslide. Analysis of shear strain variation rate of two mon-
We can know from figure 4 that the sliding plane itoring points after run-through of plastic zone is
in under layer of the slope is in plastic state. In other presented in figure 6. The variation curve of plastic
words, at this very time the run-through of plastic zone shear strain with strength reduction coefficient shows
is formed and prerequisite for slope failure is provided; that when reduction coefficient reaches 1.20, relatively
Figure 5, vector digram of displacement distribution obvious turning points of plastic shear strain of criti-
shows that relatively large deformations mainly con- cal points appear, which can be considered as factor of
centrate on interface of landslide trailing edge and rock safety, namely in critical instability state. Thus poten-
tial anti-sliding bearing capacity is brought into full
consideration.
According to shear strain rate nephogram figure 7
and displacement vector figure 8 when slope is in crit-
ical instability state (FS = 1.20), whole rock mass
above the underlayer of sliding plane slides down and
residual thrusting force acts on rock mass 5 and results
in greater deformation. When landslide is in critical
instability state, the rock mass 5 in landslide front
has been in critical state of yielding failure. Seeing
from the variation curve of plastic shear strain of 2#
monitoring point around shear outlet, plastic strain
Figure 3. Vector diagram of displacement distribution when
strength reduction coefficient is 1.0. acutely increases at this time. It is resulted by the land-
slide changing from limit equilibrium state to shear
flow state after yielding. Great displacement sliding
failure occurs in landslide above under layer sliding
plane.
According to sliding plane confirmed by nephogram
of shear strain variation rate when the slope is in critical
instability state, limit equilibrium method is adopted
to calculate the stability of the landslide. The factor of
safety with Morgenstern-Price method is 1.143, which
is 4.75% different from calculation result with failure
criterion from this paper. It is obvious that the failure
Figure 4. Nephogram of slope shear strain rate when the criterion of this paper is rational and feasible and can
run-through of plastic zone is formed (Fs = 1.8). meet the needs of engineering projects.

5 CONCLUSION

Firstly, this paper deeply discusses the existing

slope failure criterion and corresponding problems in
numerical calculation and brings forward ‘‘improved
slope failure criterion’’: on the basis of the run-through
of plastic zone in slope mass, this paper focuses on
the variation rate of plastic shear strain of the cru-
cial points on sliding plane with strength reduction
Figure 5. Vector diagram of displacement distribution when coefficient. When the catastrophe of the changing rate
the run-through of plastic zone is formed (Fs = 1.8). of plastic shear strain occurs, the corresponding state

995
 strength reduction coefficient
1.17 1.18 1.19 1.2 1.21 1.22 1.23 1.24 1.25 1.26
0.00
5.00
10.00
Plastic shear strain

15.00
20.00
25.00
30.00 1# monitoring point/10^-3
35.00 2# monitoring point/10^-3
40.00 S.F
45.00

Figure 6. Variation curves of plastic shear strain of two monitoring points along with strength reduction coefficient.

equilibrium method. The result of such compar-

ison shows that the improved slope failure crite-
rion suggested by this paper is rational and feasi-
ble and can basically meet the needs of practical
cases.

REFERENCES

Cala M. & Flisiak J. 2003. Complex geology slope sta-

Figure 7. Nephogram of slope shear strain rate when bility analysis by shear strength reduction [A]. Flac
slope is in critical instability state (Fs = 1.2). and numerical modeling in geomechanics, Richard
Brummer & Patrick Andrieux, 99–102. A.A. Balkema.
Lian Zhen-ying. 2001. Studies on some problems of
three-dimensional FEM numerical analysis in founda-
tion pit engineering[D]. Dalian, Dalian University of
Technology.
Zheng Hong, Li Chun-guang, LI Zhuo-fen, et a1. 2002.
Finite element method for solving the factor of safety.
Chinese Journal of Geotechnical Engineering, 24
(4):323–328.
Luan Mao-tian, Wu Ya-jun & Nian Ting-kai. 2003.
A criterion for evaluating slope stability based on devel-
opment of plastic zone by shear strength reduction FEM.
Chinese Journal of Disaster Prevention and Mitigation
Engineering, 23 (2):1–8.
Figure 8. Vector diagram of displacement distribution Griffiths D.V. & Lane P.A. 1999. Slope stability analysis by
when slope is in critical instability state (Fs = 1.2). finite elements. Geotechnique, 49 (2):387–403.
Zhang Ying-ren, Zhao Shang-yi & Zhang Lu-yu. 2002. Slope
stability analysis by strength reduction FEM. Chinese
is critical instability state and the relevant strength Journal of Engineering Science, 4 (10):57–61.
reduction coefficient is critical factor of safety of Ge Xiu-run, Ren Jian-xi, Li Chun-guang, et a1. 2003.
3D-FEM analysis of deep sliding stability of 3# dam
the slope. foundation of left power house of the Three Gorges
Secondly, According to ‘‘improved slope fail- Project. Chinese Journal of Geotechnical Engineering,
ure criterion’’, this paper determines landslide crit- 25 (3):389–394.
ical instability state and its stable factor of safety Zhao Shang-yi, Zheng Ying-ren & Zhang Yu-fang. 2005.
and compares with the calculation results with limit Study on slope failure criterion in strength reduction

996
 finite element method. Chinese Journal of Rock and Soil Dawson E., You K. & Park Y. 2000. Strength-reduction
Mechanics, 26 (2): 332–336. stability analysis of rock slopes using the Hoek-Brown
Zienkiewicz O.C., Humpheson C & Lewis R.W. 1975. Asso- failure criterion[A] In: Labuz J.F. ed. Geotechnical Spe-
ciated and Non-Associated Visco-Plasticity and Plasticity cial Publication: Trends in Rock Mechanics[C]. [s.l.],
in Soil Mechanics. Geotechnique, 25 (4):671–689. [s.n.], 65–77.
Matsui T. & San K.C. 1992. Finite element slope stability Zheng Hong, Liu De-fu & Luo Xian-qi. 2004. Determina-
analysis by shear strength reduction technique. Soils and tion of potential slide line of slopes based on deforma-
Foundations, 32 (1):59–70. tion analysis. Chinese Journal of Rock Mechanics and
Dawson E.M., Roth W.H. & Drescher A. 1999. Slope sta- Engineering, 23 (6):708–716.
bility analysis by strength reduction, Geotechnique, 49 Wang Yong-gang. 2006. The gradual destruction mechan-
(6):835–840. ical model and time-dependent deformation analysis of
Dawson E., Motamed F., Nesarajah S. et a1. 2000. Geotech- landslide with double sliding planes and reverse rock-
nical stability analysis by strength reduction[A]. In: Gri- layer[D]. Wuhan, Institute of rock and soil mechanics,
fiths D.V. ed. Geotechnical Special Publication: Slope Chinese academy of sciences.
Stability 2000-Proceedings of Sessions of Geo-Denver
2000[C].[s.l.], [s.n.], 99–113.

997
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Unsaturated seepage analysis for a reservoir landslide during impounding

J.B. Wei & J.H. Deng

State Key Laboratory of Hydraulics and Mountain River Engineering, College of Water Resources and Hydropower,
Sichuan University, Chengdu, China

L.G. Tham & C.F. Lee

Department of Civil Engineering, The University of Hong Kong, Hong Kong, China

ABSTRACT: Xietan landslide is a pre-existing landslide in the Three Gorges Reservoir area and is constituted
of the slide body, the slip zone, the sliding-disturbed zone, and the bedrock from up to down. In hydrogeology, the
slide body, the slip zone and the sliding-disturbed zone can be regarded as porous continuous medium, while the
permeable bedrock is fractured rock. In this paper, an equivalent continuum model was used to model all strata,
and a finite element model was established to analyze the unsaturated seepage of Xietan landslide. Then the back-
propagation neural network was used to substitute finite element seepage analysis. The saturated permeability
coefficient of every stratum was optimized by genetic algorithm based on monitored data during reservoir
impounding. The saturated permeability coefficient of the slide body by back analysis is 4.89 × 10−2 cm/s,
which is consistent with field testing results (1.78 × 10−2 cm/s to 3.2 × 10−2 cm/s); the saturated permeability
coefficient of the slip zone by back analysis is 4.66 × 10−5 cm/s, which is very larger than indoor testing results
(2.74 × 10−7 cm/s to 5.73 × 10−7 cm/s). The scale effect of soil samples was reflected by this difference.

1 INTRODUCTION function of unsaturated soil. The soil-water charac-

teristic curve is usually obtained by tests or empirical
For reservoir landslide, the temporal and spatial distri- procedures, and the unsaturated permeability function
bution of pore water pressure in the landslide with the can be predicted by using the saturated permeabil-
change of reservoir water level is of great important to ity coefficient and the soil-water characteristic curve.
the displacement and stability of landslide. At present, However, the result of seepage calculation using test
phreatic surface is usually as the base of stability anal- parameters usually differ with the measurement data
ysis of reservoir landslide, but the methods adopted for because the test parameters are affected by sample
determining the ground water tables are mostly empiri- location, test method and scale effect. Though the back
cal and varied from one project to another (Zheng et al. analysis method is an effective way to determinate the
2004). Along with the development of unsaturated parameters of geotechnical materials, it is mostly used
seepage theory, unsaturated seepage analysis method in the displacement back analysis of slope or tunnel,
has been widely used in the seepage flow analysis of the application of back analysis method on seepage
landslide (Peng et al. 2002; Zhang et al. 2003), but analysis of landslide is not so common.
related researches mostly assume that the slide body Take the Xietan landslide in the Three Gorges
of landslide is permeable and the slip zone and below Reservoir area as an example, the finite element model
of landslide is impermeable. The research of pore of unsaturated seepage of Xietan landslide is estab-
water pressure of landslide with complex hydroge- lished in the first, and then the saturated permeability
ological structure under the condition of reservoir coefficient of every stratum is obtained by back anal-
fluctuations is not considered in many jobs, and the ysis method using BP network and genetic algorithm
comparative analysis on measurement and calculation based on monitoring data during impounding, at last
is few also. the obtained permeability coefficients are used in the
The primary work of unsaturated seepage of finite element model to model the change of ground-
landslide is determining the unsaturated hydraulic water level during impounding. Through comparing
parameters of every stratum, which include the soil- the results of numerical modeling and measurement,
water characteristic curve (SWCC) and permeability the feasibility of above method is discussed.

999
2 ANALYSING THEORY can find the potential mapping relation between input
and output values.
2.1 Unsaturated seepage modeling BP network is constituted by one input layer (X ),
one output layer (Y ) and one or more hidden layer
Based on Darcy’s law and continuity principle of mass
(Z). Each layer contains several nodes, and the nodes
conservation, the governing differential equation of
are interconnected via weighted links, as shown in
two-dimensional unsaturated seepage can be derived
Figure 1 (Rogers, 1997). The training process of BP
as below (Lei, 1988):
network contains forward propagate of input values,
  backward propagate of errors and adjust of link val-
∂ ∂H ∂ ∂H ∂θ ues. Input values propagate forward to the hidden
kx + ky +Q = (1)
∂x ∂x ∂y ∂y ∂t layer nodes via weighted links, and the input values
in hidden layer transform into output values through
where H = pressure head; kx = permeability coeffi- transfer function and propagate forward to the output
cient in the x-direction; ky = permeability coefficient layer via weighted links. The errors between desired
in the y-direction; Q = applied boundary flux; θ = outputs and actual outputs are propagated backward
volumetric water content; t = time. through original interconnected links. Then the link
The main hydraulic parameters needed in unsatu- values are adjusted according to given learning rate
rated seepage analysis include the soil-water character- (α) and momentum term (β), and the output values
istic curve and permeability function. In this paper, the in output layer are adjusted. When the error values
soil-water characteristic curve of every stratum refers between desired outputs and actual outputs of all pat-
to Fredlund empirical procedure (Fredlund & Xing, terns are less than given tolerance (ε), the training of
1994a), as written below: BP network is finished.
The process of BP network simulating FEM unsat-

ln(1 + ψ/ψr ) urated seepage is as follows. At first, a BP network
θ = 1− should be constructed. The number of nodes (n) in the
ln(1 + 1000000/ψr )
input layer is the same as the number of parameters
θs to be solved, the number of nodes (m) in the output
× (2)
{ln [e + (ψ/a)n ]}m layer is the same as the number of monitored ground
water level, the number of nodes (p) in the hidden layer
where θ = volumetric water content; θs = saturated can be specified either manually or by an optimiza-
volumetric water content; ψ = suction; ψr = the tion method. Usually, only one hidden layer is needed.
suction corresponding to the residual water content; Secondly, a number of training patterns should be
a = air-entry value of the soil; n and m are curve created to train the network. The input valves of train-
fitting parameters. ing patterns can be created by random method or by
The unsaturated permeability function is predicted orthogonal method, the desired output values of train-
by using the saturated permeability coefficient and ing patterns are the corresponding calculated results
the soil-water characteristic curve (Fredlund et al.,
1994b):
 7 Y1 ... Yk ... Ym
b
θ (ey ) − θ (ψ)  y

k(ψ) = ks θ e dy
ln(ψ) ey w w w
w w j1 w p1 1k jk pk w 1m w jm w

11 pm
b
θ (ey ) − θs  y

θ e dy (3)
ln(ψaev ) ey

where k(ψ) = calculated permeability coefficient for

a specified suction ψ; ks = saturated permeability
Z1 ... Zj ... Zp
coefficient; b = ln (1000000); ψaev = air-entry value v v v
of the soil; θ  = derivative of formula (1). v 11 v
i1 v n1 1j ij nj v 1p v
ip vnp

2.2 BP network
BP network (back-propagation neural network) is
one of the most widely used artificial neural net- X1 ... Xi ... Xn
works because of its powerful function-mapping abil-
ity (Deng et al. 2001). Through training, BP network Figure 1. Structure of BP network.

1000
 t=0 clay with coarse gravels of sub-rounded shape. Its
permeability is relatively good. The bottom layer
is bedrock; the upper part of bedrock is relatively
Coding initial generation P(t)
permeable because of fracture.
A 2D finite element software SEEP/W is used
Fitness evaluation of P(t) for the unsaturated seepage analysis. In the seepage
model, sliding-disturbed zone and permeable bedrock
Yes are combined into one stratum, namely slip bed, and
t=T? Output equivalent continuum model is used to model all strata.
No Monitoring data shows that the groundwater level in
t=t+1 the rear of the landslide is not affected by the reservoir
water level during impounding period, so the calcu-
lating range is limited in the middle-front part of the
New generation P(t) by genetic operators landslide, as illustrated in Figure 3.
The monitored groundwater level before filling is
Figure 2. Flow chart of genetic algorithm. applied as initial condition and the monitored reservoir
level during filling is applied as boundary condition
in seepage analysis. The calculation steps and corre-
by FEM. Lastly, a number of testing patterns should sponding reservoir level and monitored borehole water
be created to test the mapping ability of trained net- level in ZK9 is listed in Table 1.
work. The construction of testing pattern is the same The input parameters are the soil-water character-
as training pattern. istic curve, saturated permeability coefficient and the
permeability function. In this paper, the saturated per-
meability coefficient is as an undetermined parameter,
2.3 Genetic algorithm the soil-water characteristic curve and the permeabil-
Genetic algorithm is an optimization algorithm that ity function is obtained by empirical formula (2) and
simulated the natural selection and natural genetics,
and codes optimization variables as a binary string.
The process of genetic algorithm is showed in Figure 2
(Goldberg, 1989). The initial generation is created by
random method. Further generations are created by
repeated application of genetic operations (reproduc-
tion, crossover and mutation), which favor the survival
of the fitness of old generations. When the time of
reproduction reached a given number (T ), the repro-
duction is stopped, and the individual with highest
fitness is the final results.
Figure 3. Unsaturated seepage model of Xietan landslide.
3 PERMEABILITY COEFFICIENT BACK
ANALYSIS OF XIETAN LANDSLIDE
Table 1. Steps of unsaturated seepage modeling and mon-
itoring data.
3.1 Unsaturated seepage modeling
Xietan landslide is a preexisting landslide that locates Reservoir level Water level
in the reservoir area of the Three Gorges Project, Step Time (m) in ZK9 (m)
48.2 km upstream of the Three Gorges dam. The strata
0 2003–5-24 19:00 80.90 80.60
of Xietan landslide can be divided into four layers 1 2003–5-25 7:00 82.43 80.86
from up to down. The first layer is slide body, which 2 2003–5-26 7:00 85.04 82.36
is mainly composed of silt and gravels of siltstone, 3 2003–5-27 7:00 88.10 84.23
limestone, and quartz sandstone or shale origin. It 4 2003–5-28 7:00 91.26 87.36
is loosely structured and quite permeable. The sec- 5 2003–5-29 7:00 94.48 91.25
ond layer is slip zone, which is around 2–3 m thick 6 2003–5-30 7:00 97.82 95.33
and composed of deep grey or light green clay con- 7 2003–6-1 7:00 105.92 105.66
taining coarse gravels. Its permeability is lower, so 8 2003–6-2 7:00 109.75 109.65
9 2003–6-3 7:00 113.06 113.02
it’s a relative impermeable stratum. The third layer
10 2003–6-4 7:00 115.60 115.61
is sliding-disturbed zone, which is composed of firm

1001
(3) respectively. The parameters of the soil-water char- testing patterns are used to test the mapping ability
acteristic curve are obtained by laboratory and field of the network. The maximum error of output values
testing results and are listed in Table 2. between FEM and the trained BP network is 0.13 m,
so the mapping ability of BP network is demonstrated.

3.2 BP network modeling of unsaturated seepage

3.3 Genetic algorithm optimizing of permeability
Based on tested and provisional calculated results, the coefficient
range of saturated permeability coefficient of every
stratum used in BP network is, slide body (3 × 10−2 − In genetic algorithm, the probability of an individual
8×10−2 cm/s), slip zone (2×10−5 −7×10−5 cm/s) and reproducing to new generation depends on its fitness.
slip bed (5×10−4 −1×10−2 cm). A total of 12 training Fitness is transferred from the value of objective func-
patterns and 4 testing patterns are created. The input tion. The smaller the value of objective function is,
values of training patterns and testing patterns are the the higher the fitness. The objective function in this
saturated permeability coefficient of every stratum as paper is:
illustrated in Table 3. The corresponding output values
are the stepped groundwater level in the location ZK9 
n

calculated by finite element method. min f = (Hi − hi )2 (4)

i=1
A single hidden layer BP network is used; the num-
ber of neurons in the input layer, hidden layer and
where Hi = the groundwater level by monitor or by
output layer is 3, 7 and 10 respectively. The basic
FEM; hi = the groundwater level by BP network.
parameters of BP network are, α = 0.6, β = 0.5,
The basic parameters of genetic algorithm are:
ε = 0.03. After the network has been trained, the
crossover probability = 0.6, mutation probability =
0.05, time of reproductions = 2000. At first, the
Table 2. Parameters of SWCC. calculated data by FEM of testing patterns are used
to inversely predict the saturated permeability coeffi-
Stratum θs a (kPa) n m cient of every stratum through genetic algorithm. The
results of inversive prediction are listed in Table 4.
Slide body 0.28 100 0.4 0.25 The comparison of saturated permeability coefficient
Slip zone 0.31 1000 0.5 0.7 between inversive prediction values and actual values
Slip bed 0.25 800 0.3 0.5 listed in Table 3 shows that the error of slide body and
slip zone is very little, while the error of slip bed is
slightly large. This is because the change range of sat-
Table 3. Input parameters of training patterns and testing urated permeability coefficient of slip bed in training
patterns. patterns is larger than the other strata. Then the mon-
itoring data of ZK9 during filling period are used to
Permeability coefficient (cm·s−1 ) predict the saturated permeability coefficient of every
Pattern
no. stratum. The results are listed in Table 4 also.
Slide body Slip zone Slip bed The inversion parameters are input in the FEM
Training model to model the seepage of Xietan landslide dur-
patterns 1 5 × 10−2 4 × 10−5 1 × 10−2 ing impounding period. The calculated and monitored
2 5 × 10−2 4 × 10−5 1 × 10−3 groundwater level is compared in Figure 4. The two
3 8 × 10−2 4 × 10−5 1 × 10−3 curves coincide very well.
4 3 × 10−2 4 × 10−5 5 × 10−4
5 8 × 10−2 7 × 10−5 1 × 10−3
6 3 × 10−2 7 × 10−5 5 × 10−4 4 COMPARISION OF PERMEABILITY
7 3 × 10−2 7 × 10−5 1 × 10−2 COEFFICIENT BETWEEN BACK ANALYSIS
8 5 × 10−2 7 × 10−5 5 × 10−4 AND TESTS
9 3 × 10−2 2 × 10−5 1 × 10−3
10 5 × 10−2 2 × 10−5 1 × 10−2 The inversion permeability coefficient of slide body is
11 5 × 10−2 2 × 10−5 1 × 10−3 4.89 × 10−2 cm, while the field testing result is 1.78 ×
12 8 × 10−2 2 × 10−5 5 × 10−4 10−2 − 3.2 × 10−2 cm (Wei et al. 2007). The inversion
Testing and testing values are in the same magnitude. The field
patterns 1 8 × 10−2 2 × 10−5 1 × 10−3 test method is double ring infiltration test. Big gravels
2 5 × 10−2 4 × 10−5 5 × 10−4
(above 200 mm) in slide body are avoided in field
3 8 × 10−2 4 × 10−5 5 × 10−4
test because of the size limitation of test instrument.
4 5 × 10−2 7 × 10−5 1 × 10−3
Because the spatial change of permeability coefficient

1002
 Table 4. Back analysis results of permeability coefficient.

Permeability coefficient (cm·s−1 )

Testing min f
pattern no. Slide body Slip zone Slip bed (m2 )

1 8.09 × 10−2 2.13 × 10−5 1.17 × 10−3 0.0377

2 5.09 × 10−2 3.95 × 10−5 5.63 × 10−4 0.098
3 8.09 × 10−2 4.0 × 10−5 3.33 × 10−4 0.087
4 5.01 × 10−2 7.01 × 10−5 8.98 × 10−4 0.178
monitored
water level 4.89 × 10−2 4.66 × 10−5 5.46 × 10−4 2.6

5 CONCLUSIONS

1. The infiltrative strata of Xietan landslide are

reduced to three strata such as slide body, slip zone
and slip bed through combining sliding-disturbed
zone and permeable bedrock into one stratum. Then
equivalent continuum model is used to model every
stratum, and finite element model of unsaturated
seepage analysis of Xietan landslide is established.
2. The application of BP network and genetic algo-
rithm in the back analysis of the hydraulic parame-
ters of landslide is discussed, and the saturated per-
meability coefficient of every stratum is calculated
Figure 4. Comparison between calculated and monitored by back analysis based on monitored data.
water level in borehole ZK9. 3. The inversion parameters are used in the FEM
model to model the seepage of Xietan landslide
during impounding period. The calculated result is
is taken into consideration, the inversion permeability agreement with monitored data.
coefficient can reflect the permeability characteristics 4. The inversion permeability coefficient of slip zone
more exactly. is great larger than laboratory tested result because
The inversion permeability coefficient of slip zone of scale effect. In this paper, the inversion para-
is 4.66 × 10−5 cm, while the lab testing result is meter can reflect the synthetic hydraulic character-
2.74×10−7 −5.73×10−7 cm. There is a big difference istic of the landslide more exactly.
between inversion value and testing value. The main
reasons of this difference lie in two aspects. At first,
remolded samples are used in laboratory permeabil- ACKNOWLEDGEMENTS
ity test. The sample size is 62.8 mm in diameter and
20 mm in height and the gravels above 2 mm are elim- Financial support from Chinese Postdoctoral
inated from the specimens. Scale effect is exists in the Science Foundation under grant 20070410387, the
test. While inversion value is the synthetic permeabil- Special Fund for Major State Basic Research Project
ity coefficient of the total slip zone, the non-uniform under grant 2002CB412702 and Research Grant
thickness and discontinuity of slip zone are reflected in Council of Hong Kong under grant HKU7015/02ERGC
the inversion value, so the inversion permeability coef- are greatly appreciated.
ficient of slip zone is larger than test results. And then,
due to the limitation of unsaturated seepage modeling
method, the fracture seepage in permeable bedrock is
modeled by equivalent continuum model, which has a REFERENCES
certain impact on the calculated results.
Though the unsaturated seepage modeling method Deng, J.H. & Lee, C.F. 2001. Displacement back analysis
for a steep slop in the Three Gorges Project site. Interna-
of landslide with complicated hydraulic structure is tional Joural of Rock Mechanics and Mining Sciences 38:
not perfect now, adopting the inversion parameters and 259–268.
seepage analysis method suggested in this paper, the Fredlund, D.G. & Xing, A. 1994a. Equations for the
calculated result is agreement with monitored result, soil-water characteristic curve. Canadian Geotechnical
so the feasibility of above method is demonstrated. Journal 31: 521–532.

1003
Fredlund, D.G., Xing, A. & Huang S. 1994b. Predicting Rogers, J. 1997. Object-Oriented Neural Networks in C++.
the permeability function for unsaturated soils using the London: Academic Press, Inc.
soil-water characteristic curve. Canadian Geotechnical Wei, J.B., Deng, J.H., Tham, L.G. & Lee, C.F. 2007. Field
Journal 31: 533–546. tests of saturated and unsaturated hydraulic parameters of
Goldberg, D.G. 1989. Genetic algorithms in search, opti- gravelly soil in Xietan landslide. Rock and Soil Mechanics
mization and machine learning. Reading, Mass: Addison- 28: 327–330 (in Chinese).
Wesley. Zhang, P.W., Liu, D.F., Huang, D.H. & Song, Y.P. 2003.
Lei, Z.D. 1988. Soil hydrodynamics. Beijing: T singhua Saturated-unsaturated unsteady seepage flow numerical
Press. simulation. Rock and Soil Mechanics, 24: 927–930 (in
Peng, H., Chen, S.F. & Chen, S.H. 2002. Analysis on unsat- Chinese).
urated seepage and optimization of seepage control for Zheng, Y.L., Shi, W.M. & Kong, W.X. 2004. Calculation
Dayantang landslide in Shuibuya Project. Chinese Jour- of seepage forces and phreatic surface under drawdown
nal of Rock Mechanics and Engineering, 21: 1027–1033 conditions. Chinese Journal of Rock Mechanics and
(in Chinese). Engineering 23: 3203–3210 (in Chinese).

1004
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

A simple compaction control method for slope construction

L.D. Wesley
University of Auckland, Auckland, New Zealand

ABSTRACT: Controlling the quality of compacted fill using dry density and water content can be difficult
for the following reasons. Firstly, some soils are highly variable, with rapid and random changes in optimum
water content and maximum dry density. Secondly, some soils do not show peak dry density and optimum water
content. Thirdly, some soils are highly sensitive and conventional compaction control may be inappropriate. The
above observations are particularly true of residual soils, whose formation processes produce quite different
properties from sedimentary soils. An alternative compaction control method is presented, making use of shear
strength and air voids in place of water content and dry density. By using these parameters, variations in soil
properties are taken account of, and the same specification applies regardless of these variations.

1 INTRODUCTION Compaction Test is normally very close to the Plastic

Limit.
Compaction of clays and silts has normally been con- This procedure is satisfactory with many soils, but
trolled by specifying limits of water content and dry certainly not all. Among ‘‘problematic’’ soil charac-
density. While this method has had almost universal teristics that make the application of this traditional
acceptance, it is not without deficiencies, and attempts method difficult are the following.
have been made over the years to develop alterna-
tive methods. Some of the difficulties in using the a. Many soils, especially residual soils, are far from
conventional method are described in the following homogeneous; their properties change rapidly over
sections, and an alternative method is described based
on undrained shear strength and air voids.
1.6
av
av

Ze
=
=

ro
5%
10

ai
%

rv

2 LIMITATIONS OF NORMAL COMPACTION

oi
ds

1.5 (modified)
(a v

CONTROL METHOD D
)l
in
Dry density (tonne/m3 = gm/cm3 )

e
(S
r
=

2.1 Conventional compaction control

10
0%

1.4
)

Conventional behaviour and compaction tests are well Modified

known and only brief comments will be made here. D (standard)
Typical results of compaction tests on a clay are shown
in Figure 1. Also plotted on the graph is the zero 1.3

air voids line; this indicates the maximum value the

dry density can have at any particular water con- Standard
tent. An important point to recognise is that there is 1.2
Optimum w/c (standard)
Optimum w/c (modified)

no such thing as a unique ‘‘optimum water content’’

for a particular soil. Almost any value of optimum
water content is possible; it simply depends on the 1.1
compactive effort involved in the compaction pro-
cess. In Figure 1, the optimum water content from
the Modified Compaction Test is about 8% below
1.0
that for the Standard Compaction Test. This is nor- 18 20 25 30 35 40 42
mal for moderate to high plasticity clays. It is worth Water content (%)
noting also that for moderate to high plasticity clays
the optimum water content from the Standard Proctor Figure 1. Typical compaction tests on clay.

1005
 short distances, so that the optimum water content
may vary over a wide range. Steel mill site:
1.3 Weathered basalt and ashes
b. Standard compaction tests on some residual soils do
not show a peak dry density and thus no optimum
water content.
c. Many residual soils are highly structured, and have 1.2

natural water contents close to, or even above

Dry Density (gm/cm3)

Ze
their Liquid Limit. Compaction processes destroy ro
ai
the natural structure of these soils, and may turn rv
oi
1.1 ds
them into very soft materials, making compaction lin
e
impossible without substantial drying of the soils.
These aspects are described in more detail in the 1.0
following sections.

0.9
2.2 Highly variable soils
Figure 2 shows the results of compaction tests on
soils from two relatively small construction sites in the 40 50 60 70
Water content (%)
Auckland region of New Zealand. The soils involved
at each site were of the same geological origin. The
upper diagram shows tests on soils derived from the Industrial site:
1.8
weathering of basaltic lava flows and ash deposits. Pleistocene deposits
The lower diagram shows tests on soils derived from
Pleistocene deposits, some of which may have under-
gone weathering since deposition. Despite the com-
mon geological origin at each site, there is clearly a
Dry Density (gm/cm3)

very large variation in the type of soil, as reflected in 1.4

the compaction curves shown in Figure 2. Determi- Ze
nation of the optimum water content in this situation ro
air
is rather meaningless, as it will only be valid for the vo
ids
lin
particular location from which the sample was taken. e
1.0

2.3 Soils without clear optimum water contents

There are some unusual soils that do not conform
to the behaviour illustrated in Figure 1. Their com- 0.6
paction curves do not show clear peaks of dry density
and thus do not indicate optimum water contents. The 20 40 60 80
Water content (%)
best known soil type that displays this behaviour are
clays that contain a high proportion of the clay min-
Figure 2. Compaction test results from two small construc-
eral allophone; this is a very unusual clay mineral tion sites (after Pickens, 1980).
derived from the weathering of volcanic ash. The non-
crystalline nature of volcanic ash results in a very
different weathering process to that which produces
‘‘conventional’’ clay minerals, and produces a poorly Figure 3 shows typical results of compaction tests
crystalline, very fine grained mineral. Allophane on two clays having high allophane contents. Three
clays are characterised by generally good engineering different procedures have been followed in obtaining
properties despite extremely high water contents and these curves. The curves labelled ‘‘natural’’ have been
Atterberg limits. Water contents between 100% and obtained by progressively drying the soil from its nat-
200% are common. Another of their distinctive char- ural state. The other curves have been obtained by
acteristics is that drying causes irreversible changes re-wetting the soil after it has been both air dried or
to their properties. In their natural state they are very oven dried. Separate samples were used for each point
fine-grained, and are of moderate plasticity. When air on the curves, so that the results do not reflect any
or oven dried, their plasticity is greatly reduced. In influence from repeated compaction.
many cases, oven drying changes them into non-plastic In the case of the second sample, an additional curve
slightly sandy silts. was obtained by only partially air-drying the soil, to a

1006
 1.2 destroy the bonds or other structural influences present
Natural
Air dried and thus weaken the soil.
Dry density g/cm3

1.0 Oven dried

Figure 4 illustrates the softening effect that com-
Ze
ro paction has on a number of volcanic ash soils in Japan.
0.8 air
v oid
s Tests have been done on a range of volcanic ash soils
0.6 in which strength measurements have been made after
the soils have been compacted using varying com-
0.4
pactive efforts. The strength has been measured using
20 40 60 80 100
Water content (%)
120 140 160 180 200 a cone penetrometer test, giving the ‘‘cone index’’ as
1.2 an indication of strength. The compactive effort has
Dry density g/cm3

Natural
Air dried
been varied by changing the number of blows of the
1.0 Oven dried
Air dried to 65 %
compaction rammer, from a minimum of 5 blows, a
Ze maximum of 120 blows. It is seen that most of the sam-
ro
0.8 air
v oid ples become softer as the blow count increases. Kanto
s
0.6
loam Sample A shows a consistent strength increase
until the blow count reaches about 70, beyond which
0.4 a decrease in strength occurs. Kanto loam Samples B
20 40 60 80 100 120 140 160 180 200
Water content (%) and C show a small initial increase in strength followed
by a large loss of strength. This softening effect pro-
Figure 3. Compaction tests from allophone clays. duced by higher compactive effort is often referred to
as ‘‘over-compaction’’.
For sensitive or structured soils at their natural water
water content of only 65%. The number of points on content, there is thus an ‘‘optimum compactive effort’’
these curves indicates that a large volume of soil was which will produce the strongest fill, as indicated in
required to carry out the tests. For a full description of Figure 4. Kanto loam Samples A, B and C show rea-
the properties of allophone clays see Wesley (2002). sonably clear ‘‘optimum compactive efforts’’ while the
remainder show only that the optimum compactive
effort is the minimum required to produce a tight fill.
2.4 Soils that soften during compaction
Some soils become softer during the compaction pro-
cess. Many soils, especially residual soils, are ‘‘struc- 16
tured’’, that is they have some form of weak bonds Arrows indicate “optimum
compactive effort” w = 110%
between their particles. In some cases, they consist A
of partially weathered parent material, which makes
them appear competent, but which is really a fragile,
sensitive material. When remoulded the structure is 12
broken down and the soil becomes softer. It is impor- Solid lines are
various Kanto loams
tant to recognise, therefore, that compaction of a soil
Kanuma soil
may have two effects, as follows: w = 220%
Cone Index qc

w
a. ‘‘Densifying’’ the soil, that is pressing the particles =1
21
%
closer together and squeezing out any air trapped 8
between the particles.
b. Remoulding the soil, causing it to soften. Nearly all B
natural soils lose some strength when remoulded,
and compaction is a form of remoulding. The w=
117
compaction process destroys the natural structure %
4
of the soil, especially any bonds between parti- C
w = 108%
cles. This is usually accompanied by release of
water trapped within the structure, or between the D
Volcanic w = 109%
particles, adding to the softening process. ash soil E
w = 59%
With non-structured soils, which are those of low or
negligible sensitivity, the effect of compaction will
0 20 40 60 80 100 120
be purely to force the particles closer together, and Number of Rammer Blows
thus increase the density and strength of the soil. With
highly structured soils, which are normally those of Figure 4. Influence of compactive effort on volcanic ash
high sensitivity, the effect of compaction will be to soils (after Kuno et al., 1978).

1007
3 ALTERNATIVE COMPACTION CONTROL properties to those obtained with conventional con-
USING SHEAR STRENGTH AND AIR VOIDS trol methods, specifying a minimum undrained shear
strength in the range of about 150 to 200 kPa would be
The principal objectives in compacting soil are nor- appropriate. In effect, this would put an upper limit
mally to create a fill of high strength and low com- on the water content at which the soil could be
pressibility, and in the case of water retaining fills, of compacted.
low permeability. It is also desirable that the fill will It is apparent that the required shear strength could
not significantly soften with time as a result of expo- be achieved by compacting the soil in a very dry state,
sure to rainfall, or other weather effects. In adopting which would generally be undesirable, as dry fills may
the traditional control method it is assumed that by soften and swell excessively when exposed to rain-
aiming for maximum density the above objectives will fall. To prevent the soil from being too dry a second
be achieved. This is not automatically true, and there parameter is specified, namely the air voids in the soil.
is no reason why other parameters will not achieve Figure 1 indicates that at optimum water content the
the intended objectives equally well. Undrained shear air voids in the soil is generally about 5%. If the soil
strength and air voids are suitable alternative param- is compacted 2 to 3% drier than the optimum water
eters, and are more directly related to the intended content corresponding to the compaction effort being
properties of the fill. The use of this method was devel- used, the air voids may be as much as 8 or 10%. Thus
oped in New Zealand in the 1960s and 1970s, and to prevent the soil from being compacted too dry an
is described by Pickens (1980). The basis for using upper limit is placed on the air voids, normally in the
undrained shear strength is illustrated in Figure 5. range of 8% to 10%.
The figure shows the results of a Standard Proctor Figure 6 illustrates how this method of controlling
compaction test on clay, during which measurements compaction relates to the traditional method. The zero
of undrained strength have been made in addition to air voids line is always the upper limit of the dry den-
the normal measurements of density and water con- sity for any particular water content, and thus applies
tent. The measurements have been made using both a to both methods. The traditional method involves an
hand shear vane and unconfined compressive strength upper and lower limit on water content, and a lower
measurements on samples of the compacted soil. The limit on dry density, and thus encloses the area shown
two strength measurements give somewhat different in the figure.
results. The alternative method involves an upper limit on
It is seen that at the optimum water content the water content, corresponding to the minimum permis-
undrained shear strength is about 160 kPa from the sible shear strength, and a line parallel to the zero air
unconfined tests and about 230 kPa from the vane voids line representing the upper limit of air voids.
tests. Conventional specifications may allow water There is no specific lower limit of water content, but
contents 2 or 3% greater then optimum, in which case the air voids limit prevents the soil from being too dry.
the comparable shear strength values would be about Experience has shown that suitable limits for the
120 and 180 kPa. Thus to obtain a fill with comparable two control parameters are as follows:
Sh

Vane tests
ea

Water content limit from

rs

1.7 Unconfined comp. Tests 300

shear strength criteria
tre

Limits from water

Undrained shear strength (kPa)

ng

content and dry

th

Shear strength

density criteria
3
Dry density gm/cm

1.6 200
Limits from shear
strength and air
voids criteria Ze
1.5 100 ro
Optimum water content

air
void
s
Shear
strength
Dry density limit
Dry density

1.4 0 limit
Air
voi
ds
lim
it
Water content
limits from compaction test
1.3
20 25 30 35 40
Water content (%) Water content

Figure 5. Standard Proctor compaction test showing mea- Figure 6. Compaction control using undrained shear
surements of undrained shear strength. strength and air voids.

1008
 Undrained shear strength (hand vane values): Not the criteria can be adjusted accordingly. For example,
less than 150 kPa (average of 10 tests) Minimum the core of an earth dam built on compressible founda-
single value: 120 kPa. tions, or in a seismic zone, may need to be plastic, or
Air voids (for ‘‘normal’’ soils): Not greater than ductile, to allow for possible deformations in the dam.
8%. For some unusual soils, such as allophone This can be achieved by adopting a lower undrained
clays, this value is too low and 12% would often shear strength; a value between about 70 kPa and
be more appropriate. 90 kPa would produce a reasonably plastic material,
assuming the clay is of moderate to high plasticity. For
These values have been found to be very satisfactory in a clay embankment being built for a new highway, it
producing firm, high quality fills. The undrained shear may be desirable that the layers closest to the surface
strength can be measured in situ by a hand shear vane, (on which the pavement itself will be constructed) have
or by taking samples for unconfined compression tests. a higher strength than those deeper down. This could
It can also be measured by empirical penetrometer be achieved by increasing the required undrained shear
tests that have been correlated with undrained shear strength to say 200 kPa.
strength. The hand shear vane and examples of hand It will be evident from the account given above that
operated dynamic and static penetrometers are shown this method of compaction control does not actually
in Figure 7. The hand shear vane is both the most require compaction tests at all. However, it is still
useful (since it is a direct measure of undrained shear useful to carry out compaction tests to determine the
strength) and the simplest of these methods. degree of drying, or wetting, that may be needed to
The air voids can only be determined by measur- bring the soil to a state appropriate for compaction.
ing the density and water content in the usual way.
The author’s experience has been mainly in temper-
ate or wet tropical climates, where it is often the 4 COMMENTS ON SOILS THAT SOFTEN
case that the soil is too wet and the undrained shear DURING COMPACTION
strength criteria is difficult to meet while the air voids
requirement is easily achieved. This means that the As mentioned earlier, many residual soils are ‘‘struc-
quality control consists essentially of checking only tured’’, and when remoulded by the compaction pro-
the shear strength. With the hand shear vane, or the cess the soil becomes softer, as shown in Figure 4. The
hand penetrometers, this checking can be done as the natural water content of such soils is generally much
compaction operation proceeds. This is a big advan- higher then the optimum water content of the soil when
tage over the conventional compaction control method tested in the normal way. When planning to construct
of measuring density and water content, as the results fills with soils of this type it is important to have a clear
of these tests are often only available the day after the understanding of two factors. Firstly, it is important to
measurements are made. This would not be the case understand the compaction behaviour of the soil by
with nuclear methods of control. evaluating it with appropriate laboratory testing and
While the criteria above are suitable for a wide range field trials, and secondly it is necessary to appreciate
of compaction operations, there are some situations clearly what properties are required in the compacted
where other properties may be more important, and fill. Taking account of both the compaction character-
istics of the soil and the required properties of the fill,
appropriate compaction criteria and procedures can be
Handle to apply
Falling weight manual push adopted. This may necessitate choosing between two
Torque Guage hammer
options as follows:
a. drying the soil and using normal compactive effort
Fixed fall
height to produce a high quality fill, with a shear strength
not less than 150 kPa (ie close to its normal
Proving ring to
optimum water content),
measure force b. accepting that substantial drying is not feasible
Vane pushed
by hand into
Graduated scale
to measure
because of weather conditions, and adopting a
the soil. penetration much lower compactive effort, and possibly a lower
shear strength, so that the soil can be effectively
compacted at, or close to, its natural water content.
Some fills do not require high undrained shear
(a) HAND SHEAR VANE (b) DYNAMIC (c) STATIC strength. Examples are highway embankments of low
PENETROMETER PENETROMETER
(FALLING WEIGHT) (DIRECT PUSH) to moderate height, and embankments or benched plat-
forms to support light structures. Cores of earth dams
Figure 7. Hand shear vane and hand penetrometers. are another example. In these situations, undrained

1009
shear strengths as low as 80 kPa or less may be accept- To summarise, the advantages of the shear strength
able. If this is the case, the need to dry the soil can be and air voids control method are as follows:
avoided by adopting the second option given above.
• Large variations in soil properties present no diffi-
To determine the feasibility of this second option,
culty in applying the method. The same specifica-
it is desirable to conduct trials involving the excava-
tion limits apply regardless of the variations.
tion, transport, and compaction of the soil. In this way
• Field control is more direct as the value of the
the ‘‘optimum compactive effort’’ discussed earlier in
undrained shear strength is known immediately the
relation to Figure 4 can be determined. Excavation,
measurements are made.
transport, and spreading should be carried out in such
• The specification is easily varied to produce fills
a way that disturbance and remoulding of the soil is
with particular properties needed in special situa-
kept to a minimum. In other words the natural struc-
tions.
ture and strength of the soil should be retained as
much as practical. The compaction operation should The method also allows for the specification to
similarly be conducted so that remoulding the soil is be related more closely to the requirements of the
minimised. Light, tracked equipment is likely to be fill itself. The undrained shear strength can be low-
most appropriate, and the compaction process may ered or raised as necessary, depending on whether
consist essentially of ‘‘squeezing’’ intact fragments of a relatively soft plastic fill or a high strength fill is
soil together to form a uniform fill. For this purpose required.
only a few passes of the compaction equipment is likely
to be preferable to a large number of passes; the latter
may progressively soften the soil and not make it any ACKNOWLEDGEMENTS
more compact or stronger.
The author wished to acknowledge that the method
described here has been developed by a number of peo-
5 CONCLUSIONS ple and agencies in New Zealand. The New Zealand
Ministry of Works, and the Consulting firm Tonkin
To handle the difficulties that ‘‘problematic’’ soils and Taylor Ltd have probably played the dominant role.
present to earthworks and compaction control, it is
necessary to be flexible in the way the soil is eval-
uated and the way in which the compaction process REFERENCES
is carried out and monitored. The use of the stan-
dard method involving water content and dry density Kuno, G., Shinoki, R., Kondo, T. & Tsuchiya, C. 1978. On
as control parameters is simply impractical for many the construction methods of a motorway embankment by a
residual soils. The alternative method described here sensitive volcanic clay. Proc. Conf. on Clay Fills, London.
using undrained shear strength and air voids provides 149–156.
Pickens, G.A. 1980. Alternative Compaction Specifications
an appropriate alternative, which, while avoiding the for Non-uniform Fill Materials. Proc. Third Australia-
difficulties associated with the standard method, still New Zealand Conf. on Geomechanics, Wellington 1.
produces a fill of equally good engineering properties. 231–235.
It is widely used in New Zealand, and is accepted and Wesley, L.D. 2002. Geotechnical characterisation and prop-
described in the local code of practice for earthworks. erties of allophone clays. Proc. International Workshop on
It has also been used on a number of major earthworks Characterisation and Engineering Properties of Natural
operations in Southeast Asia. Soils. Singapore, Dec. 2002.

1010
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Numerical analysis of soil-arch effect of anti-slide piles

Yuanyou Xia, Xiaoyan Zheng & Rui Rui

School of Architecture & Civil Engineering, Wuhan University of Technology, Wuhan, Hubei, China

ABSTRACT: This paper aims at solving the problems in the calculation and design of anti-slide piles presently.
The basic principle of soil arching is expatiated, FLAC3D is applied to build the 3D model of anti-slide piles,
gliding masses and anchorage zone, and the numerical simulation is carried out. Through numerical tests, the
soil arching along the whole pile is studied. According to the displacement contours, the effects are mainly
appears in the soil under and among the piles. The effect under different depths of thrust and gliding masses
is also studied. The results can provide referential information for engineering design and theoretical study of
anti-slide piles.

1 INTRODUCTION ‘‘soil-arch effect’’ for the first time. Smid, Novosad,

Matuttis, Vanel, Lawrence etc. studied the stress trans-
As collapse, failure and debris flow, landslide is also fer in granular media, and obtained some useful results
a natural disaster commonly encountered in moun- for soil-arch effect mechanics (Low, 1994, Lawrence,
tain area construction. As landslide happens, traf- 2002).
fic interruption, river obstruction, factory and mine Although soil-arch effect plays an important role
destruction, as well as village and town buried are in various geotechnical engineering, such as tunnels
usually accompanied, which will cause serious dam- and protecting piles of foundation pits, the definitions
age. Therefore, the study on landslide prevention has of it are not the same in different structure (Zhang,
always been a concern for all countries. Compared 2004). For anti-slide pile engineering, soil-arch effect
with traditional measures of landslide prevention, such transfers the stress that soil subjects to piles. Hence, it
as drainage, load reducing and retaining wall, anti- can be defined as a phenomenon of shear stress trans-
slide pile engineering is the most convenient and effi- fer (Chi, 2004). In 1980, Padfield (1980) mentioned
cient way, so it is widely used in engineering practice. arching during his research of riverbank stability. Wu
Presently, in the calculation and design of anti-slide Zishu studied the mechanism of soil-arch effect, and
pile, many assumptions are made for pile-soil inter- presented the minimum embedded soil depth and the
action. As a result, the load transfer of pile-soil and ultimate arch span (Wu, 1995). Based on the research
soil-arch effect along the whole pile cannot be faith- of Liang, Zeng, Chen, and Martin, Zhang Jian-xun etc
fully reflected, which causes a conservative design and (Zhang, 2004, Nogami, 1991, Hesham, 2000) found
a waste of money. So, it has theoretical and realistic that the major mechanisms for stabilizing soil due to
meaning to study the soil-arch effect in anti-slide pile. lateral movements by passive piles was soil-arch effect.
Blocking effect, such as traffic jams, compaction And then they used a finite element analysis software
of granular materials, crowds jams when panic and package, Plaxis 8.1, to study the causes of soil-arch
so on, is unavoidable in the array of all systems. It effect in passive piles.
is the root of soil-arch effect (Garcimartin, 2004). So far, in the field of geotechnical engineering,
As far back as 1884, an English scientist, Roberts although the monitoring data, experimental models
discovered that the pressure acting on barn bottom and theory relating to soil-arch effect become more and
attained the maximum value and kept constant when more, the interaction between passive piles and mov-
the accumulation height of the grain reached a certain ing soil mass is a complex problem that hasn’t been
value, and he called this phenomenon soil effect. In sufficiently studied as yet, much less the research on
1943, using ‘‘Trap door’’ test, Terzaghi (1943) expli- pile-soil interaction in soil slope. So it is necessary to
cated the phenomenon of stress transfer between soil further the study of soil-arch effect problem of passive
mass and the rigid edge beside, and he called this piles.

1011
2 FORMATION PRINCIPLE OF SOIL-ARCH
EFFECT

In slope engineering, after anti-slide pile construction,

soil mass between piles is tending to move outside the
slope when the deformation of anti-slide piles itself
occurs to restrain slope displacement. The trend fur-
ther develops after the excavations of the soil before
piles. Because the lateral displacement of anti-slide
pile is less than the slope’s, the soil mass after piles in
a partial area continuously extrudes piles and uneven
earth pressure is generated. The displacement of soil
mass between piles is different due to the pile restrain:
the displacement of soil mass by piles is small, while
the displacement of soil mass away from piles is big. Figure 2. 3D meshing graph.
Under this condition, ‘‘wedging’’ function (Wu, 1995,
Zhang, 2004, Zhou, 2004) is occurred due to the uti-
lization of shear resistance of soil mass before and after front and rear slip surfaces are considered; calculated
piles, which is also known as soil-arch effect. depth is 26 m.
According to the conditions above, the model is
meshed. And ‘‘interface element’’ is used to simulate
3 SIMULATION CALCULATION the slip surface and pile-soil interface. The meshing
figure is as shown in Figure 2.
3.1 Procedure for anti-slide pile simulations
In this paper, 3D computational model in FLAC3D 3.2 Numerical test analysis of soil-arch effect
is applied to simplify the problem, and the exam- Soil-arch effect, which is the main mode of action in
ple in paper which named Design and Calculation pile-soil interaction, has a bearing on whether anti-
of Pre-Stress Anchor Slide-Resistant Pile is adopted. slide pile can give its full play to resist slide. There
By symmetry, the part between two piles is consid- are many factors that can influence soil-arch effect,
ered. Constraint in the z-direction is defined along the such as pile-soil interface parameters, thrust, physical
model bottom; constraint in the x-direction is defined mechanic parameters of gliding mass and anchorage
along the opposite sides of the model; constraint in the zone, as well as the anchorage depth, size and strength
y-direction is defined along the front and rear slip sur- of piles. The stress and deformation of piles is differ-
face of the model; stress boundary constraint, whose ent, as a result, the development degree of soil-arch
value is lateral self weight of the gliding mass, is effect is not the same either. This numerical test is car-
applied above the slip surfaces. As shown in Figure.1, ried out under different thrusts to analyze the soil-arch
the depth of slip surface is 10 m; rectangle pile with effect of gliding mass.
cross section of 1 × 1.5 m is adopted, and its length Design data:
is 16 m; pile spacing is 4 m; 4 m in y-direction of the Accumulative formation of soil, block stone,
Macadam are above the slip surface and their defor-
mation increases from the top downward, γ1 =
1900 kg/m3 , ϕ1 = 26◦ . Mudstone strongly efflo-
resced and shale, which can be considered as com-
pacted soil layer, are below the slip surface, γ1 =
2100 kg/m3 , ϕ1 = 42◦ 10 . Thrust is as a triangular
distribution.

3.2.1 Stress analysis

Arching effect, which is an inherited nature of par-
ticulate material, forms a stable structure to prevent
particles from flowing when blocking. The soil-arch
effect of anti-slide pile is complexity. Because the dis-
placement of soil mass after piles caused by thrust is
blocked by piles, blocking effect and extrusion occur.
And stress redistributes due to the extrusion. The
Figure 1. 3D model graph (unit:m). direction of principal stress changes to the direction

1012
Table 1. Parameter choosing.

Density/ Poisson’s
Categories Thickness kg/m3 E/Pa ratio C/Pa ϕ/◦ K/Pa G/Pa

Detritus 10 1900 1.49E+07 0.3 1.00E+03 26 1.24E+07 5.71E+06

Strong
weathered
mudstone
and shale 16 2100 1.87E+07 0.4 3.10E+04 42.17 3.11E+07 6.67E+06
Pile 2500 2.60E+10 0.167 1.30E+10 1.11E+10

that approximately parallel with the tie of the circular

arch between two contiguous piles. Therefore, an arci-
form compacted complex is formed, and some of the
thrust is transferred to anti-slide piles. A certain exter-
nal force corresponds to a certain structure, that is to
say, if the force is not the same, the structure formed
will be different either. As to gliding mass with anti-
slide piles, the thrust is different in the depth direction.
As a result, the soil arch is varied with the depth, and
that is why soil-arch effect can not be well reflected
by 2D. In this test, the maximum principal stress con-
tour sketches of each plane under different thrust are
mapped through simulation calculation.
Through calculation, the variation of displacement
and stress on plane z = 0, z = −2, z = −4,
z = −6, z = −8 is simulated. For easy of display,
the stress of pile elements is not protracted in some
contour sketches, while the omitting will not affect
the research. The contours of principle stress under
the condition that thrust is 400 kN/m are adopted to
analyze.
The contour sketches is as shown in Figure 3. It
reveal some laws. The color of the upper part, which
corresponds to the soil mass before piles, is homo-
geneous, and the stress value it indicates is small. Figure 3. Contours of maximum principle stress of each
The form of the lower part, which corresponds to plane when thrust is 400kN/m.
the soil mass after piles, looks like an anticlinal fold:
with the increase of the depth, an ‘‘inner core’’ with
high stress appears in the middle; and as the depth of the ‘‘inner core’’ increases and gradually expands to
increases to −8 m, an arching area appears between surrounding area, some of the soil mass between and
and below the two piles, and the contours bending after piles is squeezed, and finally, an arching com-
relieves. The contours of maximum principle stress paction area is formed. The contours of z = −8 m is
varied with depth fully demonstrate the difference of the most obvious, and the principle stress is nearly the
soil-arch effect in depth. The manifestation of soil- same in the compaction area.
arch effect can be divided into 2 types: gliding mass Both of the two types of soil-arch effect come into
edges forward with pressure and friction force due to being in all thrust and depth. In this test, numeri-
its movement towards the pile is blocked, and soil-arch cal simulations are done when thrust is 100 kN/m,
effect is characterized by tractive force of soil mass; 200 kN/m, 400 kN/m and 800 kN/m. It can be seen
as the thrust & depth increase to a certain degree, the that, soil-arch effect is not actually formed when the
relative displacement between soil and pile increases, thrust is small, and pile-soil cooperation is character-
the direction of principle stress deflects, and the prin- ized by tractive force due to pile blocking and friction.
ciple stress in the middle increases continuously, and As thrust & depth increases, the relative displacement
finally, a ‘‘inner core’’ with high stress generated. between soil and pile increases, and then a compacted
With the increase of thrust & displacement, the stress soil arch is formed in the middle of the soil mass after

1013
piles to resist the thrust. By this time, soil-arch effect when it reaches a certain position before piles. From
is the main manifestation of pile-soil cooperation. the calculation result, the influence sphere caused by
deformation difference is smaller than the pile spac-
3.2.2 Deformation analysis ing, which indicates the pile spacing chosen is proper
The calculation results are introduced to SURFER at the same time.
to process. A section per 2 m, the y-direction strain From the analysis above, the movement caused by
sketches of each section under different thrust are thrust of the soil mass after piles is blocked by piles,
obtained. And the section z = −8 m is taken to the soil mass is squeezed, and soil-arch effect is func-
analyze. tional. From the contour sketches, soil-arch effect
In Figure 4, all of the contours have the trend mainly appears in the soil mass after and between piles.
of decreasing from negative to positive; the contour Soil-arch effect of soil mass after piles is character-
sketches twist in the positive direction in the middle, ized by a compact arch zone due to the deflection of
and the bent degree is the biggest between the piles principle stress, while soil-arch effect between piles is
with gradual sparse contours. From the y-direction realized by frictional force on pile-soil interface and
strain sketch of each section under different thrust, tractive force. With the increase of depth, the bend of
the contours of the soil mass after piles become more contours between piles becomes smaller and smaller
and more compact, and the bend between the piles until reversed bend appears, while the contours of the
become less with the increase of the depth. This phe- soil mass after piles become more and more com-
nomenon shows that, the amount of deformation of pact, which means soil-arch effect is more obvious: the
the soil mass after piles is bigger than the soil mass relative displacement between soil and piles reduces
before, as a result, the soil mass after piles is squeezed; to a microscopic value, the thrust soil mass subjects
and with the increase of depth, the extrusion become caused by the gliding mass after piles is very small, and
more and more obvious. The displacement on the axle soil arch transfers the thrust to piles. Reversed bend
is bigger than the gliding mass near the piles, but appears due to the deformation of the soil mass before
along the direction of before-pile, the displacement piles. Meanwhile, the friction on pile-soil interface is
becomes congruous, and the difference disappears very small.
The displacement data calculated is plotted: the
4
maximum displacement of the gliding mass varied
4
0.0223 with depth is as shown in Figure 5; the relative
0.0106 0.0219
3
3
0.0215
displacement between piles and soil is as shown in
2 0.0102
2
0.0211 Figure 6; the maximum displacement of the gliding
0.0207

1
0.0098 0.0203 mass and relative displacement varied with thrust is
1
0.0199
0.0195
as shown in Figure 7; the displacement of piles varied
0.0094
0
0 0.0191 with depth is as shown in Figure 8.
0.0187
-1
0.009
-1 0.0183 From Figures 5–8, the maximum displacement of
-2
0.0086
0.0179
0.0175
gliding mass under different thrust decreases linearly
-2
0.0171 with the increase of depth in general, while the relative
0.0082 0.0167
-3
-3
0.0163 displacement in the superficial part is small, and the
0.0078 0.0159
-4
-4
0 1 2 3 4
0 1 2 3 4
0.0155

(a) 100kN/m (b) 200kN/m

displacement of gliding mass/mm
4 4 0.1156 0
0.1144
0.1132
0.0476
0.112 0 50 100 150 200 250 300
3 3
0.0468 0.1108
0.046 0.1096 -2
0.1084
2 0.0452 2 0.1072
0.0444 0.106
0.1048
1
0.0436
1
0.1036 -4
0.0428 0.1024
0.1012
depth/m

0.042 0.1
0 0.0412 0 0.0988 100kN
0.0976 -6
0.0404 0.0964 200kN
0.0396 0.0952
-1
0.0388
-1
0.094 400kN
0.038
0.0928
0.0916
-8 800kN
-2
0.0372
-2 0.0904
0.0892
0.0364 0.088
-3 0.0356 -3 0.0868 -10
0.0856
0.0348 0.0844
0.034 0.0832
-4 -4 0.082
0 1 2 3 4 0 1 2 3 4 -12
(c) 400kN/m (d) 800kN/m
Figure 5. Maximum displacement of gliding mass varies
Figure 4. Displacement contours of plane z = −8 m. with depth.

1014
 displacement between pile and soil/mm seen that the piles turn about the point −13.15. Form
0 the maximum displacement of gliding mass and the
0 5 10 15 20 25 30 relationship between thrust and relative displacement,
-2 the displacement increases linearly with the increase
of thrust.
-4 100kN
200kN
depth/m

400kN
-6 800kN 4 CONCLUSIONS
-8
In view of the anti-slide pile design features, the
-10 three-dimensional model is created and the parame-
ter numerical test is studied by using the rock & soil
-12 engineering three-dimensional numerical calculation
software FLAC3D. The pile’s total depth of earth-arch
Figure 6. Relative displacement between pile and soil varies effect is studied and the establishment of the earth-arch
with depth. is analyzed too.
Concluding the above study, we may state the
following:
250
– The movement caused by thrust of the soil mass
displacement of gliding mass/mm

200 after piles is blocked by piles, the soil mass

is squeezed, and soil-arch effect is functional.
From the contour sketches, soil-arch effect mainly
150
appears in the soil mass after and between piles.
Maximum relative displacement
between pile and soil
– As the thrust & depth increase to a certain degree,
100
Maximum displacement of gliding the relative displacement between soil and pile
mass
increases, the direction of principle stress deflects,
50 and the principle stress in the middle increases con-
tinuously, and finally, a ‘‘inner core’’ with high
0 stress generated. With the increase of thrust & dis-
0 200 400 600 800 1000
thrust/kN/m
placement, the stress of the ‘‘inner core’’ increases
and gradually expands to surrounding area, some
Figure 7. Maximum displacement of gliding mass and of the soil mass between and after piles is squeezed,
relative displacement between pile and soil vary with thrust. and finally, an arching compaction area is formed.
– Both of the two types of soil-arch effect come
into being in all thrust and depth. Sliding different
displacement/mm depth of the soil arching effect of different forms
0
-50 0 50 100 150 200 250 and different thrust of soil arching different forms.
-2 It can be seen that, soil-arch effect is not actu-
-4 ally formed when the thrust is small, and pile-soil
-6 cooperation is characterized by tractive force due
to pile blocking and friction. As thrust and depth
depth/m

-8
increases, the relative displacement between soil
100kN
-10 and pile increases, and then a compacted soil arch
200kN
-12 400kN is formed in the middle of the soil mass after piles
-14
800kN to resist the thrust. By this time, soil-arch effect is
the main manifestation of pile-soil cooperation.
-16
-18

Figure 8. Pile displacement varies with depth. REFERENCES

Chi Yue-jun & Song Er xiang et al. 2003. Experimental study

on stress distribution of composite foundation with rigid
value increases with the increase of depth, but when piles. Rock and Soil Mechanics, 24 (3):339–343.
the depth exceeds −8 m, the value decreases. The tri- Garcimarti, A., Zuriguel, I. & Maza, D. 2004. Jamming in
angular distribution of thrust has much to do with this granular matter Proceeding of 8th Experimental Chaos
phenomenon. Meanwhile, from Figure 8, it can be Conference: 279–288.

1015
Hesham, M., Naggar, E.I. & Kevin, J. Bentley. 2000. Terzaghi, K. 1943. Theoretical soil mechanics. New York:
Dynamic analysis laterally loaded piles and dynamic p-y John Wiley and Sons.
curves. Canadian Geotechnical Journal, 37: 1166–1183. Wang Nian-xiang. 2000. A summary of interaction between
Lawrence. 2002. The mechanism of load transfer in granu- passive pile and soil mass. Journal of Nanjing Hydraulic
lar materials utilizing tactile pressure sensor. University Research Institute, 3: 69–76.
of Massachusetts Lowell. Wu Zi-shu, Zhang Li-min & Hu Ding. 1995. Studies on
Low, B.K., Tang, S.K & hhoa, V. 1994. Arching in piled the Mechanism of Arching Action in Loess. Journal of
embankments. Journal of Geotechnical Engineering, ChengDu University of science and technology, 83 (2):
120(11): 1917–1937. 15–19.
No. 2 Survey and Design Institute. 1983. Design and Calcu- Zhang Fan-rong. 2004. The Studies of design and Construc-
lation of Pre-Stress Anchor Slide-Resistant Pile. Beijing: tion Technique for Anti-Slide Structures. Ms D Thesis.
China Railway Press. Southwest Jiaotong University.
Nogami, T., Jones, H.W. & Mosher, R.L. 1991. Seismic Zhang Jian-xun, Chen Fu-quan & Jian Hong-yu. 2004.
response analysis of pile-supported structure: assess- Numerical analysis of soil arching effects in passive piles.
ment of commonly used approximations. Proceedings of Rock and Soil Mechanics, 25 (2): 174–184.
2nd international conference: Recent advancements of Zhou De-pei, Xiao Shi-yu & Xia, xiong. 2004. Discussion on
geotechnical earthquake engineering and Soil dynamics, rational spacing between adjacent anti-slide piles in some
[s.l.]: [s.n.] 931–940. cutting slope projects. Chinese Journal of Geotechnical
Richard, L.H. 1985. The arch in soil arching. Journal of Engineering, 26 (1): 132–135.
Geotechnical Engineering, 111(3): 302–318.

1016
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Determination of the critical slip surface based on stress distributions

from FEM

Daping Xiao & Chunqiu Wu

Institute of Foundation Engineering, China Academy of Building Research, Beijing, China

Hong Yang
MWH Americas, Inc., Walnut Creek, California, USA

ABSTRACT: A method is proposed to extend the classic slip line theory and determine the critical slip surface
of a slope based on stress distribution from Finite Element Method (FEM). In the proposed method, the slip line
field in the slope is determined from the orientation relationship between the slip lines and the principal stress
tracing lines. Factors of Safety (FS) are then calculated for each slip line and the critical slip surface is identified
to be the one with the minimum FS. The method has taken into account the possible existence of both elastic
and plastic zones in the slope and the effect of different flow rules. The validity and applicability of the method
are demonstrated using case studies of both a homogeneous slope and a slope with a weak layer.

1 INTRODUCTION in the classic slip line theory. Zhang et al. (2003) pro-
posed the concept of potential slip line field based on
Evaluation of slope stability based on the results of the extension of the classic slip line theory. In addition
FEM has long been studied in geotechnical engineer- to these previous studies, the current study developed
ing. The evaluation involves two interrelated aspects, a numerical modeling method to determine the criti-
namely the calculation of a FS for the slope and deter- cal slip surface of a slope based on extended slip line
mination of the critical slip surface. The key issue is to theory and FEM results. The numerical method is
determine the critical slip surface based on the FEM capable of solving slope stability problems involving
results, which can be solved based on the slip line the- both elastic and plastic zones and different materials
ory. The slip line method in geotechnical engineering using associated or non-associated flow rules for any
is used to determine the stress distribution of the plastic slope geometry and boundary conditions. The numer-
zone and the ultimate load based on equilibrium equa- ical method, therefore, extends the applicability of the
tions, failure criteria and boundary conditions. Once slip line method.
the shapes of the slip lines are determined, the critical
slip surface is identified as the slip line corresponding
to the minimum FS.
2 CLASSIC SLIP LINE THEORY
The slip line field indicates the stress state either
AND ITS EXTENSION
along the slip lines or in the localized plastic zones
adjacent to loading. The slip line field does not indicate
2.1 Classic slip line theory
the stress state of the entire domain of slope stability
problem. Therefore, it is necessary to extend the slip Under plane strain and plastic state, two intersecting
line field to the entire domain of slope in order to apply shear failure planes exist in every point of the plane.
the slip line method to slope stability analysis. The By connecting these shear planes at all points, two
extended stress field should fully comply with both clusters of failure lines can be obtained, namely the
limit equilibrium and failure criteria. To achieve this, α- and β- slip lines, as shown in Figure 1. The tan-
an appropriate approach is to determine the slip line gent of any point on a slip line is the direction of the
field from the information of a stress field generated slip line.
from elasto-plastic finite element analyses. The direction of any point on the slip line is related
Zhen (2000) studied the mathematic model needed to the tracing line of principal stress. Two clusters of
to satisfy the critical slip surface and proposed a dif- orthogonal curves (i.e. the so-called tracing lines of
ferential equation which is the equation for slip lines principle stresses) can be obtained by connecting the

1017
 The nonlinear differential Equations (1) through
y β
(4) for the slip lines with unknown curve coordinates
1 have to be solved using numerical methods, as well as
2μ = through an iteration method (Gong 1990).
2

2.2 Potential slip line field in elastic zone

Classic slip line theory was developed based on plas-
tic mechanics and applies only to the plastic zone. The
α θ β elasto-plastic finite element analysis incorporates both
2
x plastic and elastic zones. In order to use FE analysis for
1 determining the critical slip surface, the concept of slip
line must be extended to the elastic zone by introducing
Figure 1. Slip lines and principal stress tracing lines. the concept of potential slip line field. Corresponding
to the definition of slip lines, the potential slip lines
are defined as the two clusters of the most danger-
fragments representing the direction of principle stress ous shearing lines (i.e., the weakest surfaces for shear
(i.e. curves 1-1 and 2-2 as shown in Figure 1). The resistance) (Zhang et al. 2003). The μ angle between
analysis of this paper assumes that the tracing line the potential slip line and the primary principle stress
of major principle stress, σ1 , is the baseline; the α- can be determined by:
line has a clockwise sharp angle to the baseline, and
the β-line has a counter-clockwise sharp angle to the π θmax
baseline. μ = − (5)
4 2
For a Mohr-Coulomb material, the angle between
the two clusters of slip lines is 2μ = π/π2 − φ, and σ1 − σ3
the angle between the slip line and tracing line of prin- in which θmax = arcsin σ1 + σ3
2
; θmax = φ when
+ cot φ
ciple stress is μ = π/4 − φ/2, in which φ is the 2
the point reaches it strength limit, and θmax < φ when
friction angle of material (Gong and Shi 2000). When the point is in elastic state.
using classic slip line theory to determine stresses,
the curve coordinates (Sα , Sβ ) constituted by the two
clusters of slip lines can be solved from the following 2.3 Slip line field with consideration
equations: of non-associated flow rule
Constitutive relationships are not considered in classic
dy slip line theory. In classic slip line theory, failure cri-
α-line: = tan(θ − μ) (1)
dx teria constitute only a constraint. However, the classic
slip line theory inexplicitly comprises the assump-
dy
β-line: = tan(θ + μ) (2) tion of associated flow rule for material, namely
dx that the dilation angle, ψ, equals the internal fric-
∂P ∂θ tion angle, φ. For geomaterials conforming to the
− sin(2μ) + 2R Mohr-Coulomb failure criteria, the effect of dila-
∂Sα ∂Sα tion will be amplified when the associated flow rule
  is adopted. Many field observations and laboratory
∂x ∂y
+ γ sin(2μ) + cos(2μ) =0 (3) testing have indicated that dilation angles for real
∂Sα ∂Sα soils are much smaller than internal friction angles
(Kabilamany and Ishihara 1990). Davis (1968) found
∂P ∂θ that the velocity characteristic lines (i.e. the slip lines)
sin(2μ) + 2R
∂Sβ ∂Sβ no longer coincide with the characteristic lines of
  principle stresses when the dilation angle becomes
∂x ∂y unequal to the internal friction angle. The slip lines
+ γ sin(−2μ) + cos(2μ) =0 (4)
∂Sβ ∂Sβ no longer satisfy Equations (1) and (2). For this case,
the angle between the slip line and major principle
where P is the averaged stress; R is the radii of stress stress becomes:
circle, and R = P sin φ + c cos φ for M-C material;
γ is unit weight; θ is the angle between major principle π ψ
μ = − (6)
stress σ1 and x-axis. 4 2

1018
 Davis further found that the normal stress σ* and
shear stress τ* on slip line has the relationship of:

F = τ ∗ − c∗ − σn∗ tan φ ∗ = 0 (7)

where the strength parameters c∗ and φ ∗ have simi-

lar meanings with those in the Mohr-Coulomb criteria
with the following relationship:
(a) (b)
c∗ tan φ ∗ cos ψ cos φ
= = (8) Figure 2. Search for slip line field. (a) Discretization of
c tan φ 1 − sin ψ sin φ
the slope; (b) Case showing spacing needs to be reduced for
searching.
As a result, non-associated flow rule can be con-
sidered in the slip line field for geomaterials. For
FE analysis using non-associated flow rule, the slip
line field obtained from a FE-generated stress field
automatically considers the non-associated flow rule.

3 NUMERICAL MODELING OF SLIP LINE

FIELD BASED ON FE ANALYSIS

With the incorporation of potential slip lines, the two Figure 3. Slip line traced from the slip line direction field.
clusters of slip lines can be uniformly represented by
Equations (9) and (10):
4 DETERMINATION OF CRITICAL SLIP LINE
dy
α-line: = tan(θ − μ ) (9)
dx On the basis of slip line theory, two clusters of slip
dy lines (i.e. the α and β clusters) can be obtained from
β-line: = tan(θ + μ ) (10) FE analysis, as shown in Figure 4. The clusters com-
dx prising the critical slip line can be easily identified
based on engineering experience. For the case shown
In this paper, rather than solving the nonlinear dif- in Figure 4, the α-slip line field has the actual slip
ferential functions (1) through (4), the slip line field is lines. The critical slip line is that corresponding to the
obtained from Equations (9) and (10). Firstly, θ and μ minimum FS, which can be calculated by the following
at each discrete point is obtained from FE results of the equation:
stress field, and the direction of slip lines can be deter-
mined from Equations (9) and (10). Since FE analysis 
considers equilibrium, geometric and physical func- Si (σ tan φ + c)ds
tions, and elastic and plastic zones; FE analysis more FS =
n
 (11)
realistically and reasonably predicts the potential slip Si τ ds
n
line clusters as compared with classic slip line theory.
As shown in Figure 2(a), the entire range of poten-
tial slope failures is divided horizontally and vertically where n is the number of sections searched for the slip
into discrete points for searching slip line paths. The lines, and σ and τ are normal and shear stresses on the
initial horizontal spacing b and vertical spacing d slip line, respectively.
depend on requirements and limitations of computa-
tion accuracy and capacity. The spacing can be further
divided using the bisection method, if needed. For the 5 CASE STUDIES FOR VERIFICATION
case shown in Figure 2(b), the horizontal spacing for
the search has to be reduced. A program for identifying the critical slip line has been
From the calculated direction field of slip lines, the develop, which is used in association with FE results
slip line at any point of the exit section can be easily obtained from the commercial software ANSYS. The
traced (see Figure 3) and the slip line field passing effectiveness of the proposed method was verified by
through the exit section can be obtained. the following case studies.

1019
 Figure 5. Slope geometry and material properties for case
(1) α-slip line field study 1.

FS values for the slip lines from bottom to top

Critical slip
surface

(2) β-slip line field Figure 6. Slip line field for case study 1.

Figure 4. α- and β-slip line fields.

The minimum FS is 1.012 with the slip surface located
at the middle, which is the critical slip surface. The
minimum FS is close to unity, which indicates the
5.1 Case study 1—homogeneous slope
effectiveness of searching for the critical slip surface
Figure 5 shows the soil properties and geometry of a within the slip line clusters.
homogeneous slope. This example has been previously Figure 7 shows the relationship between the dis-
studied by Zhang et al. (2002) using FEM and by Wang tribution of plastic zones and the distribution of slip
and Yin (2002) using limit analysis and limit equilib- lines. The distributions of slip lines and plastic zones
rium methods. The results of these former studies are are very similar. The critical slip surface (the 6th from
compared with the results from this study. the bottom) just passes through the plastic zone of the
Assuming the dilation angle of the soil is ψ = 0 slope, which indicates the location of the slip surface
and using the non-associate flow rule and strength is reasonable. Also, the four slip surfaces (5th to 8th)
reduction method, the stress distributions and slip sur- with FS values close to unity are all located in the
faces in the slope under critical state, as determined plastic zone, which indicates that the slope can fail
using ANSYS, are shown in Figure 6. The FS cal- along any of the slip surfaces. This suggests that the
culated using FEM (i.e., the coefficient of strength slip occurs in a zone rather than along a single surface.
reduction under limit state) was 1.98, which is the This zone is referred to as the shear zone and has been
same value calculated by Zhang et al. (2002). Mean- documented with many field observations.
while, the FS calculated by Wang and Yin (2002) was
1.97 using limit analysis and 2.03 using Morgenstern-
5.2 Slope with a weak layer
Price method based on limit equilibrium analysis. This
agreement indicates that the calculation using finite Figure 8 shows a slope with a 2 m thick weak layer
element method in this study is correct, since the below the toe of the slope. The geometry of the slope
results are close enough with those obtained using and the soil properties are shown in the Figure 8. The
other methods by other researchers. FS value was calculated to be 1.24 using the strength
The FS for each slip surface was calculated using reduction method assuming the materials conform to
Equation (11) and is shown in Figure 6. The strength the non-associated flow rule.
parameters are based on strength reduction. The figure The slip surfaces calculated from limit analysis and
shows that the FS initially decreases and then increases FEM are shown in Figure 9. The FS values were cal-
as the slip surfaces go up from the bottom to the top. culated from Equation (11). The slip surface with the

1020
 Shear zone

Figure 7. Relationship of the locations for the plastic zone Figure 10. Relationship of the locations for the plastic zone
and slip lines for case study 1. and slip lines for case study 2.

minimum FS value (the 6th from left to right) of 1.029 in the plastic zone, which forms the shear zone, as
is the critical slip surface. The FS is close to unity mentioned previously.
indicating the effectiveness of the search method for Another case example of slope with a weak layer
the slip surfaces. It can be seen from the figure that, was also studied using both associated flow rule and
because of the existence of the weak layer, the shape non-associated flow rule (Wu 2004). The critical slip
of the slip surface is more like a logarithmic spiral surfaces calculated are all in the weak layers with the
curve rather than a circular curve which is commonly FS close to unity. The results are not presented here
observed or a homogeneous slope. due to the length limit of the paper.
The relative locations of the plastic zones and the
slip surfaces at limit state condition are shown in
Figure 10. The shapes of the slip surfaces match the 6 CONCLUSIONS
distribution of the plastic zones well. The critical slip
surface and those slip surfaces with FS close to unity This paper extends the classic slip line theory and
(6th to 10th form the left to the right) are all located presents a method to determine the critical slip surface
based on stress distribution using FEM. The method
determines the slip surfaces using the relationship of
the slip line and principal stress tracking line to cal-
= 19.5kN/m3 culate the FS for each slip surface. The slip surface
10m

= 20kPa with the minimum FS is the critical slip surface. Both

G = 15
elastic and plastic zones of a slope are considered, and
2m

different flow rules were also adopted. The method

2m

was proved to be effective through case examples for

= 19.5kN/m3 = 19.5kN/m3 both homogeneous slope and a slope with a weak layer.
= 20kPa =0
= 15 = 15

Figure 8. Slope geometry and material properties for case REFERENCES

study 2.
Davis, E.H. 1968. Theories of plasticity and the failure of soil
masses. In Lee I. K. eds., Soil Mechanics, Butterworth,
FS values for the slip lines from bottom to top
London, p. 341–380.
Gong, X.N. 1990. Soil Plastic Mechanics. Press of Zhejiang
University, Hangzhou, China (in Chinese).
Gong, X.N. Shi, M.D. 2000. Numerical solutions for slip
Critical slip line field, in Computer Analysis for Geomechanics. China
surface Construction Press. p. 193–214 (in Chinese).
Kabilamany, K. Ishihara, K. 1990. Stress dilatancy and hard-
ening laws for rigid granular model of sand. Journal of Soil
Dynamics and Earthquake Engineering, 9(2), 66–77.
Wang, Y.J., Yin, J.H. 2002. Slope stability analysis using
a method with associated and non-associated flow rules.
Proceedings of International Conference on Soil Mechan-
Figure 9. Slip line field for case study 2. ics and Geotechnical Engineering.

1021
Wu, C.Q. 2004. Nonlinear Finite Element Method in Soil Zhang, L.Y., Shi, W. M., Zhen, Y. R. 2002. Slope stability
Stability Problems. Ph.D. dissertation, Wuhan University analysis using FEM under plane strain conditions. Chinese
(in Chinese). Geotechnical Engineering Journal, 24(4), 487–490 (in
Zhang, G.X., Liu, X.H., Wei, W. 2003. Two-dimensional Chinese).
slip surface of slope and stability analysis using elasto- Zhen, H. 2000. Several Types of Non-linear Problems in
plastic FEM. Chinese Railway Journal, 25(2), 79–83 Geomechanics. Ph.D. dissertation, China Academy of
(in Chinese). Science (in Chinese).

1022
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Effect of drainage facilities using 3D seepage flow analysis reflecting

hydro-geological structure with aspect cracks in a landslide –
Example of analysis in OODAIRA Landslide area

Masao Yamada
Japan Conservation Engineers Co., Ltd., Japan

Keizo Ugai
Gunma University, Japan

ABSTRACT: Stratum water and crack water through cracks of bedding planes and of joint surfaces give a
great influence on the amount of displacement of moving mass in a landslide area. In this paper, at first, we
examined the three-dimensional modeling of seepage flow analysis reflecting aspect cracks and then calculated
water pressure distribution specifically by melting snow water in order to grasp a change of water pressure
distribution of before-and-after construction of drainage facilities of well and drainage pipe. On that occasion,
we made accurate modeling structure of various cracks and drainage facilities of well and drainage pipe. Using
the results derived by the seepage flow analysis, we carried out a three-dimensional stability analysis. Based on
the result of the stability analysis, we evaluated the stability of the landslide slope and the effect of drainage
facilities.
In OODAIRA Landslide area where landslides were reactivated by melting snow, by applying the above
mentioned analysis, we were able to obtain the acceptable results that matches field situation between fluctuation
of water pressure and displacement of moving mass. We were then able to simulate effectiveness of drainage
facilities.

1 INTRODUCTION maintaining its form and size as much as possible and

of collecting boring pipe and drainage pipe as line
Groundwater consists of stratum water and crack crack element.
water. The hydro-geological structure where ground- It is difficult to set up the coefficient of perme-
water flows in crack zones and stratums affects ability of ground layers and cracks in order for three-
landslide displacement. dimensional model with cracks to match observed
For the purpose of examining how the flow of the water level. So at first, we set the coefficient of ground
ground water vary along the direction and continuity permeability by using both two-dimensional steady
of element in the hydro-geological structure, we car- analysis and two-dimensional unsteady analysis that
ried out a three-dimensional seepage flow analysis by is a sensitivity analysis for melting snow water. Sec-
setting the slide surface and joint with crack as aspect ondly, we set the coefficient of permeability of the
crack, and collecting boring pipe and drainage pipe as slide surface, joint with crack and high-permeability
line crack. zone by using three-dimensional unsteady analysis.
The characteristics of our three-dimensional seep- In this paper, in OODAIRA Landslide area where
age flow analysis are as follows. landslides were reactivated by melting snow, we car-
At first, we examined a three-dimensional model- ried out a three-dimensional seepage flow analysis
ing of seepage flow analysis reflecting not only aspect reflecting the hydro-geological structure with aspects
cracks of the slide surface and joint with crack but also cracks. We then examined not only the validity of
high-permeability zone. a three-dimensional seepage flow analysis model by
We made modeling of the slide surface and joint comparing investigation between analysis value and
with crack as a joint element and high-permeability observed value and but also examined effectiveness of
zone as a series of solid elements with high- before-and-after construction of drainage facilities in
permeability. Also, we made modeling of well facility melting snow period.

1023
Figure 1. Detailed map of study region.

2 OUTLINE OF OODAIRA LANDSLIDE AREA 2.2 Geological structure

Geological structure in the landslide area consists
2.1 Outline of study area
of alternate strata of sandstone and mudstone of
OODAIRA Landslide area is located in 8 kilometers KURUMA group, and pyroclastic materials are found
away from OTARI village in NAGANO prefecture in the raising place such as the crest upside and
in NNW direction. Its geometry is a large horseshoe KAMISUKI stock farm.
geometry that the northern end is flat in KAMISUKI Alternate strata of sandstone and mudstone is rela-
stock farm and TUCHIZAWA River runs along the tively hard in the stream bed of TUCHIZAWA River
southern end. and is crushed considerably on the slope of both sides
A1 block where the landslide movement is active of the river, particularly interbedded coal-seam group
exists in the designated landslide prevention area (see is in rock-fragmented state and weathered.
Figure 1). The strike of bedding plane which were observed in
From middle to upper-side of A1 block is a gentle the scarp zone and the observation well runs in about
slope where slope angle is 20–30◦ with patches of flat EW direction and the angle of bedding plane is 20–40◦
terrain in some places. In the lower-side, the steep and agrees with the stratum angle seen on the slope of
escarpment with slope angle of 40–50◦ is seen toward TUCHIZAWA River.
TUCHIZAWA River. Looking at the failed slope in the lower side of A1
The steep escarpment forms a collapse area. This block, there is a layer interface between sandstone-
collapse caused by a heavy rainfall occured in the mudstone that is in fragmented rock state and con-
lower part of the slope of A1 block in 1995 and the tains relatively fresh sandstone-mudstone although has
collapse area was further expanded by melting snow multiple crack. The height of layer interface comes
in 1999. gradually down from north to south.

1024
 right-side wall (joint with crack) and its outer high-
permeability zone, inflow from upper slope, infiltra-
tion from ground surface, and so on. And it is known
that the left-side wall and the outer ground have capped
rock structure and bedding plane is inclined toward the
right-side wall (see Figure 3).
The landslide area has underground hydrology char-
acteristics where melting snow and rainfall produces
groundwater with pressure through the slide sur-
face. We made modeling of hydro-geological structure
by connecting levels of stratal layers at bore holes
sites.
Figure 2. Critical water pressure (BV12–4).
Details of a method for making stratal struc-
ture model from boring survey data is reported in
The strike and angle of bedding plane of the YAMADA (2006).
layer interface agrees with those of KURUMA group When we carry out FEM seepage flow analysis, it
(E-W30◦ –40◦ S). That is to say, it is considered that is necessary that the slide surface, joint with crack,
the landslide has dip slope structure regulated by bed- drainage facilities (well, drainage pipe) are expressed
ding plane of bed rock from left-side wall to end site. by appropriate elements.
This is reinforced by boring survey and some boring During our analysis, we kept in mind to make mod-
observation. Existing of regulatory aspect is showed by eling aspect cracks and well facilities to maintain exact
reading of aerial photograph and homogenious slope of form and size as much as possible.
N-S32◦ E in right-side wall of the landslide runs
straight. 3.2 Setting of the slide surface and joint with crack

2.3 Subsurface hydrology and activity The slide surface is a continuous plane of main dis-
of the landslide placement shear planes and a zone with constant
thickness (SHIN 1995).
An example of relationship between landslide and We will treat the slide surface as aspect crack phys-
ground water flow is shown in Figure 2 (BV12–4). ically where ground water flows mostly along plane
From the figure, it is clear that the strain at GL- and in FEM analysis, the slide surface is expressed as
22.9 m on the strain-gauge change significantly since joint element with constant thickness. Joint element
6 March 2002. with thickness is shown in Figure 4. Joint element with
In melting snow period many strain gauges indi- thickness has only value of thickness as a parameter
cated noticeable accumulation of strain. Observation (although in Figure 4 nodal point seems apart, but in
by GPS is performed since 2003 and the landslide fact nodal point is overlapped).
movement was recorded in melting snow period. The
variation volume of the strain in melting snow period
tends to calm down yearly by the increase in the num-
ber of well facilities. The direction of the landslide
movement agrees with the direction of the longitudinal
field line.
From the past observation, rainfall in non-snow
period does not exceed critical water level. Only in
melting snow period the landslide movement becomes
active.
Namely, melting snow affects the increase of water
pressure rather than rainfall.

3 THREE-DIMENSIONAL FEM SEEPAGE

FLOW ANALYSIS

3.1 Modeling of three-dimensional

hydro-geological structure
From the result of past survey, supply resources of
the groundwater include seepage from scarp zone, Figure 3. Analytical model.

1025
 Table 1. Coefficient of permeability of geologic layer and
cracks.

Figure 4. Joint element with thickness.

4 SETTING OF THE COEFFICIENT OF coefficient of permeability of high- permeability zone

GROUND PERMEABILITY AND CRACK as 1×10−4 m/s.
ELEMENT

4.1 Setting of ground the coefficient of permeability 5 TREATMENT OF DRAINAGE FACILITIES

IN 3D FEM ANALYSIS
Boundary conditions of analysis area were set as
follows: 5.1 Setting of well
1. Upper side as fixed boundary condition that keeps We made modeling of 8 wells (No.1–No.8), drainage
constant water pressure, pipes in order to reproduce drainage facilities.
2. River as seep boundary aspect condition, Although the well is essentially in circular form,
3. Valley as boundary condition where the flow vol- modeling is very simplified if the well is modeled as
ume is 0. square section, because solid element of the ground is
square form.
Because parameter used for seepage flow analysis As a result of comparing investigation about cir-
is not test value, we gave that unsaturated parameter, cular section and square section of equal areas, we
specific storage coefficient, saturated volume water were able to obtain same results in case both sections.
content and minimum water holding capacity are con- Therefore we made modeling of the well as square
stant and specific storage coefficient = 1 × 10−0.4 , section.
saturated volume water content = 0.4, minimum
water holding capacity = 0 in order to improve 5.2 Setting of collecting boring pipe
computational performance. and drainage pipe
Coefficient of permeability of ground layer is
shown as Table 1. We made modeling of collecting boring pipe and
drainage pipe as a line crack model. A line crack model
has permeability only along crack direction and there
is no outflow from one-dimensional crack element to
4.2 Setting of crack element
surrounding ground element.
Thickness and coefficient of permeability of the In the landslide area, volume of melting snow is
slide surface were respectively given with 15 cm, considerable and the distribution of melting snow
2 × 10−3 m/s for joint element. Also thickness joint volume is not uniform and because collected water vol-
element with crack is decided as about 5 cm and its ume from collecting boring pipe is different for each
coefficient of permeability were given as 2×10−4 m/s setting direction. Therefore we gave the collecting bor-
because it is considered its coefficient of permeability ing pipe coefficient of permeability that is appropriate
is smaller than the slide surface. to observed collecting volume of well.
Because the coefficient of permeability in high- Also because the gradient of drainage pipe that
permeability zone is small than the coefficient of connects wells and discharges groundwater to sur-
permeability of joint with crack (2 × 10−4 m/s) and face is low, we modeled drainage pipe as line crack
is larger than the coefficient of permeability of the that section areas are different as collecting bor-
ground (5 × 10−6 − 5 × 10−4 m/s), hence we gave the ing pipe.

1026
6 RESULT OF 3D SEEPAGE FLOW ANALYSIS
AND ITS VALIDATION

6.1 3D steady seepage flow analysis

of hydro-geological structure model
We carried out a three-dimensional steady seepage
flow analysis for the following cases.

1. In case of setting the slide surface and joint with

crack,
2. In case of setting high-permeability zone with joint,
3. In case of setting all aspect cracks.

Decreasing values of pore water pressure in II-line

are largest where the slide surface and joint plane are
crossing. The elevation of this part is lowest among all
lines and the gradient of this part is lowest among all
lines.
As a result of the survey, we set high-permeability
zone in between A-line and B-line as an upper-site
and D-line as lower-site. But effective area of high-
permeability zone is seen in lower side of the middle
point of A-line and B-line.
Figure 5. Results of three-dimensional steady analysis.
This is the inflow point of groundwater thorough
rock in upper-side of the slope and surface water
through crack. In case of field model setting all aspect
cracks decrease values of pore water pressure is largest
in between II-line and IV-line.
Because slide surface becomes deeper toward
right-side and right-side is made of hard bed rock,
hydro-geological structure that groundwater collects
in between II-line and IV-line reflects field hydro-
geological regulated condition.

6.2 3D unsteady seepage flow analysis

in before-and-after construction
of drainage facilities
We examined a three-dimensional seepage flow analy-
sis before-and-after construction of drainage facilities
in melting snow period.
5 wells were completed in September 2004 and
additional 3 wells were completed in December 2005
(8 wells in total).
Both analysis values in BV13-1 and BV13-5 mostly
match to the chronological water level. Also the
decrease of pore water pressure after construction of
the well in BV13-1 is larger than that of BV13-5.
It can be said that analysis result can explain varia-
tion of chronological water level against melting snow
volume, considering that melting snow volume is not
uniform in reality and modeling of geological struc-
ture in the field has restriction while observation data Figure 6. Effect of well facilities on amount of melted snow
can contain errors to some extent. during the melting season of 2006.

1027
7 EVALUATION OF SLOPE SAFETY USING
3D STABILITY ANALYSIS

7.1 3D analysis method

UGAI (1987) proposed a three-dimensional stability
analysis by the simplified Janbu method applicable
to slope that has arbitrary terrain and slide surface
profile. We evaluated slope safety by calculating three-
dimensional safety factor using the three-dimensional
Janbu method for chronological water pressure that
is derived from a three-dimensional seepage flow
analysis.
When performing a three-dimensional stability Figure 7. Effect of drainage facilities (calculated using data
analysis it is necessary to use the profile of the sur- from the melting season of 2006).
face and slide surface, three-dimensional mesh data
of water pressure and shear strength of the soil. Water
pressure plane can be provided by an isopleth line map 8 CONCLUSION
of water pressure produced from water pressure data of
the element derived by FEM analysis. We made three- We carried out a three-dimensional seepage flow
dimensional mesh data of ground surface, slide surface analysis considering cracks for OODAIRA Landslide
and water pressure plane by using GIS software. Here, reactivated during melting snow period and verified
mesh interspace is 5 m reflecting microtopography of availability of a three-dimensional seepage flow anal-
the landslide block. ysis reflecting hydro-geological structure with aspect
cracks.
7.2 Shear strength of slide surface We presented it was possible to reproduce the ris-
ing height of pore water pressure in three-dimensional
Angle of shear resistance is derived as ϕ = 8.53◦ from steady and unsteady analyses. Especially we were
the result of slide surface shear test. The landslide area able to derive the result that matches to observation
has a peculiar geological structure that the slide surface value by using a model reflecting real hydro-geological
is regulated by and inclined toward the right-side wall. structure model. We carried out a three-dimensional
And so we treat the resistance part of the slide surface stability analysis based on positional information of
as cohesion c’. water pressure plane derived from a three-dimensional
The landslide has been reactivated by accumulated seepage flow analysis and were able to derive the result
strain due to rapid increase of melting snow volume suitable to variation of strain value.
since 6 March 2002. As a result of calculating c’ where However, we could not absolutely confirm analysis
safety factor of the landslide slope in 6 March 2002 is value to observed value, though we could reproduce
defined as 1.0, c’ is derived as c’=18.11 kPa. water pressure by the analyses. The reasons can be
explained as follows.
7.3 Evaluation of drainage facilities 1. There are in no small part observation bore hole of
In the previous section we calculated safety factor of water level that is unable correspond to theoretical
the slope in the three different periods. When time variation of melting snow volume. For it is inher-
frame differs, melting snow volume also varies. So we ently unreasonable to use melting snow of only one
calculated safety factor of the slope by using the same place when chronological data of melting snow is
melting snow volume in order to evaluate the levels of different each place.
construction effects of the drainage facilities. 2. We can hardly reproduce water pressure distribu-
Figure 7 shows the result of calculating safety factor tion by correcting the coefficient of permeability
of the slope for case A, case B and case C by using and hydro-geological section in two-dimensional
the same melting snow volume from 1 March 2006 to analysis. However, there is a limit to correct the
9 May 2006. coefficient of permeability because water flow
Safety factor of the slope with 5 wells increases toward cross-section occurs in a three-dimensional
by 0.01–0.05 compared to that with no wells. Safety analysis.
factor of the slope with 8 wells further increases by In the future, if we can calculate safety factor of
0.004–0.005. the slope using water pressure observed by automatic

1028
observation system, it will be possible to explain con- SHIN, J. (1995). Landslide engineering – new topics —,
struction effect of drainage facilities in real time by pp. 3–9, sankaido.
using easy-to-understand evaluation term of safety YAMADA, M., YAMAZAKI, T., YAMASAKI. (2000).
factor. Relationship between groundwater flow and landslides,
landslide, Vol. 36, No. 4, pp. 22–31.
YAMADA, M. SHINGO, S. (2006). Construction of land-
slide GIS, landslide, Vol. 42, No. 4, pp. 51–62.
REFERENCES

UGAI, K. (1987). Three-dimensional slope stability analysis

by simplified Janbu method, landslide, Vol. 24, No. 3,
pp. 8–14.

1029
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Back analysis of soil parameters: A case study on monitored displacement

of foundation pits

B. Yan, X.T. Peng & X.S. Xu

School of Civil Engineering & Architecture, University of Ji’nan, Ji’nan, Shangdong, China

ABSTRACT: The elastic-reaction method has been widely used in practical engineering for its convenience and
validity. The solutions calculated with the elastic-reaction method are effective on condition that the m-values and
proportional coefficients of lateral subgrade reaction are consonant with the field soils. The befitting m-values
are impossibly obtained in laboratory soil tests. The paper presented an effective back-analysis model to obtain
befitting m-values. Since practical m-values are found, the lateral displacements of foundation pits are well
predicted.

1 INTRODUCTION 2 THE MODEL OF BACK-ANALYSIS

With high buildings rising, retaining and protect- 2.1 The objective function of back-analysis
ing of foundation excavation have been an important
Based on field data, the objective function is
phase in constructions in crowded cities. Piles in
expressed as
rows, diaphragms, soil nailing walls, sail anchors and
cement-soil walls and so on are often used in deep N  2
excavation engineering (JGJ120-99) (GB5007-2002).  si
J = min (1)
All academic solutions, including bearing forces or
i=1
sit − 1
displacements of sides of foundation pits, should be
accurate to ensure the security of engineering and sur-
roundings around foundation pits. One of the major where N = total of monitoring points; si = calculated
problems facing us is that whether soil parameters lateral displacement at ith monitoring point; sit =
accord well with the fields or not (Arai et al 1983, monitored lateral displacement at ith monitoring point.
1986). Figure 1 shows the analysis of search soil parame-
The paper presents a sort of finite elements method ters, m-values.
(FEM) adopted in calculating internal forces and lat-
eral displacements for piles in row or diaphragms 2.2 The model for the elastic-reaction method
(Milligan, 1983). The method has been widely used in
practical engineering for its convenience and validity. Based on the Technical Specification for Retaining and
The solutions calculated with the FEM are effective Protecting Building Foundation Excavation of China
on condition that the m-values, proportional coeffi- (JGJ120-99), Figure 2 shows the horizontal loads,
cients of lateral subgrade reaction, are consonant with active earth press, exerted on the bearing piles. The
the field soils. The befitting m-values are impossibly active earth press above bottom of the pit can be esti-
obtained in laboratory soil tests, but in fields. Field mated with Rankin or Coulomb’s Earth Press Theory.
data, such as lateral displacements or internal forces For crushed stone soil and sandy soil above the
of bearing piles, are utilized to estimate the m-value to ground water surface, the standard active earth press
comport with practical fields (Arai et al 1984, 1987). is formulated as,
That is back-analysis, which is a valuable method to 
get befitting parameters, m-values. eajk = σajk kai − 2cik kai (2)

1031
 where
Start
σrk = rmj zj
Input monitored
displacements σ0k = q0 (5)

Input initial b0
σ1k = q1
m-values b0 + 2b1
All the functions mentioned above may be well
Run elastic- known by consulting JGJ120-99.
reaction program The soils inside of pits simulate the Winkler soil
model for bearing piles. The model is composed of
Calculate objective lumped horizontal springs. It’s assumed that the mod-
function ulus of the springs rise at a proportional coefficient,
value, along the piles from the top down.
Judgment of the No Continue running
precision objective back-analysis 2.3 Back-analysis of m-value
Ok The back-analysis is actually a mathematical opti-
Output of practical mization problem. The simplex method is used in the
m-values paper. Based on Eq. (2), the monitored lateral displace-
ment, sit , should be all-sided and accurate along the
Figure 1. Process of back-analysis. piles. Before back analyzing practical m-values, ini-
tial values will be estimated by one’s experience. The
calculated lateral displacements, si , may be estimated
with the elastic-reaction method program. Obviously,
Bearing piles si is often unequal to sit . After running the program to
cycle time again, that si deviates sit is getting smaller
Active earth press and smaller. When the objective value, J , is much
smaller than a satisfying value (ε) at that moment,
the m-values back-analyzed are well consonant with
Bottom of pit practical soils.

Winkler
3 APPLICATION IN PRACTICAL
soil model
ENGINEERING

3.1 Outline of the project

Figure 3 shows a shallower foundation pit, with 90.0 m
Figure 2. The model for the elastic-reaction method. in length, 54.0 m in width and 5.0 m in depth, including

Cement deep mixing piles

under the ground water surface, it is formulated as,


eajk = σajk kai − 2cik kai
+ [(zj − hwa ) − (mj − hwa )ηwa kai ]γw (3)

for silt and clay, the same formula in (2), where σajk =
vertical standard stress outside of foundation pits. It
can be shown below,
Monitoring point

σajk = σrk + σ0k + σ1k (4) Figure 3. Plane of the foundation pit.

1032
a deeper pit, with 39.6 m in length, 35.0 m in width and Table 2. m-value back-calculated and comparing lateral
10.7 m in depth. The water plane is 1.5 m below the displacements back-calculated with lateral displacements
ground level. The artesian aquifer with 14 m hydraulic monitored for one united soil layer.
head is found at the—18.0 m elevation.
m-value Displacements Displacements
The bracing system is composed of piles in row, Monitoring calculated calculated monitored
top beams on the top of bearing piles and internal points (kN/m4 ) (mm) (mm)
struts. The piles in row are formed with bored cast-
in-place piles, with its diameter 1000 mm. The piles A 3300 26.42 22.92
tip arrive −24.5 m elevation. To prevent ground water B 2170 29.20 27.20
from seeping into the deep foundation pit, water tight C 3200 18.11 21.11
screens, with a diameter of 600–700 mm, are set near D 2600 24.44 27.94
the bearing piles outside. The shallower pit is consol-
idated with cement deep mixing piles, with 4.0 m in
width, 15 m in depth, which can also act as water tight
screens.
Figure 4 shows a geological profile composed of parameters are feasible on condition that the behav-
five distinct soil layers. Table 1 shows their corre- iors of soil layers are approximately the same. Only
sponding index properties. approximate soil layers can be united so that the back-
In the procedure of construction, the ground water analyzed results are minimally influenced. Compare
and artesian aquifer are availably controlled at 1.0 m ‘‘one united soil layer’’ with ‘‘approximately united
under the bottom of the pit. soil layers’’ as follows.

3.2 Back-analysis of m-value in practical (1) One united soil layer

engineering
Figure 3 gives the positions of monitoring points,
The m-value used in the elastic reaction method is tra- such as A, B, C, D. Lateral displacements at monitoring
ditionally estimated one value, that is to say, several points, sit which can be obtained by deflection incli-
soil layers are treated as one same layer. The simple soil nometers, are practical displacements when the pit is
excavated at the level of −10.7 m. Introducing field
-0.5 data into Eq. (1), the practical m-values are searched
Internal strut -2.5 by back-analysis program.
Clay layer 1 Table 2 shows the solutions back-analyzed. When
Bearing -7.0 the passive earth press areas are supposed to be the
-10.7 same soils, the lateral displacements back-analyzed
Slush layer 2 have relation to the total monitoring points.
-16.8
-19.5 Silty clay 3
(2) Approximate soil layers united

-22.8 Silty clay layer 4 In Figure 4, the soil layer ① and ② belong to softer
soil, uniting both of them. The Soil layer ③, ④ and ⑤
Silty clay layer 5
belong to stiffer soil, uniting three of them. Now, five
-24.5 layers become two layers. Table 3 shows the solutions
back-analyzed in the case.
Figure 4. Geological profile.
Conclusions back-analyzed are as follows:
Table 1. Index properties for soil layers. The monitoring points should be set as many as
possible. Since behaviors of soil layers in the field
Soil Height ω have large distinctness, unite five layers into two
layers (m) % Ip IL γ φ c layers but one. The top layer is softer soil and
the bottom one is stiffer layer. Obviously, the top
Clay layer 6.5 35.2 17.9 0.81 18.7 10.6 24.3 layer has a smaller m-value and the bottom layer
Slush layer 9.8 43.4 20.2 1.05 17.9 9.5 21.3 has a larger one. By means of the m-value back-
Silty clay 2.7 27.7 12.0 0.89 19.4 15.2 24.8
Silty clay 3.3 27.4 14.5 0.60 19.7 16.3 50.6
analyzed, the lateral displacements calculated with
Silty clay 11.8 22.5 11.6 0.52 20.8 21.2 56.0 the elastic-reaction method are well consonant with
displacements monitored.

1033
Table 3. The m-values back-calculated and comparing on condition that the m-values, proportional coeffi-
lateral displacements back-calculated with displacements cients of lateral subgrade reaction, are consonant with
monitored for approximate soil layers united. the field soils. The paper presents an effective back-
analysis model. Only approximate soil layers can be
m-value
united so that the results back-analyzed are influenced
calculated in a minor degree. Monitoring points should be set as
(kN/m4 ) Displacements Displacements many as possible to ensure reliable results. Since befit-
Monitoring calculated monitored ting m-values are found, the lateral displacements for
points Above Under (mm) (mm) foundation pits can be well predicted.

A 1020 4500 25.01 22.92

B 1300 3850 27.60 27.20 ACKNOWLEDGMENTS
C 1040 4160 19.61 21.11
D 880 4660 27.94 27.94
The supports from Natural Science Foundation of
Shandong Province (G0637) and University of Jinan
(B0421) are appreciated.
Table 4. Comparing lateral displacements predicted with
lateral displacements monitored.
REFERENCES
Monitoring Displacements Displacements
points predicted (mm) monitored (mm) Arai, K., Ohta, H. & Yasui, T. 1983. Simple optimization
techniques for everlasting deformation module from field
A 18.92 22.92 observations. Soils and Foundations. 23 (3):107∼113.
D 25.94 27.94 Arai, K., Ohta, H., Kojima, K. & Wakasugi, M. 1986. Appli-
cation of back-analysis to several test embankments on
soft clay deposits. Soils and Foundations. 26(2):60∼72.
Arai, K., Ohta, H. & Kojima, K. 1987. Estimation of nonlin-
ear constitutive parameters based on monitored movement
4 FORECASTING THE LATERAL of subsoil under consolidation. Soils and Foundations.
DISPLACEMENT AT THE NEXT STAGE 27(1):35∼49.
Arai, K., Ohta, H. & Kojima, K. 1984. Estimation of soil
The lateral displacements change with the schedule parameters based on monitored movement of foundations.
of construction. Those future lateral displacements 24(4):95∼108.
known beforehand can effectively direct the next stage China Academy of Building Research. 1999. Technical
of construction. Based on displacements monitored at Specification for Retaining and Protection of Building
monitoring points A and D at a certain depth of the pit, Foundation Excavations. Beijing: China Architecture &
befitting m-values has been reached. Then the lateral building press.
Milligan, G.W.E. 1983. Soil deformation near anchored
displacements of the pit at the design elevation will be sheet-pile walls. GeotechniqueNo.1: 44∼55.
calculated by means of the elastic-reaction method. Ministry of Construction of the PRC. 2002 Code for design
Table 4 shows that the predicted lateral displace- of building foundation. Beijing: China Architecture &
ments are almost identical to the monitored one. building press.

5 CONCLUSIONS

The elastic-reaction method has been widely used in

practical engineering for its convenience and validity.
The solutions calculated with the method are effective

1034
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

3D finite element analysis on progressive failure of slope due to rainfall

G.L. Ye
Department of Civil Engineering, Shanghai Jiaotong University, China

F. Zhang
Department of Civil Engineering, Nagoya Institute of Technology, Japan

A. Yashima
Department of Civil Engineering, Gifu University, Japan

ABSTRACT: A soft rock slope in Tokai-Hokuriku Expressway of Japan failed due to a heavy rain. The
three-stage failure procedure indicates that it was a typical progressive failure. In this study, by using a modified
elastoplastic model with Matsuoka-Nakai failure criterion, which not only can describe the strain-hardening
and strain-softening behavior of soft rock but also can take into consideration the influence of the intermediate
principle stress, a 3D soil-water coupled finite element analysis is conducted to investigate the mechanics of the
progressive failure. The 3D shape of the slope, the geological conditions and the initial ground water level are
delicately considered during 3D modeling. The change of the ground water level during rainfall is simulated
by increasing the water heads of elements from initial level to the ground surface. In the discussions of the
calculation results, the characteristics of slope failure, such as the development of shear strain, the deformation
of ground, the propagation of shear band and the failure zone are discussed in detail. It is found that 3D soil-water
coupled analyses based on the modified elastoplastic model can simulate the progressive failure of a slope to an
engineering acceptable accuracy.

1 INTRODUCTION movement of underground water, careful ground sur-

vey is necessary to acquire a satisfactory accuracy of
Slope failure induced by heavy rain is a very impor- the analysis.
tant but difficult problem for geotechnical engineers. In this paper, a slope failure due to heavy rain
Progressive failure, which is a typical time-dependent is simulated by 3D soil-water coupled finite element
behavior, is often observed in slope failure. Clarify- analysis. The observed failure was a typical progres-
ing the mechanical behaviors of progressive failure is sive failure and the heavy rainfall is the main reason for
indispensable for predicting when and where the slope the failure. Zhang et al. (2001) conducted a 2D soil-
failure will occur. From the viewpoint of soil mechan- water coupled finite element analysis on this slope,
ics, the soil skeleton-pore water coupled behavior which is based on an elastoplastic model proposed by
and/or the inherent viscosity of soil skeleton of the soft Oka and Adachi (1985).
rock should be considered when dealing with long- The model proposed by Oka and Adachi (1985),
term stability of a soft-rock slope concerning rainfall. however, has a Cam-Clay type failure criterion, which
Accordingly, a soil-water coupled numerical analysis as had been pointed out that it could not predict cor-
and/or an elasto-viscoplastic model is necessary for rectly the strength of geomaterials under general load-
simulation. ing condition except axisymmetrical condition. Zhang
In most cases, the failure pattern is strongly related et al. (2003) proposed a modified model by intro-
with the 3D profile of the slope. However, numer- ducing Matsuoka-Nakai failure criterion (SMP failure
ical simulation on progressive failure by 3D FEM is criterion) to that model. In current paper, based on the
seldom conducted due to the complexity of the numer- modified model, 3D soil-water coupled finite element
ical modeling such as the geometry and geological analysis is conducted to investigate the progressive
mesh modeling, determination of material parameters, failure of the slope. The aim of the research is to try to
and above all, the time consumption of calculation. establish an applicable way to assess the progressive
Especially in the case considering the effect of the failure of slope under 3D topographic condition.

1035
2 CASE HISTORY OF SLOPE FAILURE Cliff gimlet
Altitude (m)
22 0
Strong weathered rock (D) 0. 3
0. 9 /s
Weathered rock(CL~CM)
2.1 General description of the slope failure Boundaries of elastic wave speeds 1 .5 /s
Failure surface
initial underwater line
The failed slope locates between the Mino Interchange 3. 1 /s

and Minami Interchange of Tokai-Hokuriku Express- Retaining wall made of concrete blocks
1 50
way in Japan. The failure happened on September 22, Tokai-Hokuriku Express Way
1999, which was occasionally recorded by a journal-
ist who was on another failure site near the slope.
The video showed that it was a typical progressive
failure. Figure 1 shows the sequence of the observed 80
(a) Geologic profiles in central section
progressive failure, which is a three-stage failure, tak-
ing about 3 hours from the first stage to the final stage. Figure 2. Geologic profile and its modeling of the central
The detailed failure process is described as following: section plane of the slope.
About at 9:22 on Sept. 22, 1999, some cracks appeared
in the concrete-block frame wall and a great deal of
water gushed out of the drainage work. At 9:45, the
first stage failure occurred, the failure area was 45 m rain reached peak value within 20∼30 hours, depend-
wide and 50 m long. The second failure occurred at ing on the permeability of the ground. Therefore it is
12:20, the failure area expanded to 120 meters wide reasonable to think that the increase of the pore-water
and 70 meters long. The final failure occurred at 13:10, pressure due to the heavy rain is a main reason for the
continued from the second failure till the crest of the failure. Since the failure happened six years after its
slope, the final failure area was 120 m wide and 125 m completion, deterioration in the strength of the ground
long, with fallen soil of almost 110,000 m3 in volume. may be possible. In present analysis, however, only the
The maximum hourly rainfall depth, the 24-hour effect of rainfall is considered.
rainfall depth and the 9-day rainfall depth before the
failure amounted to 63 mm, 182 mm, and 619 mm,
respectively. These rainfall depths were recorded at 2.2 Geologic conditions of the slope
AMeDAS Mino observatory station of Japan Mete- As plot in Figure 2, the geologic information from
orological Agency (JMA), where is very near to the field investigations of the slope is summarized as
site of the slope failure. The failure happened 6 hours following:
after the maximum hourly rainfall depth was reached.
It is reported (Tamura and Matsuka, 1999) that exces- 1. Top layer of the ground is strongly weathered rock
sive pore-water pressure in a ground of slope usually that is about 5 m in depth on summit and a max-
started to increase at the time 10 hours after a heavy imum depth of 17 m in the middle-low part of
the failed slope. This layer, classified as D-grade
rock, is composed of weathered clay and fresh frag-
ments of rock, with an elastic-wave velocity of
(1) Before failure (2) First failure
(9:55) 0.3∼0.9 km/s. The failure surface formed within
To Mino IC this layer. (2) The layer beneath the top layer is a
weathered rock with a depth of 20 m∼25 m and an
m
elastic-wave velocity of 1.4∼1.6 km/s, classified
To Minami IC 150 as CL∼CM-grade rock.

3 NUMERICAL ANALYSIS CONDITIONS

(3) Second failure (4) Final failure
(12:20) (13:10) In the formulation of soil-water coupled analysis, the
finite element method is used for the spatial dis-
cretization of the equilibrium equation, while the finite
difference method is used for the spatial discretization
of the pore-water pressure in the continuity equation.
Figure 3 shows the finite element mesh adopted in
the analysis of the slope. The size of the ground is
300 m in width, 310 m in length and 214 m in height.
Figure 1. Observed progressive failure sequence of the The numbers of the node and 8-node isoparametric
slope. element are 8190 and 7000 respectively.

1036
 residual state in a conventional triaxial compression
Node: 8190 test) is dependent on the confining stress. Rf takes a
Element: 7000 relatively large value at small confining stress. For this
reason, the value of Rf of the strong weathered rock
will be given in 3 different values under different ini-
tial confining stress in such a way that, the closer the
place is to the surface, the larger the Rf will be. By this
Initial water level treatment, it is possible to overcome the discrepancy
mentioned above.

3.3 Material parameters

Figure 5 shows the comparison of stress-strain-
dilatancy relationships obtained from the theory
under drained triaxial and plane-strain compres-
sive/extensive conditions. It shows that the peak stress
difference (q = σ1 − σ3 ) of plane-strain condition is
larger than those of conventional triaxial compression
Figure 3. 3D mesh model. condition and the volumetric strain of plane-strain con-
dition is less than those of conventional compression
condition. In the figure, the material parameters of the
3.1 Boundary conditions
For displacement, it is fixed at the bottom in both
horizontal and vertical directions, only the normal Rf=15.0

direction of the side boundaries is fixed. Rf=7.0

For drainage condition, the ground surface is per- Rf=4.5

elastic
meable while the others are impermeable.

3.2 Initial conditions

Due to the difficulty to determine the initial stress field
for a soft rock ground, the initial stress fields in both
calculations are simply assumed to be a gravitational
field. Initial total water head, is directly given by the
value of underground water level measured from bore-
hole tests in such a way that the underground water Central section plane No.14 row section plane

level of the elements within one column are the same,

and that those elements above the underground water Figure 4. Central section and No.14 row section planes of
level are also saturated and possess the same value as the 3D mesh.
those of the elements below the underground water
4000 8 4000 8
level. Stress difference (q) Stress difference (q)
3000 6 3000 6
In the calculation, the first stage is to calculate the
Stress difference q (kPa)
Stress difference q (kPa)

Volumetric strain (%)

Volumetric strain (%)

Volumetric strain ( v) Volumetric strain ( v)

2000 4 2000 4
initial stress field; then a prescribed increment of total Compression Compression
1000 2 1000 2
water head is applied in the second stage and then the
0 0 0 0
calculations are continued to simulate the change of
-1000 -2 -1000 -2
E.P.W.P with a time interval of 2 sec/step. 0 2 4 6
Axial strain (%)
8 10 0 2 4 6
Axial strain (%)
8 10

Figure 4 shows the central cross-section of 3D 4000 8 4000

Stress difference (q)
8

Stress difference (q)

mesh. Because the initial mean effective stress near 3000 6 3000 6
Stress difference q (kPa)
Stress difference q (kPa)

Volumetric strain ( v)
Volumetric strain (%)
Volumetric strain (%)

Volumetric strain ( v)

the ground surface is very small in the initial stress 2000

Extension
4 2000
Extension
4

field calculation, some soil elements near the surface 1000 2 1000 2

may have already failed at initial stress state, if the 0 0 0 0

stress ratio at residual state is given in the same value -1000

0 2 4 6 8 10
-2 -1000
0 2 4 6 8 10
-2

Axial strain (%) Axial strain (%)

disregard the magnitude of its confining stress. This (a) Conventional triaxial condition (b) Plane strain condition
is not coincident with the real situation. According to
the laboratory tests (Adachi, T. and Ogawa, T., 1980), Figure 5. Comparison of stress-strain-dilatancy relation-
the stress ratio at residual state Rf (Rf : (dσ1 /dσ3 )f at ships under different loading conditions (σ3 = 200 kPa).

1037
Table 1. Material parameters of slope.
t=4000 sec t=6000 sec
Weathered rock Strong weathered
Parameter Unit (CL ∼CM ) Rock (D)
Shear strain 2I2

E (kPa) 3.0 ∗ 106 1.0 ∗ 106

ν 0.25 0.33
γ (kN/m3 ) 15.7 (γ  ) 25.6 (γsat )
k (m/sec) 1.0 ∗ 10−6 1.0 ∗ 10−6 t=8000 sec t=11450 sec

G 100.0 100.0
τ 0.050 0.050
b (kPa) 2000 1000
σmb (kPa) 20000 10000
α 1.00 1.00 Central section
D −0.30 −0.30 plane
Rf +∞ Upper 15.0
(Elastic) Middle 7.0
Bottom 4.5
Figure 6. Development of shear strain in the slope with time
(central section plane).

rock are the same as those of strong weathered rock

listed in Table 1 except Rf that takes a value of 5.0. t=4000 sec t=6000 sec
The physical meanings and detailed description of the
way to determine these parameters can be referred to
with the reference (Zhang et al., 2003).
In the numerical calculation, the change of pore-
water pressure due to a rainfall is simulated under the
assumption that after a heavy rain, the total water head t=8000 sec t=11450 sec
increases from the initial underground water level to
ground surface. And the increment of the pressure
water head in each element column is applied within
six hours.
10 cm

4 RESULTS AND DISCUSSION Figure 7. Development of displacement vector in the slope

with time (central section plane).
4.1 Investigation on overall behavior of ground
Figure 6 shows √ the changes in the distribution of the
shear strain ( 2I2 , I2 : second invariant of deviatoric move downhill along the shear band when the shear
plastic strain tensor). It can be seen that he shear strains strain reached a certain value, as shown in Figure 6.
developed very slowly in the first 8000 seconds, after Therefore, based on the direction of the displacement
then, the shear strains turned to develop rapidly until of the ground, it is possible to judge whether a slope
total failure. A relatively large shear strain occurred at is in the danger of failure.
the middle of the slope. Then a shear zone, in which Figure 8 shows the comparison of the predicted
larger shear strain developed, propagated from local shear band and the observed failure surfaces. In real-
area to neighboring areas and finally formed a shear ity, the slope failed in three stages, taking about
band. The shape of the shear band coincided with the 3 hours from the first stage to the final stage. Due to
boundary between the strongly weathered rock (D) the limitation of finite element method that based on
and rather hard rock (CL ∼CM ). It is found that if the the continuum theory, however, the multi-stage failure
maximum shear strain within the shear band reached a cannot be reproduced in simulation. However, the cal-
certain value, the failure abruptly happened along the culated shear band coincides well with the observed
shear band. final failure surface. Therefore, it can be said that
Figure 7 shows the change of displacement vectors present analysis can simulate the progressive failure
of the ground with time. The ground moves upward of a cut slope to a reasonable accuracy. On the other
at first 4000 seconds due to the floating force from hand, from the development of shear band, we can
the rising of the underground water, then it turned to judge the degree of the progressive failure.

1038
4.2 Investigation on the behavior of individual be achieved simultaneously, resulting in an increase
element of effective confining pressure in the element. On
the other hand, strain softening may happen in some
Figure 9 shows the behavior of element 3, where
neighboring elements, resulting in some redistribution
the formation of the shear band was ignited. Fluc-
of stresses due to the lose of strengths, these redis-
tuations were seen in the stress-strain relations and
tributed stresses have to be resisted by the element.
strain rate with time, that is, strain hardening occurs
As the result, the element will experience strain hard-
at first, and then softening occurs and followed by
ening again. The repetition of above process results
succeeding hardening again and fails at last. The rea-
in the fluctuation. In the figures, stress path and
son of the fluctuation can be explained in following
stress history path represent the trails of stress state
way: as strain softening develops in an element, dila-
and stress history state in different times. The stress
tancy occurs while the supply of pore water cannot
state may overpass the residual strength line in over-
consolidated state while stress history state cannot
overpass the residual strength line. If it reaches the
Bird view
line, it means that the soil reached residual failure
state.
The stress path and the stress-history path of ele-
Observed final failure zone
ment 3 are also showed in Figure 9. It can be
Observed final failure surface
seen that the shear stress and the shear stress-history
are decreasing throughout the whole process. The
same tendency can be seen for the mean effective
stress. When the stress history overpasses the residual
Section view strength line, the element fails.
Figure 10 shows the behavior of element 7, which
locates outside of the shear band. Different from the
Figure 8. Comparison of the calculated shear zone and the element located on shear band, the element did not
observed failure zone.
-4
-4
4.0 10
4.0 10 1/2
1/2

Shear stress ratio (t /t )

S N
Shear stress ratio (t /t )

Shear strain rate (2I )

S N

2
Shear strain rate (2I )

-5
3.0
2

10
-5 10
3.0

-6
10
-6 2.0 10
2.0

-7 -7
1.0 10 1.0 10

-8 -8
0.0 10 0.0 10
0 0.002 0.004 0.006 0.008 0.01 0 2000 4000 6000 8000 10000 12000 0 0.002 0.004 0.006 0.008 0.01 0 2000 4000 6000 8000 10000 12000
1/2 1/2
Shear strain (2I ) Time (sec) Shear strain (2I ) Time (sec)
2 2

-0.002 250 -0.002 250

Positive
v

Expansion
v

-0.001 125 -0.001 125

Volumetric strain

E.P.W.P. (kPa)

Volumetric strain

E.P.W.P. (kPa)

Expansion Positive
0.000 0 0.000 0

Negative
0.001 Compression -125 Compression Negative
0.001 -125

0.002 -250
0 2000 4000 6000 8000 10000 12000 0 2000 4000 6000 8000 10000 12000 0.002 -250
0 2000 4000 6000 8000 10000 12000 0 2000 4000 6000 8000 10000 12000
Time (sec) Time (sec) Time (sec)
Time (sec)
250 250

Stress path Central section plane Stress path Central section plane
Stress history path 200 Stress history path
200
of 3D analysis of 3D analysis
t ,t (kPa)
t ,t (kPa)

failure 150 almost stationary in

150
stress history path
S
S

*
*

100
S

7
S

100

50
Residual strength line
3 50
Residual strength line

0 0
0 50 100 150 200 250 0 50 100 150 200 250
* *

Figure 9. Behavior of element locate on shear band (in the Figure 10. Behavior of element outside of shear band (in
central section plane). the central section plane).

1039
show fluctuation in stress-strain relations. And the No.14 row
magnitudes of the fluctuation in shear strain rates are section plane H1 H2 H5 H6
H3 H4
almost the same, which indicates that the soil-water
interaction of the element outside the shear band is
not so violent as those on the shear band. Needless to
say, the predicted stress history paths show the same
result that element did not fail.
Above discussion shows that the behaviors of ele- 0.010 0.015
ments inside and outside of shear band are different. 3306
H3 3656
H4

1/2

1/2
0.008 2955
H2 H5
4005
It can be concluded that accurate prediction of stress-

Shear strain (2I )

Shear strain (2I )

2

2
2610
H1 H6
4360
0.010
dilatancy relation will play a very important rule in 0.006

describing the progressive failure of a slope related to 0.004

0.005
soil-water coupled problem. This is why a modified 0.002
elastoplastic model (Zhang et al., 2003) is adopted in
the analysis. 0.000
0 2000 4000 6000 8000 10000 12000
0.000
0 2000 4000 6000 8000 10000 12000
Time (sec) Time (sec)
Figure 11 shows the development of shear strain in -4 -4
10
the elements along the shear band of the slope in the H3
3306
10
H4
3656

1/2

1/2
middle plane of the failure zone. From the change of H2
2955 H5
4005

Shear strain rate (2I )

Shear strain rate (2I )

-5
H1 H6

2
-5
10

2
2610 10 4360
both the strains and strain rates, it can be seen that the
accelerating stage first occurs at element 3, and then 10
-6
10
-6

it propagates toward two sides. -7 -7

10 10
Figure 12 shows the development of the shear strain
in the elements along the shear band in an orthogonal 10
-8
10
-8
0 2000 4000 6000 8000 10000 12000
0 2000 4000 6000 8000 10000 12000
vertical section. From the change of both the strains
and strain rates, it can be seen that the accelerating
stage first occurred at element H3 and H4, and then it Figure 12. Propagation of failure in horizontal plane.
propagated toward two sides, leftward in the sequence

of element H2 to H1 and rightward in the sequence

5 of element H4 to H5. Therefore, a progressive fail-
Central section plane of 3D analysis
4 ure in two orthogonal vertical planes, a typical 3D
behavior, can be clearly simulated. From this figure, it
3
2 can be seen that 3D calculation can give a more com-
1 prehensive description of the mechanical behavior of
progressive failure than 2D calculation can do.
0.015 0.015

3515
3 3
3515
1/2

1/2

3656
2 4
3537
Shear strain (2I )

Shear strain (2I )

1
3586 5
3552
2

0.01 0.01
5 CONCLUSIONS
0.005 0.005
In this paper, based on an elastoplastic model consid-
ering the influence of intermediate stress and strain
0
0 2000 4000 6000 8000 10000 12000
0
0 2000 4000 6000 8000 10000 12000
softening, 3D soil-water coupled finite element-finite
Time (sec) Time (sec) difference analysis was conducted to simulate a large-
10
-4
10
-4
scale slope failure due to heavy rain. From the
3
3515 3515
3
analyses, the following conclusions can be given:
1/2

2
1/2

3656 4
3537
-5 1 -5
Shear strain rate (2I )
Shear strain rate (2I )

10 3686 10 5
3552
2
2

10
-6
10
-6
1. By adopting 3D soil-water coupled finite element
-7 -7
methods based on a suitable elastoplastic model for
10 10
soft rock, and properly 3D modeling, the progres-
10
-8
10
-8 sive failure of a real slope due to heavy rainfall can
0 2000 4000 6000 8000 10000 12000 0 2000 4000 6000 8000 10000 12000
be simulated to an engineering acceptable accuracy.
2. From the stress-strain relations and the time
Figure 11. Propagation of failure in the central section changes of strain rate in soil elements along the
plane. shear band, it is found that the formation and the

1040
 propagation of the shear band were directly caused REFERENCES
by the strain-softening behavior of the ground.
3D analysis is able to describe the propagation in Adachi, T. & Ogawa, T. (1980): Mechanical properties and
two orthogonal vertical planes, being much more failure criterion of soft sedimentary rock, Proceedings of
realistic than 2D analysis. In other words, 3D anal- JSCE, 295, 51–62 (in Japanese).
ysis can give a more comprehensive description of Oka, F. & Adachi, T. (1985): A constitutive equation of
geologic materials with memory, Proc. 5th Int. Conf.
the progressive failure of a slope. on Numerical Method in Geomechanics, Balkema, 1,
3. An overall failure of a slope was ignited by the 293–300.
propagation of shear band within the ground of Tamura, E. & Matsuka, S. (1999): Automatic measurement of
slope. pore water pressure in the hard-rock slope and the sliding
4. Accurate prediction of stress-dilatancy relation will weathered-rock slope-field survey in mountainous region
play a very important rule in describing the progres- in Shikoku Island, Japan. Proc. of Int. Conf. on Slope
sive failure of a slope related to soil-water coupled Stability Engineering (IS-Shikoku), Matsuyama, Japan,
problem. Balkema, 1, 135–140.
5. Present analysis cannot simulate the process of Zhang, F., Yashima, A., Sawada, K., Sumi, T., Adachi, T. &
Oka, F. (2001): Numerical analysis of large-scale slope
multi-stage progressive failure observed in the real failure, Computational Mechanics—New Frontiers for
failure event, the predicted failure surface coin- New Millennium (APCOM’01), Australia, Elsevier, 1,
cides only with the observed failure surface at final 527–532.
stage. This is due to the discrepancy that in contin- Zhang, F., Yashima, A., Ye, G.L., Adachi, T. & Oka, F.
uum solid mechanics, medium is considered to be (2003): An elastoplastic strain-softening constitutive
continuum while in reality, after its first-stage fail- model for soft rock considering the influence of inter-
ure, the failed soils had already gone away from the mediate stress, Soils and Foundations, 43 (5), 107–117.
slope.

1041
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Block-group method for rock slope stability analysis

Zixin Zhang, Ying Xu & Hao Wu

Department of Geotechnical Engineering, School of Civil Engineering, Tongji University, Shanghai, China
Key Laboratory of Geotechnical & Underground Engineering, Ministry of Education, Tongji University, Shanghai,
China

ABSTRACT: This paper provides an extension for the stereo-analytical method, called ‘the block-group
method’. The stereo-analytical method can consider both convex and concave blocks, but a group of blocks, if
considered as a whole, might indeed be unstable and more dangerous compared with individual key blocks. It
is revealed that the block group must satisfy the conditions: the block group contains at least one key block;
neighboring finite blocks have common face with the key block or key block-group. Block-group types can be
determined with the stereo-analytical block method. Then the kinetic energy law is implemented to analyze the
movement behavior of each key block or key block-group. Finally, a rock slope project is demonstrated to prove
that the block-group method yields more realistic results than the basic key block method.

1 INTRODUCTION block geometry and define if the blocks are finite

or infinite. The second is to identify which finite
Rock mass is the total in-situ medium containing bed- block intersects the excavation can form a remov-
ding planes, cracks, faults, joints, folds and other able block. The third step is to perform to stabil-
structural features. Rock slope, which is divided into ity analysis to distinguish the stable and unstable
many rock blocks by various discontinuities, is of blocks.
static equilibrium state in nature stress field. But However, the key block method only considers the
some rock blocks may slide along structural features key blocks. If no such blocks exist, the method con-
or free falling, when rock slope is excavated and cludes that the rock mass is stable. Yet in reality, group
disturbed. Even, large-scale rock blocks gradually of blocks if considered as a whole, may indeed unsta-
move. To a great extent, rock slope stability depends ble. Few authors have actually attempted to improve
on mechanical characteristic, geometrical characteris- the key block method. Wibowo (1997) sought to take
tic and distribute orderliness of discontinuities. Now, the secondary key blocks into account for a key block
many analysis methods of discontinuous mechanics analysis based on the Goodman-Shi method. Xu &
have been put forward and developed, such as limit Wang. (2000) studied the most dangerous sliding-
equilibrium method, key-block method, numerical block combination of rock by enlightening search
analysis method, and so on. of information with precedence of tree-like structure
The key block method has been widely used in depth. Yarahmadi & Verdel (2003) proposes ‘‘key-
many projects for conducting quick analysis of rock group method’’, that considers not only individual
mass media stability. The method is base on two but also groups of collapsible blocks into an iterative
hypothesizes: rock blocks are rigid and run through and progressive analysis of the stability of discontin-
the whole study rock mass project. Implement of uous rock slopes in 2D. To take intra-group forces
the key block method has been manifested into two into consideration, Yarahmadi & Verdel (2005) also
main forms: the vector technique by Warburton (1981) implemented the Sarma method within the key-group
and the graphical technique developed by Good- method to generate a Sarma-based key-group method
man & Shi (1985). Zhang & Kulatilake (2003) put for rock slope reliability anlyses in 2D. This paper
forward a new stereo-analytical method, which is a proposes ‘‘block-group method’’, which can analyze
combination of the stereo-graphic method and ana- key block-groups and non-movable block-groups. Fur-
lytical method, having applicability to both convex thermore, the kinetic energy law is implemented to
and concave blocks. There are three main steps in determine the movement behavior of blocks in key
the key block method. The first is to describe the block-groups.

1043
2 BLOCK-GROUP METHOD Table 1. The ∝ operations.

2.1 Grouping technique of block-group ∝ −1 0 1

−1 −1 −1 0
It is not impossible to consider all possible groups of 0 −1 0 1
two or more blocks to produce block-groups, yet a 1 0 1 1
number of combinations would quickly limit that con-
sists of defining the conditions required for assembling
block-group. Firstly, the key blocks must be the base
J5 EP SP
blocks for assembling block-group, because the stabil-
J3 J3 J5
ity of key blocks is the precondition of the stability of J1
total rock slope. Secondly, there are neighboring finite
blocks having common joint with key block or key J1
block-group. Thirdly, after the common joint is con-
JP1
nected as a whole, the stereo-analytical block method J2 JP2
J4
is adopted to analyze stability of block-group.
In order to operate and analyze easily, signed J4 J2
numbering method is adopted to show the topology
between block and joints. ‘‘+1’’ denotes that block
is in the upper half-space of the joint. ‘‘−1’’ denotes
Figure 2. One joint rock slope.
that block is in the lower half-space of the joint. ‘‘0’’
denotes that block does not include the joint.
Figure 1 shows one joint rock slope, which has three
‘‘∝’’ denotes the operations shown in table1. So, the
joints J1 , J2 and J3 . The vertical face is free face. Left
code of bigger key block-group can be expressed as:
half-space of the free face is excavation pyramid (EP).
Right half-space of the free face is space pyramid (SP).
D12 = D1 ∝ D2 = [1 − 1 0] (1)
According to the key block method, block 1 is a key
block and block 2 is a finite block. Their signed num-
bering codes are D1 = [1 0 −1] and D2 = [1 −1 1]
respectively. The finite block 2 is a neighboring block 2.2 Progressive grouping
of the key block 1. They have the common joint 3.
The goal of block-group method is to analyze larger
In other words, block 1 and block 2 are in opposing
hazardous unstable group of blocks instead of just a
half-space of joint 3. Correspondingly, the number pair
few key blocks. The block-group method consists first
‘‘+1’’ and ‘‘−1’’ is shown in the third position of their
of analyzing and confirming the key blocks. Then
signed numbering codes. When we hide joint 3, joint
identify if there are neighboring finite blocks for key
rock slope only has joint 1 and joint 2. According to
block or key block-group exist. Subsequently, stabi-
the key block method, the big block is a key block. It
lity analysis for block-group is performed according
means that we can assemble the block 1 and the block
to the stereo-analytical block method. For instance, the
2 to be a bigger key block-group 12. Here a sign ‘‘∝’’
process of progressive grouping is as follows:
is used to operate D1 and D2 to express D12 . The sign
Five joints J1 , J2 , J3 , J4 , J5 and a free face divided
rock slope into four joint blocks. Their codes are D1 =
[1 0 0 −1 0], D2 = [0 −1 1 0 0], D3 = [0 0 −1 1 −1]
and D4 = [1 −1 0 0 1] respectively (Figure 2). Based
on the key block method, block 1 and block 2 are key
blocks. Block 3 is a tapered block. Block 4 is a finite
block. So we can get block-groups from block 1 and
EP SP
block 2 respectively.
J1 For key block 1, joint 1 and joint 4 are composing
J1 joints. But the number pair of ‘‘+1’’ and ‘‘−1’’ does
JP2 not appear in the first position in the codes of block 1
and other blocks. In other words, there is no neighbor-
J2 JP1
J3 ing finite block having common joint 1 with block 1.
J3
So we do not need to hide joint 1 to assemble block-
J2 group. The number pair of ‘‘+1’’ and ‘‘−1’’ appears in
the forth position in the codes of block 1 and block 3.
So block 1 and block 3 can be assembled into a block-
Figure 1. One joint rock slope. group. Then the joint 4 can be hidden. According to

1044
the key block method, the block 13 is a non-movable
J3 EP SP
block-group (Figure 3). The code of block-group 13 J3
can be expressed as: J1

D13 = D1 ∝ D3 = [1 0 −1 0 −1] (2) J1

For block-group 13, joint 1, joint 3 and joint 5 are J2 JP2

composing joints. But the number pair of ‘‘+1’’ and
‘‘−1’’ does not appear in the first position of block J2
codes. So we do not need to hide joint 1 to assem-
ble block-group. The number pair of ‘‘+1’’ and ‘‘−1’’
appear in the third position in the codes of block- Figure 5. Non-movable block-group 134.
group13 and block 2. So the block-group 13 and
the block 2 can be assembled into block-group123. EP SP
Then the joint 3 can be hidden. According to the key J5
J1
block method, the block-group 123 is a key block-
group (Figure 4). Its code is D123 = D12 ∝ D3 = J1
[1 −1 0 0 −1].
Also the number pair of ‘‘+1’’ and ‘‘−1’’ appear in J2
JP1234
the fifth position in the codes of block-group13 and
J2
block 4. So the block-group 13 and the block 4 can be
assembled into block-group134. According to the key
block method, after the joint 5 is hidden, the block- Figure 6. Key block-group 1234.
group 134 is a non-movable block-group (Figure 5).
Its code is D134 = D13 ∝ D4 = [1 −1 −1 0 0]. block 2 is a neighboring block. The non-movable block
For the key block-group 123, finite block 4 is 134 and finite block 2 can also be assembled into
a neighboring block. The key block 123 and finite block-group 1234. According to the key block method,
block 4 can be assembled into block-group 1234 the block-group 1234 is a key block-group. Its code is
(Figure 6). For the non-movable block-group 134, key D1234 = D123 ∝ D4 = [1 −1 0 0 0].
The grouping steps for key block 2 is almost the
EP SP same as that for key block 1 and the results are as fol-
J5
J3 J3 J5
lows: non-movable block-group 23, key block-group
J1 123, non-movable block-group 234 and key block
group 1234.
J1
3 MOVEMENT BEHAVIOR OF BLOCKS
J2 JP2 FOR KEY BLOCK-GROUP
J2
3.1 The theory of movement behavior analysis
Due to the key block-group is considered as a whole,
the theory of key block method is difficult to ana-
Figure 3. Non-movable block-group 13. lyze the movement behavior of blocks in key block-
group. The blocks are assumed rigid blocks. So we
implement the kinetic energy law to achieve the aim.
J5 EP SP
Movement behavior of any block must be meet kinetic
J5
J1 energy law:

J1 Ek = W (3)

J2 Where Ek is the kinetic energy, W is the force work for

JP123
the block.
J2 Ek can be denoted as equation:

1 2 1 2 1 2
Ek = mu̇ + mv̇ + Jẇ (4)
Figure 4. Key block-group 123. 2 0 2 0 2 0

1045
W can be denoted as equation: Fy = Fy (t), 0≤t≤1 (13)

W = Wp + Wl + Wb (5) where t is a variant along the loading line. The work

of the line force is
where m represents mass of block. J represents rotate 1
inertia. The variable u0 and v0 are the rigid body trans- W1 = e0 [Fx (t)u + Fy (t)v]ldt (14)
lation of a specific point (x0 , y0 ). The variable w0 is
the rotation angle of the block with the rotation center When the line force (Fx (t), Fy (t))= (Fx , Fy ) is constant,
at (x0 , y0 ). Wp is the work of point force. Wl is the the work is derived as follows:
work of line force. Wb is the work of body force.
1 1

3.2 Displacement transformation formula of block

Wl = Fx e0 uldt + Fy e0 vldt (15)

For a single block, the displacement of arbitrary point In order to computer the integration, the elements of
of block can be ascertained based on one reference the integration have to be computed:
point of this block. Generally, the shape center of block
1 1
is used to be the reference point. The displacements e0 (x − x0 )dt = e0 ((x2 − x1 )t + (x1 − x0 ))dt
(u, v) of any point (x, y) of a block can be repre-
sented by three displacement variables (u0 , v0 , w0 ). 1
The displacements of any point can be written as = (x2 + x1 − 2x0 ) (16)
2
1
u = u0 − (y − y0 )w0 (6) e0 (y − y0 )dt = 21 (y2 + y1 − 2y0 ) (17)

v = v0 + (x − x0 )w0 (7) Therefore

 1
3.3 Point force work W1 = l u0 Fx + v0 Fy − w0 (y2 + y1 − 2y0 )Fx
2
When a block moves in a key bock-group, it certainly 1 
will contact other blocks. The point contact may occur. + w0 (x2 + x1 − 2x0 )Fy (18)
2
The point loading force affects the movement of the
blocks. The point force (Px , Py ) acts on point (x, y) of
a block. The work of point force is: 3.5 Body force work
Assuming that (fx , fy ) is the constant body force acting
Wp = Px u + Py v on the body of the block, (xG , yG ) is the center of
= Px [u0 − (y − yo )w0 ] + Py [v0 + (x − x0 )w0 ] gravity of this block. Then

(8) Sx
xG = (19)
S
3.4 Line force work Sy
yG = (20)
S
The surface contacts between blocks engender line
force, such as friction. Assume the force is distributed where
on a straight line from point (x1 , y1 ) to point (x2 , y2 ).
The equation of the force line is
S= eedxdy (21)
x = (x2 − x1 )t + x1 (9)
Sx = eexdxdy (22)
y = (y2 − y1 )t + y1 , 0≤t≤1 (10)
Sy = eeydxdy (23)
The length of this line segment is
 The work of the constant body force (fx , fy ) is
l= (x2 − x1 )2 + (y2 − y1 )2 (11)

The force is
Wb = ee(f x u + fy v)dxdy (24)

where (u, v) is the displacement of gravity center of

Fx = Fx (t) (12) body. When the block is homogeneous, the gravity

1046
center is the same as shape center. So the work of quickly determine the movement behavior and failure
body force is mode of blocks in key block-group. The main advan-
tage of using the block-group method over the other
Wb = S(f x u0 + fy v0 ) (25) discontinuous analysis methods lies in its processing
speed.
3.6 Dynamic analysis
Based on the formula (4), (8), (18), (25), the formula 4 BLOCK-GROUP METHOD STEP
(3) can be rewritten as:
Based on grouping technique of block-group and
1 2 1 2 1 2 movement behavior analysis of blocks in a key block-
mu̇ + mv̇ + Jẇ group, the block-group method steps are illustrated in
2 0 2 0 2 0
Figure 7.
= (Px + lFx + Sf x )u0 + (Py + lFy + Sf y )v0
1
+ [Px (y0 − y) + Py (x − x0 ) − lFx 5 VALIDITY OF BLOCK-GROUP METHOD
2
1 Shown in Figure 8 is an example of a discontinuity rock
(y2 + y1 − 2y0 ) + lFy (x2 + x1 − 2x0 )]w0
2 slope. It consists of four joints J1 , J2 , J3 , J4 and one
(26) free face. The mechanical properties are as follows:

It can be written as ρ = 2500kg/m3 , c = 0.03MPa, ϕ = 27◦

1 2 where ρ is the rock density, c is the cohesion of the

mu̇ − (Px + lFx + Sf x )u0 = 0
2 0 discontinuities, and ϕ is the friction angle of the dis-
1 2 continuities. According to the key block method, rock
mv̇ − (Py + lFy + Sf y )v0 = 0 slope includes key block 1, tapered block 2, non-
2 0
movable block 3, finite block 4, finite block 5 and five
1 2 infinite blocks. However, Based on the block-group
Jẇ − [Px (y0 − y) + Py (x − x0 )
2 0 method, the biggest key block-group is key block-
1 group 123 (Figure 9). The comparison between the
− lFx (y2 + y1 − 2y0 )
2
1 Assembly of blocks {B}
+ lFy (x2 + x1 − 2x0 )]w0 = 0 (27)
2

The result of displacements (u0 , v0 , w0 ) of a block Find key blocks {K}

can be got through analysis. So the movement behavior
of blocks in a key block-group is also determined. no
{K}
For a block in key block-group, there are several
forces caused by counteraction with neighbor blocks. Find moveable blocks yes
Combine force F and combine moment M are engen- Grouping block-groups
dered for shape center of block by these force loading. Update assembly
If F and M are not equal to zero, the unequal F and of blocks
M make block move meeting kinetic energy theory. Find the bigger key block-groups
Because large displacement is involved in stability
analysis, the block position and contacts will change
following time steps. We have to know all the pairs Remove key block-groups
of blocks which will possibly meet together during the
next step. And the counteract force must be got accord-
ing to a definite method. The method is defined in the
conventional ways like Distinct Element Method. For Find the biggest key block-groups
the key block-group, calculation is done for all blocks
based on time step iteration. Comparing with dis-
tinct element method and other discontinuity analysis Movement behavior analysis of blocks
methods, the block-group method not only considers a
single block but also group of blocks. The method can Figure 7. Flow chart of block-group method.

1047
 J3

J2 J1 J4

Figure 8. Discontinuity rock slope. Figure 10. Failure mode of rock slope provided by UDEC.

Table 2. Material properties used in the numerical analysis

with UDEC.

Property Value Units

Blocks
Density 2500 Kg/m3
Bulk modulus 12 GPa
Shear modulus 8 GPa
Discontinuities
J3 Joint normal stiffness 25.3 GPA/m
Joint shear stiffness 8.5 GPa/m
Cohesion 0.03 MPa
Tensile strength 0.01 MPa
Friction angle 27 Degree

J2 J1 J4
rock slope. In addition, kinetic energy law is developed
to determine the movement behavior of blocks in the
Figure 9. Key block-group 123.
key block-group. Compared with other discontinuity
numerical analysis methods, the main advantage of
using the block-group method lies in its processing
block-group method and the distinct element analy- speed and can quickly confirm the most hazardous
sis performed using UDEC is conducted (Figure 10). area and failure mode of joint rock slope.
Table 2 displays the properties used in the numeri-
cal analysis with UDEC. We find they make a good
agreement. So we can consider block-group 123 as the
REFERENCES
most hazardous area of rock slope. This proved that the
block-group method yields more realistic results than Cundall, P.A. 1971. A computer model for simulating pro-
the basic key block method. gressive large scale movements in blocky systems. Pro-
ceeding of the symposium of the international society of
rock mechanics. Nancy, France 1:2–8.
6 CONCLUSIONS Goodman, R.E. & Shi, G. 1985. Block theory and its
Application to Rock Engineering. Prentice-Hall: New
Jersey.
Owing to taking a greater number of blocks into Shi, G.H. 1988. Discontinuous deformation analysis new
account in the stability analysis and more realistic, the numerical model for the static and dynamics of block
block-group method proposed in this paper is supe- system. Berkeley: Department of civil engineering, Uni-
rior to the key block method for stability analysis of versity of California.

1048
Warburton, P.M. 1981. Vector stability analysis of an arbi- Yarahmadi Bafghi, A.R. & Verdel, T. 2005. Sarma-based key-
trary polyhedral rock block with any number of free faces. group method for rock slope reliability analyses. Interna-
International Journal of Rock Mechanics and Mining tional Journal for Numerical and Analytical Methods in
Science & Geomechanics 18:415–427. Geomechanics 29:1019–1043.
Wibowo, Johannes L. 1997. Consideration of secondary Zhang, Zixin & Kulatilake, P.H.S.W. 2003. A new stereo-
block in key-group analysis. International Journal of Rock analytical method for determination of removal blocks
Mechanics and Mining Science & Geomechanics 34:3–4. in discontinuous rock masses. International Journal for
Xu, Mingyi & Wang, Weiming. 2000. Research on the dan- Numerical and Analytical Methods in Geomechanics 27:
gerous sliding-block combination of rock slopes. Rock 791–811.
and soil mechanics 21(2):148–151.
Yarahmadi Bafghi, A.R. & Verdel, T. 2003. The key-
group method. International Journal for Numerical and
Analytical Methods in Geomechanics 27:495–511.

1049
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Quantitative study on the classification of unloading zones of high slopes

Da Zheng & Run-Qiu Huang

State Key Laboratory of Geohazards Prevention, Chengdu University of Technology, Chengdu, China

ABSTRACT: Unload and unloading zones of high rock slope are familiar phenomenon in southwest area of
China, which influence stability of rock slope and other rock engineering. On the basis of analyzing the actual way
of classification of unloading zones, the paper puts forward a divisiory way of adopting crack rate, opening crack
rate and summation of crack width as quantitative indexes according to formation mechanism and geological
exhibition of unloading zones. After large numbers of locale measurement of cracks, analytical result indicates
that there is a good corresponding connection between the three indexes and unloading degree, and it is feasible to
regard the three indexes as quantitative standard of classification of unloading zone. Furthermore, combining to
qualitative geological characteristics of unloading zones, the paper presents suggestions for dividing unloading
zones of actual projects.

1 GENERAL INSTRUCTIONS have been accepted by practical engineering, and the

criterion for estimating unloading zones is established
Unloading of the rock mass slopes is that due to river by the corresponding exploration survey code (Lu,
erosion or manpower excavation free faces’ occur- 1995). However, due to obvious difference of the
rence destroys initial stress balance of the rock mass, geological conditions and complex unloading there
as leads to stress releasing at the superficial layer of is uncertain to divide unloading zones in practical
the slopes and unloading rebound. During this process operations. Therefore, the geological index and meth-
a new rupture system is formed. These constitute rock ods to divide unloading zones still wait for more
mass relaxation and tensile fracture phenomena at the discussions.
shallow slopes (Huang et al, 1994). In the paper several kinds of unloading divisions
The slope unloading is a common dynamical phe- are studied. And some basic rules have been found
nomenon for the high rock slopes in the southwest according to many field investigations. Hereby cor-
area of China. Unloading breaks the whole rock mass, responding suggests on geological basis and quan-
and reduce the qualities of the rock mass. At the titative index to divide unloading zones are brought
same time, the channels are formed due to weathering forward.
force and exogenic process of underground. Therefore,
unloading is meaningful for analyzing the stability of
the rock slopes and related rock engineering. In the 2 IDENTIFYING AND DIVIDING METHODS
region of engineering geology and rock mechanics FOR ACTUAL SLOPE UNLOADING ZONES
the researches on unloading and unloading zones are
always concerned. Evaluation and division on the slope unloading zones
Unloading is usually divided into strong, weak are firstly based on the unloading phenomenon, crack
and centralized ones. Strong unloading and weak opening, filling condition, weathering of the crack
unloading are familiar. In recent years the central- wall and underground in the light of the criterion
ized unloading is a special phenomenon for the high offered by the codes. Then farther check for the locale
rock slopes in the gorges of southwestern China. The evaluation of the unloading zones is made in terms of
paper mainly discusses strong unloading and weak the wave velocity of the adit walls.
unloading. At present, there is no uniform standard for dividing
Forming mechanism of the unloading zone, and the unloading zones. Hou (2000) researched changing
classification of the unloading zone and engineer- characteristics of unloading crack growth density and
ing unloading characteristics of the rock mass have opening degree with adit depth. And in strong unload-
been researched by many scholars (Huang, 2000; ing rock mass cracks are at large open, and full with
Zhang, 1993; Li, 2001; Wu, 2001). Their findings secondary mud, width of which ranges from 0.5 cm

1051
to 1.5 cm. Crack width of the weak unloading rock 4 APPLICATION AND DISCUSSION
mass is commonly less than 1 cm, and much full with
calcareous and muddy matter (Hou, 2000). Besides the Using these quantitative indexes, we divide the unload-
above indexes, due to unloading the rock structure is ing zones for certain Hydropower Station.
relaxed, and the rock mass loses the integrality, and the
structure plane becomes loose, as results in enhancing
water transmitting ability of the rock mass. There- 4.1 Geological general situations
fore, coefficient of permeability obtained by forced The dam of the Hydropower Station is located in the
water tests reflects the conditions of rock mass’s crack lower braches of Jinshan River. The river valley is nar-
growth, joint opening and unloading. It is used to row and the bank slope is steep, with anisomerous
divide the unloading zones of the rock mass. The Code V shape. It is composed of the layer from upper Emeis-
For Water Resources and Hydropower Engineering han basalt of Permian System. The consequent slopes
Geological Investigation in China (GB5087–99) has can be seen in the left bank, and its landform shows a
regulated classification of rock mass permeability in ladder shape. In the right bank the slope is steep. There
detail, as indicates that the penetrability of rock mass are preferable engineering geological conditions in the
is related with opening degree of the cracks. Quantita- dam area. The rock structure in the region of the dam
tive index of the rock mass unloading zones has been mainly develops in the interlaminar, disturbed belt,
applied into practical projects. For example, longitu- small fault and rockbed crack in the basalt. The dis-
dinal wave velocity and crack opening degree are used turbed belt with good connectivity goes through the
for dividing the unloading zone of the dam abutment, whole dam region, which composes the basic struc-
Xiaowan Hydropower Station. In addition, opening tural frame of the rock structure in the area of the dam.
degree of centralized unloading crack, its growth den- The intrastratal disturbed belt with gently dip angles
sity, rock mass structure and sound wave velocity of the popularly grows in the basalt, which is an important
rock mass are considered as the standard for dividing part of the rock structure in the region of the dam.
the unloading zones of Xiluodu Hydropower Station
in Jinsha River.
4.2 Division of the slope unloading zones
Both quantitative and qualitative methods are com-
bined for dividing the unloading zones in the area of
3 SELECTION OF QUANTITATIVE DIVISION the dam. The quantitative indexes used in the paper
OF SLOPE UNLOADING ZONES are crack ratio, opening crack ratio and sum of crack
width. These indexes are measured by setting the
Researches show that slope rock mass unloading measured lines along the adit wall.
immediately gives rise to the results that the rock mass In order to achieve full statistical data, the paper
in shallow layers is relaxed, original structure planes amply measured more than 40-cm-long cracks over
open, secondary cracks are generated, and even rock ten adits, and obtained many data about the cracks. At
mass failures (Ju, 2000). The data statistic indicates the same time, the data are analyzed to divide unload-
that number of rock slope cracks always decreases with ing zones. For example, Adit 39 is used as a case. Fig.
increase of the horizontal depth far from the slope sur- 1 describes changing rule of crack ratio and opening
face. And the number and opening degree are stable crack ratio with the adit depth. Changing rule of crack
when the horizontal length is more than a fixed value.
So the paper thinks that identifying and dividing the
unloading zones mainly depends on number, open-
ing width of the cracks, and relaxation of the rock
mass structure. Therefore, we select change of crack
ratio, opening crack ratio, crack opening degree or
secondary mud width full with the cracks to describe
unloading conditions of the rock mass, and divide
unloading zones.
Crack ratio is the crack number per unit length
along the measured adit wall. Opening crack ratio is
the number of opening cracks per unit length mea-
sured. Degree of crack opening is described using
‘‘crack width sum’’. The three indexes are all quan-
titative ones. In practical engineering, the indexes can
be obtained through careful measurements, thus are Figure 1. Curve of crack rate and opening crack rate with
exercisable. depth in the adit 39.

1052
 Crack ratio and open crack ratio
Figure 3. Curve of crack rate and opening crack rate with
Figure 2. Curve of summation of crack width with depth in
depth in the adit 57.
the adit 39.

width sum with the adit depth can be seen in Fig. 2.

Crack ratio and opening crack ratio decrease with
increase of the adit depth. At the position from 52 m to
54 m, the two parameters rapidly fall at the same time.
The results indicate that the crack is close, and the rock
mass no longer is affected by unloading when the adit
depth is more than 54 m. The three indexes above are
accordant for describing slope unloading. In the light
of the results 54 m is considered as unloading depth of
the adit.
Figure 4. Curve of summation of crack width with depth in
It should be pointed out that the marked relationship
the adit 57.
between crack ratio and unloading is related with the
properties of the basalt. The basalt under no unloading
conditions contains many hidden cracks. As the basalt
is affected by unloading the hidden cracks emerge.
According to quantitative parameters we again dis-
cuss division of strong unloading and weak one. Fig. 1
and Fig. 2 show that only one of the three indexes well
changes at the position from 20 m to 22 m deep in the
adit. For 0 m to 20 m, the crack ratio is 25 lines per
2 m, opening crack ratio is 15 lines per 2 m, and open-
ing crack lines account for 60 percent. The crack width
sum is up to 26.8 mm per 2 m. The results indicates that
from 0 m to 21 m the rock mass is under strong unload-
ing zones. From 21 m to 54 m deep in the adit, crack
ratio and opening crack ratio evidently fall, opening
crack lines is less than 10 percent of the total number,
and crack width sum is averagely 0.84 mm per 2 m. Figure 5. Classification result of unloading zones at I-I
Therefore, the segment from 21 m to 54 m is con- section of some hydropower’s dam site.
sidered as weak unloading zone. As the adit depth is
more than 54 m, crack width sum is zero. This shows The results of unloading division using field empir-
that unloading no longer affect the segment more than ical method and the above quantitative method are
54 m. basically accordant. Therefore, crack ratio, opening
Other adits’ unloading is divided according to the crack ratio and crack width sum can be used as geolo-
same method. Fig. 3 and Fig. 4 describe that crack gical quantitative indexes for dividing unloading zones
ratio and crack width sum for Adit 57 change with of the high slopes’ rock mass.
adit depth. The conclusion is consistent with Adit 39. Adopting the above methods we divide unloading
The strong unloading ranges from 0 m to 56 m, and zones of all adits in the Profile I-I, distribution of which
weak one changes from 56 m to 102 m. can be seen in Fig. 5.

1053
Table 1. Index of quantification on classification of unloading zone in dam site of some hydropower.

Unloading zone Strong unloading zone Weak unloading zone

Crack ratio (/2 m) 20∼25 10∼15

Opening crack ratio (/2 m) 10∼15 5∼8
Crack width sum (mm/2 m) 15∼20 1∼5
Unloading characteristics Rock mass is loose, cracks and secondary mud Unloading cracks well develop with
appear, structural cracks are widely open, and small width, and no mud fills
full width widely open, and full with debris and the cracks
secondary mud. Steep dip angles are found
Rock character Basalt

On the basis of the findings, suggest schemes for Huang Runqiu, Lin Feng, Chen Deji, et al. Formation mech-
dividing unloading zones of the damsite are list in anism of unloading fracture zone of high slopes and its
Table 1 in terms of geological qualitative analysis engineering behaviors. Journal of Engineering Geology,
and analogue of Xiaowan, Xiluodu and Xiangjiaba 2001, 9 (3): 227–232 (in Chinese).
Hydropower Stations’ experiences. Ren Guangming, et al. A Quantitiative Study on the Classi-
fication of Unloading Zones of Rock mass Slope. Journal
of ChengDu University of Technology, 2003, 8: 235–238.
(in Chinese).
5 CONCLUSIONS Nie Dexin, Han Aiguo & Ju Guanghong. Study on Integrated
Zoning of Weathering Degree of Rock mass. Journal of
1. Unloading of the rock mass currently represents Engineering Geology, 2002, 10 (1): 20– 25. (in Chinese).
opening of the cracks and relaxation of the rock WU Gang. The present and expectation of the study to mecha-
mass structure. The researches in the paper indicate nism of engineering rock broke unloading. The Journal of
that increase of crack number is an important char- Engineering Geology, 2001, 9 (2): 174–181. (in Chinese).
acteristic of unloading. It, altogether with crack The Ministry of Water Resources of People’s Repub-
lic of China. The Code For Water Resources and
opening, shows like changing rules with increase Hydropower Engineering Geological Investigation in
of adit depth. China (GB5087–99). Beijing: China Plan Press, 1999.
2. On the basis of research findings about rock mass Hou Zhibin. Unloading characteristics of sand-shale
unloading, changes of crack ratio, opening crack rock mass in the pivot area of Chezhuang Hydraulic
ratio, and crack opening degree (crack width sum) Engineering. Shanxi Hydropower, 2000, 16 (3): 4–9. (in
are selected to describe unloading degree of rock Chinese).
masses. The changes of the three indexes are highly Lu Hong. Engineering geological survey of the high arch
consistent with adit depth. dam. Yunnan hydroelectric technique, 1995 (4): 4–15. (in
3. The three parameters are used to divide unload- Chinese).
Huang Runqiu, Wang Shitian, Hu Xiewen et al. Study on
ing zones in the practical engineering regions. The the main Engineering Geological Problems of Xiaowan
results are basically consistent with those divided high arch dam. Chengdu: Southwest Jiao-Tong University
by empirical geology engineers. Therefore, crack Press, 1996. (in Chinese).
ratio, opening crack ratio and crack width sum can National Laboratory of Geohazards Prevention of Chengdu
be considered as geological quantitative index and University of Technology. Study on the rock mass qualities
basis for dividing unloading zones of high slope and selection of the base surface in the dam site of Xiluodu
rock mass. Hydropower Station. Chengdu: Chengdu University of
Technology, 2001. (in Chinese).
Ju Guanghong. Study on engineering geology of weath-
ered unloading zones for deep down-cutting gorge granite
REFERENCES in Laxiwa Hydropower Station, Yellow River. Chengdu:
Master thesis of Chengdu University of Technology. 2002.
Huang R.Q., Zhang Z.Y. & Wang Shitian. 1994, Research on (in Chinese).
rock structure epigenetic reformation. Hydrogeology and
Engineering Geology, 1994, 21 (4): 17–21. (in Chinese).

1054
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Investigations on the accuracy of the simplified Bishop method

D.Y. Zhu
School of Civil and Water Resources Engineering, Hefei University of Technology, Hefei, China

ABSTRACT: This paper describes the reason why the simplified Bishop method always gives factors of safety
of circular slip surfaces in good agreement with those given by the rigorous methods. The absence of vertical
interslice force in the factor of safety equation in this method only means a summation regarding the vertical
interslice forces has been neglected. An appropriate set of vertical interslice forces could be found that: (1)
allows this term to be zero, (2) renders the satisfaction of the horizontal equilibrium condition, and (3) gives
the same factor of safety. In other words, the simplified Bishop method implicitly satisfies the rigorous limit
equilibrium conditions.
Key words: simplified Bishop method, limit equilibrium, slope stability, factor of safety

1 INTRODUCTION and Ei+1 , the vertical interslice forces are Xi and Xi+1 ,
the normal force at the base is Ei , the shear resistance
The simplified Bishop method (Bishop, 1955) has at the base is Ti , the pore water pressure at the base
been widely used in slope stability analysis and is is ui , the effective internal friction angle and cohesion
regarded as the best method of limit equilibrium are φi and ci , respectively. The factor of safety along
for calculating the factors of safety of circular slip the slip surface is Fs .
surfaces. In this method, the interslice forces are According to Mohr-Column failure criterion and
assumed to be horizontal, or the vertical interslice the principle of effective stress:
forces are neglected, the vertical force equilibrium and
the moment equilibrium about the centre of the circu-
lar slip surfaces are satisfied, but the horizontal force
equilibrium is not considered. Thus, the simplified
Bishop method is still regarded as one of the non-
R
rigorous limit equilibrium methods of slices. However,
the simplified Bishop method always gives factors xb
of safety of circular slip surfaces in good agreement y bi
(n)

with those given by other rigorous methods of slices (i) X i+1

such as the Morgenstern-Price method (Morgenstern Ei+1
o x Ei
and Price, 1965) and the Spencer method (Spencer, xa Wi
(1)
1967). In the profession of geotechnical engineering, (2) Xi
Ti
the simplified Bishop method has been accepted as the
accurate method of slices, although it dos not satisfy i Ni
all the limit equilibrium conditions (Duncan, 1996).
The problem why the simplified Bishop method is so
accurate has puzzled the profession for over 50 years. X
This study attempts to give a theoretical study on this
issue.
xa xm xb

X/b
2 FUNDAMENTALS

A sliding body about a circular slip surface of radius

R is divided into n vertical slices, as shown in Fig. 1.
For the ith slice, the width is bi , the angle of base is αi ,
the weight is Wi , the horizontal interslice forces are Ei Figure 1. Sliding body and forces acting on the slice.

1055
Ti = (Ni − ui bi sec αi ) tan φi /Fs + ci bi sec αi /Fs (1) still identical to that of the simplified Bishop method.
Furthermore, if that set of vertical interslice forces are
Consider the vertical force equilibrium condition so selected that the horizontal force equilibrium con-
dition is satisfied, then the factor of safety computed
Ni cos αi + Ti sin αi = Wi + Xi+1 − Xi (2) by equation (6) or (7) corresponds to that of rigorous
method of slices since all the three equilibrium condi-
From equations (1) and (2), one obtains tions are completely satisfied. The key issue is to find
such a set of vertical interslice forces that not only
Ni = (Wi + Xi+1 − Xi + ui bi tan αi tan φi /Fs satisfy the horizontal force equilibrium condition but
also equation (8).
− ci bi tan αi /Fs )/mαi (3) Consider the horizontal force equilibrium for the
 whole sliding body:
Ti = (Wi + Xi+1 − Xi − ui bi ) tan φi /Fs
 
n
+ ci bi /Fs /mαi (4) (Ti cos αi − Ni sin αi ) = 0
sin αi tan φi /Fs
i=1
mαi = cos αi + (5)
and substitute equations (3) and (4) into the above
Consider the moment equilibrium with respect to the
equation, one obtains:
centre of circular slip surface:
n 


n 
n
Ti R = Wi R sin αi (Wi + Xi+1 − Xi )(cos αi tan φi /Fs − sin αi )
i=1
i=1 i=1
:
− (ui bi tan φi − ci bi )/(Fs cos αi ) mαi = 0
Substituting equations (3) and (4) into the above
equation gives (9)

n   Now what remains is to find the distribution of ver-

i=1 (Wi + Xi+1 − Xi − ui bi ) tan φi + ci bi /mαi
Fs = n tical interslice forces X that simultaneously satisfy
i=1 Wi sin αi equations (8) and (9).
Similar to the Correia method (Correia, 1988), the
(6) distribution of vertical interslice forces is of the form
The simplified Bishop method assumes that the con- as follows
tribution of vertical interslice forces to the factor of
safety is neglected, hence X = λf (x) (10)
n
[(Wi − ui bi ) tan φi + ci bi ]/mαi where f (x) is interslice force function and λ is the
Fs = i=1 n (7)
i=1 Wi sin αi
scaling factor. The difference between the vertical
interslice forces on the two sides of the slice is

3 SATISFACTION OF HORIZONTAL FORCE Xi+1 − Xi = Xi = λfi bi (11)

EQUILIBRIUM
Substitute of equation (11) into (8) leads to
The vertical interslice forces do not appear in the factor
of safety equation of the simplified Bishop method. 
n
This does not mean the vertical interslice forces should fi bi · tan φi /mαi = 0 (12)
be zero. In fact, from equation (6), we can see if the i=1
following equation holds
The value of λ is determined according to the hori-

n
zontal force equilibrium. Substitute of equation (11)
(Xi+1 − Xi ) · tan φi /mαi =0 (8) into (9) gives
i=1

equation (6) is identical to equation (7). n

Ai/mαi
In other words, if a set of vertical interslice force λ = ni=1  (13)
exist that satisfy equation (8), the factor of safety is i=1 i bi /mα
f

1056
in which 4 EXAMPLE

Ai = −Wi (cos αi tan φi /Fs − sin αi ) A homogeneous slope has a circular slip surface, as
shown in Fig. 2. The unit weight is 18.4 kN/m3 , the
+ (ui bi tan φi − ci bi )/(Fs cos αi ) (14) cohesion and the internal friction angle are 25 kPa and
22◦ respectively.
For a circular slip surface, the interslice force func- The factor of safety is 2.08 computed by the sim-
tion f (x) satisfying equation (12) can be found, giving plified Bishop method. When the interslice force
the factor of safety that is identical to that of the simpli- function f (x) is chosen as equation (15) with xm =
fied Bishop method. The value of λ is then determined 36.9, equation (12) is satisfied. From equation (13),
from equation (13), resulting in the distribution of the the scaling factor λ is calculated as 1988.62. The dis-
vertical interslice force satisfying the horizontal force tribution of the vertical and horizontal interslice forces
equilibrium condition. is shown in Fig. 2.
For simplicity, the interslice force function f (x) can
be chosen as (see Fig. 1)
⎧ x−x 5 CONCLUSION
⎪
⎪
a
xa ≤ x ≤ xm
⎨ xm − xa
f (x) = (15) The simplified Bishop method, which neglects the
⎪
⎪ xb − x vertical interslice forces and takes no account of the
⎩ xm < x ≤ xb
xb − xm horizontal force equilibrium condition, has accuracy
as high as the rigorous limit equilibrium methods
Thus of slices. This has been confirmed by experiences
⎧ of many practitioners, but not been explained theo-
⎪ 1 retically. This study shows that the factor of safety
⎪
⎨ xa ≤ x ≤ xm
x − xa equation of the simplified Bishop method does not
f  (x) =
m
(16) require that the vertical interslice forces be zero, but a
⎪
⎪ 1
⎩− xm < x ≤ xb term designated by Eq. (8), be zero. This condition can
xb − xm be realized by assigning an appropriate distribution of
Since the simplified Bishop method implicitly sat- vertical interslice forces that satisfies the horizontal
isfies all the complete equilibrium conditions through force equilibrium condition simultaneously, and lead
proper selection of vertical interslice force distribu- to a factor of safety identical to that of the simplified
tion, it is not strange that the simplified Bishop method Bishop method. Since the approach satisfies the com-
has the accuracy that compares with that of other plete equilibrium conditions, this method has accuracy
rigorous methods of slices. so high that can compares with the Morgenstern-Price
method and the Spencer method.
(18, 66)

REFERENCES
=18.4kN/m3
c =25kPa R =85m
30.0m Bishop, A.W. The use of the slip circle in the stability analysis
=22˚ of earth slopes. Géotechnique 1955, 5(1): 7–17.
2 Correia, R.M. A limit equilibrium method of slope stabil-
1 ity analysis. Proc. 5th Int. Symp. Landslides, Lausanne,
0.0m 1988, 595–598.
Duncan, J.M. State of the art: limit equilibrium and finite-
element analysis of slopes. J. Geotech. Engrg., ASCE,
1996, 122(7): 577–596.
Morgenstern, N.R., and Price, V.E. The analysis of the stabi-
kN/m lity of general slip surfaces. Géotechnique, 1965, 15(1):
10000 79–93.
8000 Spencer, E. A method of analysis of the stability of embank-
E
ments assuming parallel interslice forces. Géotechnique,
6000
1967, 17(1): 11–26.
4000 Zhu, D.Y., Lee, C.F. and Jiang, H.D. Generalized framework
2000 X of limit equilibrium methods for slope stability analysis.
0 Géotechnique, 2003, 53(4): 377–395.
-20 0 20 40 60 80 m

Figure 2. Slope and interslice forces distribution.

1057
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Author index

Abdullah, C.H. 1873 Burlon, S. 687 Chi, S.Y. 1639

Abe, S. 939 Burns, S.F. 1979 Ching, J. 639
Abolmasov, B. 1939 Cho, Y.C. 933
Aceto, L. 373 Cai, D. 1817 Choi, Y.S. 1633
Adler, R.F. 1991 Cai, F. 593, 851, 1455 Chou, Y.H. 1493
Aghababaei, H. 857 Cai, J.T. 1659 Chu, B.L. 1409
Akhundjanov, A.M. 1247 Cai, Y. 1165 Clément, C. 331, 1143
Akin, M. 1647 Cai, Y.J. 2019 Clarke, D. 1553
Al. Zaitsev, A. 963 Calabrò, N. 647 Coe, J.A. 1899
Allison, R.J. 865 Calcaterra, D. 299, 1095, Coelho Netto, A.L. 409
Almeida Saraiva, A.L. 1397 Cola, S. 647
561 Calò, F. 299, 1095 Colombo, A. 1233
Alonso, E.E. 67 Calvello, M. 495, 599, 607 Conversini, P. 1609
Aminpoor, M. 553 Cao, G.J. 2027 Corominas, J. 237, 1517
Anarbaev, E.M. 1475 Cao, L. 449 Cotecchia, F. 1525
Araiba, K. 1321 Cao, P. 815 Cottaz, Y. 319
Asano, S. 1375, 1439 Cappa, F. 331, 711 Couture, R. 1151
Auray, G. 1721 Casarano, D. 513 Critelli, S. 373
Cascini, L. 599, 607, 1103, Crovelli, R.A. 1899
Bach, D. 1991 1893 Cucchi, A. 1233
Badv, K. 1489 Castellanos Abella, E. 1879 Cui, P. 263
Bai, A. 737 Catalano, E. 1433 Cui, X. 1481
Bai, J. 1071, 1077 Ceccucci, M. 1111
Bai, S. 1067 Cecinato, F. 615 Dahlhaus, P.G. 267
Barradas, J. 567 Cha, K.S. 1763 Dai, W.J. 1367
Baum, R.L. 179 Chae, B.G. 933 Damiano, E. 1157
Bazarov, Sh.B. 1247 Chai, H.J. 1675, 1799 Dang, L.C. 1195
Becker, R. 1083 Chamra, S. 1653 Daryono, M.R. 1615
Bell, R. 1083 Chang, C.H. 1559 De Roeck, G. 573
Bencardino, M. 365 Chang, M. 1493 Delacourt, C. 395
Beniston, M. 217 Che, A. 307, 449 Demir, B.M. 531
Bhandari, R.C. 231 Chen, C.Y. 313 Deng, J.H. 821, 999
Bianchi, M. 1233 Chen, G.F. 2101 Deparis, J. 323
Bica, A.V.D. 293 Chen, H. 2005 Derevenets, Ph.N. 1225
Bièvre, G. 395 Chen, J. 535 Dhungana, I. 1463
Bigarré, P. 331, 1925 Chen, K. 1501 Di Crescenzo, G. 1951
Blikra, L.H. 1089 Chen, L.H. 631 Di Maio, C. 495
Bojorque, J. 573 Chen, S. 789 Di Martire, D. 299
Bomont, S. 1713 Chen, S.F. 1559, 2019 Di Nocera, S. 1103
Bonnard, Ch. 217 Chen, X.D. 1507 Diano, G. 585
Bottiglieri, O. 1525 Chen, Y.H. 1639 Díez, J.A. 347
Bovenga, F. 513 Chen, Y.M. 731 Dijkstra, T.A. 1553
Bozzano, F. 579, 585, 1381, Chen, Z. 25, 2011 Dima, C. 1397
1389, 1905 Chen, Z.Y. 623, 631, 655 Ding, J. 1533
Bressani, L.A. 293 Cheng, Q. 1133, 1985 Ding, X.L. 1367
Bretschneider, A. 579 Cheng, Q.G. 1119 Dixon, N. 1553
Burghaus, S. 1083 Cheng, Y. 1127 Doanth, B.T. 1475

1059
Dong, S.M. 1675 Gens, A. 67 Hu, B. 979
Du, L. 1775 Ghanbari, E. 1405 Hu, G.T. 1119
Du, Y. 1781 Gharouni-Nik, M. 1725 Hu, X.W. 249
Duan, Q.W. 655, 663 Ghazanfari, M. 1911 Hu, Y. 1481
Dünner, C. 331, 1143, 1925 Ghazipoor, N. 491 Hu, Y.G. 639
Durmaz, S. 531 Giannico, C. 1233 Huang, G.W. 1367
Dyer, M. 1309, 1965 Gillarduzzi, A. 1547 Huang, R. 389, 403, 1355
Dykes, A.P. 339 Glade, T. 1083 Huang, R.Q. 1051, 1177,
Glendinning, S. 1553 2037
Eberhardt, E. 39 Gon, L.S. 1571 Huat, B.B.K. 1919
Effendiantz, L. 323 Gong, X.N. 1659 Huffman, G.J. 1991
Eisenberger, C.N. 293 González, D.A. 237 Hughes, D.A.B. 1553
El Bedoui, S. 421, 1273 González-Gallego, J. 1731 Hughes, P.N. 1553
Ellis, W.L. 179 Gori, P.L. 1173 Hui, T.H.H. 1769
Emami, K. 1489 Grasso, E. 1397 Hungr, O. 145
Enoki, M. 671 Greco, R. 1157 Huo, Z.T. 1321
Entezam, I. 491 Grelle, G. 365
Eom, Y.H. 1297 Gress, J.C. 361, 687 Il, M.Y. 1571
Esaki, T. 893 Grimaldi, G.M. 607 Ishii, Y. 697
Esposito, C. 355, 585, 1905 Guadagno, F.M. 365 Ivanchenko, I.V. 963
Estaire, J. 347 Guglielmi, Y. 331, 711,
Evangelista, A. 913 1273 Jafari, M.K. 1427
Gui, M.W. 1559, 1737 Jamaludin, S. 1919
Fan, J. 543 Gullà, G. 373, 599 Janik, M. 1083
Fan, X.-M. 1681 Gunzburger, Y. 1143 Jeon, S. 463
Fasani, G.B. 355, 1905 Guo, B. 679 Jia, G.W. 731
Faure, R.M. 319, 687, 1721 Guo, H.X. 1507 Jiang, J.-C. 747
Fei, L.Y. 1639 Jiang, L. 1481
Feng, H. 1863 Haelterman, K. 761 Jiang, L.-W. 1681
Feng, M. 2061 Halliday, G.S. 1565 Jiang, R. 737
Feng, S. 307 Hamidi, A. 469 Jiang, X. 741
Feng, X. 803 Han, J. 1071, 1077 Jiang, X.L. 815
Feng, X.Y. 693 Han, K.K. 1737 Jiang, Y.H. 1335
Ferlisi, S. 1103, 1893 Han, Y. 463 Jibson, R.W. 1447
Flentje, P. 267 Hansen, A. 457 Jin, W. 1361
Fletcher, C.J.N. 457 Harp, E.L. 1447 Jindal, V. 957
Floris, M. 1905 Hassani, H. 1911 Jing, R. 991
Fonseca, A.P. 409 Haugen, E.D. 381 Jing, Z.D. 783
Fornaro, G. 1103 He, H. 1839 Jomard, H. 421, 951
Franz, J. 1165 He, M. 1329 Jongmans, D. 323, 395
Fricout, B. 323 He, S. 795, 1745 Joo, Y.S. 1763
Frolovsky, Y.K. 963 He, S.M. 809 Ju, L.I. 1571
Fujisawa, K. 697 He, Y. 719, 737 Ju, N. 403
Fukuoka, H. 443 Hencher, S. 427 Jung, S.H. 257
Hencher, S.R. 457 Jung, Y.B. 1633
Gaeta, M. 1381 Heng, C. 1825 Jworchan, I. 1757
Gaffet, S. 711 Hernandez, M. 421, 725,
Galzerano, C.M. 1095 725 Kang, X.B. 1689
Gao, Q. 1795, 2075 Highland, L.M. 1173 Kao, T.C. 193
Gao, W. 535, 705 Higuchi, K. 971 Kawabata, H. 939
Garcia, C.F. 1463 Ho, K.K.S. 243, 1581, Kaynia, A.M. 381
Garcia, R.A.C. 1831 1595, 1769 Kazama, M. 1241
García de la Oliva, J.L. Hong, W.P. 1751 Keersmaekers, R. 761
1731 Hou, S. 1067 Kim, B.S. 257
Gattinoni, P. 1539 Hsu, H. 1621 Kim, H.W. 1183
Genda, M. 939 Hsu, S.C. 1409 Kim, J.H. 1297

1060
Kim, M. 1811 Li, X.P. 809 Manzella, I. 835
Kim, N.K. 1763 Li, Z.C. 1335 Maranto, G. 1111
Kim, S.K. 1763 Li, Z.D. 443 Martin, C.D. 3, 769
Kim, T.H. 1751, 1763 Lian, J.F. 2055 Martín, R. 1517
Klein, E. 1925 Liang, S.Y. 1189 Martini, E. 1609
Knappett, J.A. 803 Liao, H.J. 443 Martino, S. 579, 585, 1381,
Kniess, U. 395 Liao, X.P. 501, 783 1389
Krishna Murty, V.V.R.S. Lin, C.C. 1409 Mastroviti, G. 1111
231 Lin, H. 815 Mathy, A. 323
Krummel, H. 1083 Lin, M.L. 193 Matsiy, S.I. 1225
Ku, C.Y. 1639 Lin, W. 543 Matsuura, S. 1439
Kuai, Z. 679 Lin, X. 435 Mazengarb, C. 267
Kuhlmann, H. 1083 Lin, Z.G. 129 Mazzanti, P. 585, 1381,
Kumar, V. 957 Ling, J. 741 1905
Kuraoka, S. 697 Liu, B. 719 McDougall, S. 145
Kuroda, S. 971 Liu, C.Z. 2055 Meisina, C. 1233
Kwong, A.K.L. 1575 Liu, F.Q. 719 Mendes, J. 1553
Liu, H. 1421, Meng, Q.S. 979
Labiouse, V. 835 1775 Meng, R. 737
Lacerda, W.A. 409 Liu, H.C. 1493 Meng, X.W. 921
LaHusen, R.G. 179 Liu, J. 1415, 1415 Meric, O. 323
Lan, H. 769 Liu, J.F. 1933, 2001 Merrien-Soukatchoff, V.
Lang, Y. 449, 777 Liu, M. 389 1143
Lang, Y.H. 1189 Liu, P. 1681 Meyer, H.J. 1979
Lau, T.M.F. 243, 1581, Liu, S.T. 1211 Mihalinec, Z. 1587
1595, 1769 Liu, W.J. 719 Minardo, A. 1157
Law, R.P.H. 415 Liu, X.L. 821 Miner, A.S. 267
Lebourg, T. 421, 725, 951, Liu, Y.H. 2055 Mitani, Y. 893
1273 Liu, Z. 1421 Mitra, D. 927
Ledesma, A. 237 Lloret, A. 67 Miura, K. 843
Lee, C. 1297 Loew, S. 39 Mizuhashi, M. 53
Lee, C.F. 999, 1575 Lokin, P. 1939 Mohamed, A. 1873
Lee, J.F. 1639 Longobardi, V. 1095 Monfret, T. 711
Lee, J.S. 1297 Lu, G. 1067 Montagna, A. 1381
Lee, K.S. 427 Lu, G.P. 1689 Monterisi, L. 1525
Lee, S.G. 257, 427 Lu, S. 1071, 1077 Moreno Robles, J. 1731
Lee, W.C. 313 Lu, X.B. 263 Morgenstern, N.R. 3
Leng, J.Y. 783 Luan, M.T. 693 Mori, T. 1241
Lenti, L. 1389 Luan, Y.J. 719 Morimasa, S. 843
Leong, E.C. 67 Lui, J.F. 1933 Mugnozza, G.S. 355
Li, B.R. 1493 Luo, L. 2019 Muhidinov, D.Z. 1475
Li, D. 1985 Luo, X. 307, 449,
Li, D.L. 1805 777 Naaim, M. 835
Li, H. 1775 Luo, Y. 507 Nadim, C. 1925
Li, H.F. 1315 Luo, Y.L. 827 Nakamura, A. 851
Li, H.G. 1805 Luong, B.X. 671 Nakamura, H. 777
Li, H.P. 1799 Lupo, G. 365 Nakashima, Y. 697
Li, J.S. 249 Lv, M.J. 1805 Nancey, A. 1721
Li, J.T. 815 Ng, C.W.W. 159, 415
Li, K. 795 Ma, H. 2061 Nicoletti, P.G. 1433
Li, S. 803 Ma, H.M. 1217 Nitti, D.O. 513
Li, T. 435 Ma, Y. 1689 Niyazov, R.A. 1247
Li, T.B. 1781 Maertens, J. 573, 761 Noorzad, A. 553
Li, T.F. 1195, 2055 Mahboubi, A.R. 553 Notti, D. 1233
Li, X. 789, 1205, Mahdavifar, M. 1427 Nunes Bandeira A.P. 1887
1745 Malone, A.W. 457 Nutricato, R. 513

1061
O’Brien, A. 1757 Qiu, Y. 741 Shen, J.M. 1595
Ochiai, H. 1253, 1375 Quental Coutinho R. 1887 Shi, B. 1283
Okada, Y. 1253 Shi, L.L. 1959
Okamoto, T. 1439 Rahardjo, H. 67 Shi, Y.X. 921
Olalla, C. 347 Ramondini, M. 299, 1095 Shimomura, T. 53
Olivares, L. 203, 1157, Raymond, P. 1603 Shou, K. 1621
1951 Redaelli, M. 1309, 1965 Singh, R. 927
Oliveira, S.C. 1831 Refice, A. 513 Singh, V.K. 281
Omar, H. 1919 Reid, M.E. 179 Smethurst, J. 1553
Onoue, A. 971 Reis, E. 1831 Soares, J.M.D. 293
Orengo, Y. 395 Ren, W.Z. 991 Song, E.X. 1507
Ortolan, Ž. 1587 Renalier, F. 395 Song, J. 679
Ou, G.Q. 1933, 2001 Revellino, P. 365 Song, X. 1177
Owladeghaffari, H. 857 Rezaur, R.B. 67 Song, Y.S. 933, 1751
Rianna, G. 1259 Sorbino, G. 599
Paciello, A. 1389 Riopel, S. 1151 Srinivasa Gopalan, P. 231
Pagano, L. 1259 Risser, V. 725 Stedile, L. 355
Pan, H.L. 2001, 2001 Ritzkowski, C.M. 1475 Strom, A.L. 285
Pan, Y.B. 443 Rizakalla, E. 1757 Su, L.J. 443
Pang, K.K. 1595 Roh, B. 463 Su, O. 507
Pardo de Santayana, F. 1731 Rojat, F. 687 Su, T.W. 1639
Parise, M. 275, 299, 1095 Rollins, G.C. 1811 Su, X. 519
Park, D.C. 427 Rosser, N.J. 1265 Suemine, A. 1667
Park, H.D. 1633 Rossi, D. 1721 Sui, H. 1283
Park, H.J. 1943 Roy, D. 927 Suk, O.T. 1571
Park, J.S. 1763 Rui, R. 1011 Sun, C.-Y. 755
Park, W. 463 Rui, Z. 755 Sun, H.W. 1581, 1769
Pastor, M. 647 Rutqvist, J. 711 Sun, J.C. 1627
Pathier, E. 395 Sun, P. 631
Paulsen, H. 1083 Sadagah, B.H. 901 Sun, S. 1289, 2075
Pecci, M. 355 Sakai, H. 1277 Sun, Y. 507, 1329
Peduto, D. 1103 Sakai, N. 907 Sunitsakul, J. 1971
Peng, H. 827 Sakajo, S. 907 Sunwoo, C. 1633
Peng, X.M. 1321 Salciarini, D. 1609
Peng, X.T. 1031 Salimi, S.N. 469 Tacher, L. 217
Perri, F. 373 Santaloia, F. 1525 Taheri, A. 481
Petley, D.J. 865 Santarém Andrade, P.G.C. Takahashi, C. 1455
Petley, D.N. 865, 1265 561 Takeuchi, A. 1321
Phoon, K.K. 639 Santo, A. 1951 Tamburi, P. 1609
Piao, C. 1283 Sarah, D. 1615 Tan, C.H. 1639
Picarelli, L. 1157, 1301, Savio, G. 1233 Tanaka, N. 939
1951 Sawatparnich, A. 1971 Tang, C. 2055
Pinheiro, R.J.B. 293 Scarascia Mugnozza, G. Tang, H. 789, 881
Pirulli, M. 145, 835 1389 Tang, H.M. 249
Pisciotta, G. 1103 Scavia, C. 145 Tang, H.X. 945
Poisel, R. 873 Schaefer, V.R. 525, 1787 Tang, W.H. 415
Popescu, M.E. 487, 1787 Schauerte, W. 1083 Tang, X. 1795, 1863
Powrie, W. 1553 Schulz, W.H. 1447 Tensay, B.G. 1475
Preh, A. 873 Schwartz, S. 395 Terranova, C. 1095
Prestininzi, A. 579, 1381 Scotto di Santolo, A. 913 Tham, L.G. 999
Pun, W.K. 85 Seely, D.D.B. 475 Theule, J.I. 1979
Selkirk-Bell, J.M. 267 Thiebes, B. 1083
Qian, J.Y. 887 Sento, N. 1241 Tian, H. 1329
Qian, T. 881 Seo, Y.S. 933 Tiwari, B. 1463
Qiao, J. 1959, 1995 Serratrice, J.F. 835 Tohari, A. 1615
Qiu, C. 893, 1343 Shao, Y. 655 Tohei, M. 697

1062
Toll, D.G. 1553 Wang, S.Y. 263 Yang, H. 525, 1017, 1811,
Tommasi, P. 203 Wang, X.B. 1781 1991
Topal, T. 1647 Wang, X.Z. 979 Yang, J. 655
Tortoioli, L. 1609 Wang, Y. 507, 2061 Yang, W.T. 1367
Tosa, S. 1349 Wang, Y.G. 991 Yang, X. 1775
Towhata, I. 53 Wang, Y.J. 655, 663 Yang, Z. 1817, 1995
Trandafir, A.C. 475, 487 Wang, Z. 1985 Yao, A. 1825
Tric, E. 725, 951 Wang, Z.C. 991 Yashima, A. 1035
Truong, Q.H. 1297 Wang, Z.F. 1343 Yazdanjou, V. 469
Tsang, C.F. 711 Wang, Z.H. 2111 Ye, G.L. 1035
Tsui, H.M. 1581 Wang, Z.Q. 1315 Yener, A.B. 531
Tyagi, K. 957 Wang, Z.W. 1211 Yin, J.-H. 1361
Wasowski, J. 513 Yin, Y. 2005
Ugai, K. 25, 593, 851, 971, Watson, A.D. 39 Yin, Y.P. 2089
1023, 1455 Webb, S. 1475 Yoon, H.K. 1297
Um, J.G. 1943 Wei, G. 1283 Yoshimatsu, H. 939
Urciuoli, G. 85, 1259, 1301 Wei, J.B. 999 Yoshiya, O. 307
Uromeihy, A. 491 Wei, Y. 741 You, Y. 1933, 2001
Utili, S. 1309, 1965 Wen, M.S. 2055 Yuan, L. 1481
Uzuoka, R. 1241 Wesley, L.D. 1005 Yuan, P.J. 1335
White, D.J. 525 Yue, Z.Q. 535, 1697
Van Gemert, D. 761 Woo, I. 1943
van Westen, C.J. 1879 Wu, B. 1621 Zeni, L. 1157
Vanicek, I. 1653 Wu, C. 1017 Zervos, A. 615
Vardoulakis, I. 615 Wu, H. 1043 Zêzere, J.L. 1831
Vassallo, R. 495 Wu, M.J. 1335 Zhan, T.L.T. 731
Vázquez-Suñé, E. 1517 Wu, S.R. 1469 Zhang, D. 1283
Veveakis, E. 615 Wu, A.W. 519, 1475 Zhang, F. 679, 1035, 1067
Villemin, T. 323, 395 Wu, X. 1329 Zhang, G. 887
Vinale, F. 1259 Wu, Y. 1745 Zhang, H. 985
Vinogradov, V.V. 963 Wu, Z. 1355 Zhang, J. 507, 795, 1289
Virieux, J. 711 Zhang, J.-M. 887
Vitolo, E. 1893 Xia, Y. 1011 Zhang, K. 2005
Xian, C. 679 Zhang, L. 1839
Wakai, A. 939, 971 Xiang, R. 1825 Zhang, L.M. 1205, 1315,
Wang, A.X. 887 Xiao, D. 1017 1703
Wang, B. 1283 Xiao, W. 1817 Zhang, M.S. 129
Wang, D. 519 Xie, M. 893 Zhang, M.-S. 755
Wang, F.W. 1321, 1469 Xie, M.W. 1343 Zhang, P.W. 663
Wang, G. 1321, 1667, Xie, X.R. 2019 Zhang, Q. 921, 1367
1855, 1863, 2061 Xu, H. 1659, 1863 Zhang, S. 507, 795
Wang, G.H. 1469 Xu, M. 1689 Zhang, X.Y. 1367
Wang, G.Q. 1627 Xu, N. 1329 Zhang, Y. 985, 1355
Wang, H. 501 Xu, Q. 1675, 1681 Zhang, Y.M. 1321
Wang, H.B. 1469 Xu, X.S. 1031 Zhang, Y.S. 1493
Wang, H.S. 1493 Xu, Y. 1043 Zhang, Z. 1043, 1825
Wang, J. 1067, 1415 Xu, Z.J. 129 Zhang, Z.P. 1217
Wang, J.J. 1799 Xu, Z.P. 693 Zhang, Z.Y. 1675
Wang, J.-Y. 755 Zhao, H. 543
Wang, K.L. 193 Yamada, M. 1023, 1349 Zhao, J. 403
Wang, L. 1367 Yamagami, T. 747 Zhao, J.B. 1847
Wang, M.Y. 1659, 1659 Yan, B. 1031 Zhao, M. 2019
Wang, N. 519 Yan, E.C. 1805 Zhao, S. 1177
Wang, R. 543, 979 Yan, M. 389, 1355 Zhao, Y. 2011
Wang, S. 985, 1355, 2075 Yan, Z. 1421 Zheng, B. 1289
Wang, S.T. 1675 Yang, G. 389 Zheng, D. 1051

1063
Zheng, G.D. 1189 Zhou, J.P. 2101 Zhu, J.G. 1799
Zheng, J. 985, 1855 Zhou, P. 1067 Zielinski, M. 1309
Zheng, X. 1011 Zhu, B. 1289 Zingariello, M.C. 1259
Zhou, C. 507 Zhu, D.Y. 1055 Zolfaghari, M.R. 1427
Zhou, C.H. 769 Zhu, H.B. 2027 Zou, L. 1863
Zhou, G.D. 415 Zhu, H.-H. 1361 Zucca, F. 1233

1064
LANDSLIDES AND ENGINEERED SLOPES
PROCEEDINGS OF THE TENTH INTERNATIONAL SYMPOSIUM ON LANDSLIDES
AND ENGINEERED SLOPES, 30 JUNE–4 JULY, 2008, XI’AN, CHINA

Landslides and Engineered Slopes

From the Past to the Future

Editors

Zuyu Chen
China Institute of Water Resources and Hydropower Research, Beijing, China

Jianmin Zhang
Department of Hydraulic Engineering, Tsinghua University, Beijing, China

Zhongkui Li
Department of Hydraulic Engineering, Tsinghua University, Beijing, China

Faquan Wu
Institute of Geology and Geophysics, Chinese Academy of Sciences, Beijing, China

Ken Ho
Civil Engineering and Development Department, Hong Kong SAR, China

VOLUME 2
CRC Press/Balkema is an imprint of the Taylor & Francis Group, an informa business

© 2008 Taylor & Francis Group, London, UK

Typeset by Vikatan Publishing Solutions (P) Ltd., Chennai, India

Printed and bound in Great Britain by Antony Rowe (A CPI-group Company), Chippenham, Wiltshire.

All rights reserved. No part of this publication or the information contained herein may be reproduced, stored
in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, by photocopying,
recording or otherwise, without written prior permission from the publisher.

Although all care is taken to ensure integrity and the quality of this publication and the information herein, no
responsibility is assumed by the publishers nor the author for any damage to the property or persons as a result
of operation or use of this publication and/or the information contained herein.

Published by: CRC Press/Balkema

P.O. Box 447, 2300 AK Leiden, The Netherlands
e-mail: Pub.NL@taylorandfrancis.com
www.crcpress.com – www.taylorandfrancis.co.uk – www.balkema.nl

ISBN set: 978-0-415-41196-7 (2 volumes + CD)

ISBN Vol.1: 978-0-415-41194-3 (hbk)
ISBN Vol.2: 978-0-415-41195-0 (hbk)
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Table of Contents

Preface XXIII

VOLUME 1

Keynote lectures
Landslides: Seeing the ground 3
N.R. Morgenstern & C.D. Martin
Limit equilibrium and finite element analysis – A perspective of recent advances 25
Z. Chen & K. Ugai
Improving the interpretation of slope monitoring and early warning data through better
understanding of complex deep-seated landslide failure mechanisms 39
E. Eberhardt, A.D. Watson & S. Loew
Effects of earthquakes on slopes 53
I. Towhata, T. Shimomura & M. Mizuhashi
Monitoring and modeling of slope response to climate changes 67
H. Rahardjo, R.B. Rezaur, E.C. Leong, E.E. Alonso, A. Lloret & A. Gens
Soil nailing and subsurface drainage for slope stabilization 85
W.K. Pun & G. Urciuoli

Special lectures
Loess in China and landslides in loess slopes 129
Z.G. Lin, Z.J. Xu & M.S. Zhang
Advances in landslide continuum dynamic modelling 145
S. McDougall, M. Pirulli, O. Hungr & C. Scavia
Deformation and failure mechanisms of loose and dense fill slopes with and without soil nails 159
C.W.W. Ng
Capturing landslide dynamics and hydrologic triggers using near-real-time monitoring 179
M.E. Reid, R.L. Baum, R.G. LaHusen & W.L. Ellis
The effects of earthquake on landslides – A case study of Chi-Chi earthquake, 1999 193
M.L. Lin, K.L. Wang & T.C. Kao
The role of suction and its changes on stability of steep slopes in unsaturated granular soils 203
L. Olivares & P. Tommasi
Prediction of landslide movements caused by climate change: Modelling the behaviour
of a mean elevation large slide in the Alps and assessing its uncertainties 217
Ch. Bonnard, L. Tacher & M. Beniston

Geology, geotechnical properties and site characterization

Geotechnical appraisal of the Sonapur landslide area, Jainita hills, Meghalya, India 231
R.C. Bhandari, P. Srinivasa Gopalan & V.V.R.S. Krishna Murty

V
The viscous component in slow moving landslides: A practical case 237
D.A. González, A. Ledesma & J. Corominas
The systematic landslide investigation programme in Hong Kong 243
K.K.S. Ho & T.M.F. Lau
General digital camera-based experiments for large-scale landslide physical
model measurement 249
X.W. Hu, H.M. Tang & J.S. Li
Shear strength of boundaries between soils and rocks in Korea 257
S.G. Lee, B.S. Kim & S.H. Jung
Cracks in saturated sand 263
X.B. Lu, S.Y. Wang & P. Cui
Some geomorphological techniques used in constraining the likelihood
of landsliding – Selected Australian examples 267
A.S. Miner, P. Flentje, C. Mazengarb, J.M. Selkirk-Bell & P.G. Dahlhaus
Rock failures in karst 275
M. Parise
Geotechnical study at Sirwani landslide site, India 281
V.K. Singh
Inferences from morphological differences in deposits of similar large rockslides 285
A.L. Strom
Movements of a large urban slope in the town of Santa Cruz do Sul (RGS), Brazil 293
L.A. Bressani, R.J.B. Pinheiro, A.V.D. Bica, C.N. Eisenberger & J.M.D. Soares
Geotechnical analysis of a complex slope movement in sedimentary successions of the
southern Apennines (Molise, Italy) 299
D. Calcaterra, D. Di Martire, M. Ramondini, F. Calò & M. Parise
Application of surface wave and micro-tremor survey in landslide investigation
in the Three Gorges reservoir area 307
A. Che, X. Luo, S. Feng & O. Yoshiya
A case study for the landslide-induced catastrophic hazards in Taiwan Tuchang Tribute 313
C.Y. Chen & W.C. Lee
Pir3D, an easy to use three dimensional block fall simulator 319
Y. Cottaz & R.M. Faure
Characterization of the fracture pattern on cliff sites combining geophysical imaging
and laser scanning 323
J. Deparis, D. Jongmans, B. Fricout, T. Villemin, O. Meric, A. Mathy & L. Effendiantz
In situ characterization of the geomechanical properties of an unstable
fractured rock slope 331
C. Dünner, P. Bigarré, F. Cappa, Y. Guglielmi & C. Clément
Properties of peat relating to instability of blanket bogs 339
A.P. Dykes
Stability problems in slopes of Arenós reservoir (Castellón, Spain) 347
J. Estaire, J.A. Díez & C. Olalla
The 22 August, 2006, anomalous rock fall along the Gran Sasso NE wall
(Central Apennines, Italy) 355
G.B. Fasani, C. Esposito, G.S. Mugnozza, L. Stedile & M. Pecci

VI
New formulae to assess soil permeability through laboratory identification and flow
coming out of vertical drains 361
J.C. Gress
Structure-controlled earth flows in the Campania Apennines (Southern Italy) 365
F.M. Guadagno, P. Revellino, G. Grelle, G. Lupo & M. Bencardino
Geotechnical and mineralogical characterization of fine grained soils affected by soil slips 373
G. Gullà, L. Aceto, S. Critelli & F. Perri
Vulnerability of structures impacted by debris flow 381
E.D. Haugen & A.M. Kaynia
Engineering geological study on a large-scale toppling deformation
at Xiaowan Hydropower Station 389
R. Huang, G. Yang, M. Yan & M. Liu
Characterization of the Avignonet landslide (French Alps) with seismic techniques 395
D. Jongmans, F. Renalier, U. Kniess, S. Schwartz, E. Pathier, Y. Orengo,
G. Bièvre, T. Villemin & C. Delacourt
Deformation characteristics and treatment measures of spillway
slope at a reservoir in China 403
N. Ju, J. Zhao & R. Huang
Sliding in weathered banded gneiss due to gullying in southern Brazil 409
W.A. Lacerda, A.P. Fonseca & A.L. Coelho Netto
Experimental and three-dimensional numerical investigations of the impact of dry
granular flow on a barrier 415
R.P.H. Law, G.D. Zhou, C.W.W. Ng & W.H. Tang
Temporal survey of fluids by 2D electrical tomography: The ‘‘Vence’’ landslide observatory
site (Alpes-Maritimes, SE France) 421
T. Lebourg, S. El Bedoui, M. Hernandez & H. Jomard
Characteristics of landslides related to various rock types in Korea 427
S.G. Lee, K.S. Lee, D.C. Park & S. Hencher
Two approaches to identifying the slip zones of loess landslides and related issues 435
T. Li & X. Lin
Testing study on the strength and deformation characteristics of soil in loess landslides 443
H.J. Liao, L.J. Su, Z.D. Li, Y.B. Pan & H. Fukuoka
Failure mechanism of slipping zone soil of the Qiangjiangping landslide in the Three
Gorges reservoir area: A study based on Dead Load test 449
X. Luo, A. Che, L. Cao & Y. Lang
Post-failure movements of a large slow rock slide in schist near Pos Selim, Malaysia 457
A.W. Malone, A. Hansen, S.R. Hencher & C.J.N. Fletcher
Characteristics of rock failure in metamorphic rock areas, Korea 463
W. Park, Y. Han, S. Jeon & B. Roh
Shape and size effects of gravel grains on the shear behavior of sandy soils 469
S.N. Salimi, V. Yazdanjou & A. Hamidi
Nonlinear failure envelope of a nonplastic compacted silty-sand 475
D.D.B. Seely & A.C. Trandafir
An investigation of a structurally-controlled rock cut instability at a metro station shaft
in Esfahan, Iran 481
A. Taheri

VII
Yield acceleration of soil slopes with nonlinear strength envelope 487
A.C. Trandafir & M.E. Popescu
Evaluation of rockfall hazards along part of Karaj-Chaloos road, Iran 491
A. Uromeihy, N. Ghazipoor & I. Entezam
Coupled effect of pluviometric regime and soil properties on hydraulic boundary
conditions and on slope stability 495
R. Vassallo, C. Di Maio & M. Calvello
Mechanical characters of relaxing zone of slopes due to excavation 501
H. Wang & X.P. Liao
Deformation characteristics and stability evaluation of Ganhaizi landslide in the Dadu River 507
Y. Wang, Y. Sun, O. Su, Y. Luo, J. Zhang, C. Zhou & S. Zhang
Landslide-prone towns in Daunia (Italy): PS interferometry-based investigation 513
J. Wasowski, D. Casarano, F. Bovenga, A. Refice, R. Nutricato & D.O. Nitti
Basic types and active characteristics of loess landslide in China 519
W. Wu, D. Wang, X. Su & N. Wang
Investigation of a landslide using borehole shear test and ring shear test 525
H. Yang, V.R. Schaefer & D.J. White
The importance of geological and geotechnical investigations of landslides occurred
at dam reservoirs: Case studies from the Havuzlu and Demirkent landslides
(Artvin Dam, Turkey) 531
A.B. Yener, S. Durmaz & B.M. Demir
An innovative approach combining geological mapping and drilling process
monitoring for quantitative assessment of natural terrain hazards 535
Z.Q. Yue, J. Chen & W. Gao
Types of cutslope failures along Shiyan-Manchuanguan expressway through
the Liangyun fracture, Hubei Province 543
H. Zhao, R. Wang, J. Fan & W. Lin

Advances in analytical methods, modeling and prediction of slope behavior

Probability limit equilibrium and distinct element modeling of jointed rock slope at northern
abutment of Gotvand dam, Iran 553
M. Aminpoor, A. Noorzad & A.R. Mahboubi
Rock block sliding analysis of a highway slope in Portugal 561
P.G.C. Santarém Andrade & A.L. Almeida Saraiva
Contribution to the safety evaluation of slopes using long term observation results 567
J. Barradas
Delimitation of safety zones by finite element analysis 573
J. Bojorque, G. De Roeck & J. Maertens
Laboratory and numerical modelling of the lateral spreading process involving
the Orvieto hill (Italy) 579
F. Bozzano, S. Martino, A. Prestininzi & A. Bretschneider
Albano Lake coastal rock slide (Roma, Italy): Geological constraints and numerical modelling 585
F. Bozzano, C. Esposito, S. Martino, P. Mazzanti & G. Diano
Superposition principle for stability analysis of reinforced slopes and its FE validation 593
F. Cai & K. Ugai
Soil suction modelling in weathered gneiss affected by landsliding 599
M. Calvello, L. Cascini, G. Sorbino & G. Gullà

VIII
Modelling the transient groundwater regime for the displacements analysis of slow-moving
active landslides 607
L. Cascini, M. Calvello & G.M. Grimaldi
Numerical modelling of the thermo-mechanical behaviour of soils in catastrophic landslides 615
F. Cecinato, A. Zervos, E. Veveakis & I. Vardoulakis
Some notes on the upper-bound and Sarma’s methods with inclined slices for stability analysis 623
Z.Y. Chen
Slope stability analysis using graphic acquisitions and spreadsheets 631
L.H. Chen, Z.Y. Chen & P. Sun
Efficient evaluation of slope stability reliability subject to soil parameter uncertainties
using importance sampling 639
J. Ching, K.K. Phoon & Y.G. Hu
Prediction of the flow-like movements of Tessina landslide by SPH model 647
S. Cola, N. Calabrò & M. Pastor
Applications of the strength reduction finite element method to a gravity dam stability analysis 655
Q.W. Duan, Z.Y. Chen, Y.J. Wang, J. Yang & Y. Shao
Study on deformation parameter reduction technique for the strength reduction finite
element method 663
Q.W. Duan, Y.J. Wang & P.W. Zhang
Stability and movement analyses of slopes using Generalized Limit Equilibrium Method 671
M. Enoki & B.X. Luong
Long-term deformation prediction of Tianhuangpin ‘‘3.29’’ landslide based on neural
network with annealing simulation method 679
F. Zhang, C. Xian, J. Song, B. Guo & Z. Kuai
New models linking piezometric levels and displacements in a landslide 687
R.M. Faure, S. Burlon, J.C. Gress & F. Rojat
3D slope stability analysis of Rockfill dam in U-shape valley 693
X.Y. Feng, M.T. Luan & Z.P. Xu
3-D finite element analysis of landslide prevention piles 697
K. Fujisawa, M. Tohei, Y. Ishii, Y. Nakashima & S. Kuraoka
Integrated intelligent method for displacement predication of landslide 705
W. Gao
A new approach to in situ characterization of rock slope discontinuities: the ‘‘High-Pulse
Poroelasticity Protocol’’ (HPPP) 711
Y. Guglielmi, F. Cappa, S. Gaffet, T. Monfret, J. Virieux, J. Rutqvist & C.F. Tsang
Fuzzy prediction and analysis of landslides 719
Y. He, B. Liu, W.J. Liu, F.Q. Liu & Y.J. Luan
LPC methodology as a tool to create real time cartography of the gravitational hazard: Application
in the municipality of Menton (Maritimes Alps, France) 725
M. Hernandez, T. Lebourg, E. Tric, M. Hernandez & V. Risser
Back-analyses of a large-scale slope model failure caused by a sudden drawdown of water level 731
G.W. Jia, T.L.T. Zhan & Y.M. Chen
Effect of Guangxi Longtan reservoir on the stability of landslide at Badu station of Nankun railway 737
R. Jiang, R. Meng, A. Bai & Y. He
Application of SSRM in stability analysis of subgrade embankments over sloped weak
ground with FLAC3D 741
X. Jiang, Y. Qiu, Y. Wei & J. Ling

IX
Strength parameters from back analysis of slips in two-layer slopes 747
J.-C. Jiang & T. Yamagami
Development characteristics and mechanism of the Lianhua Temple landslide,
Huaxian county, China 755
J.-Y. Wang, M.-S. Zhang, C.-Y. Sun & Z. Rui
Modeling landslide triggering in layered soils 761
R. Keersmaekers, J. Maertens, D. Van Gemert & K. Haelterman
Numerical modeling of debris flow kinematics using discrete element method combined with GIS 769
H. Lan, C.D. Martin & C.H. Zhou
Three dimensional simulation of landslide motion and the determination of geotechnical
parameters 777
Y. Lang, X. Luo & H. Nakamura
Stability analysis and stabilized works of dip bedded rock slopes 783
J.Y. Leng, Z.D. Jing & X.P. Liao
A GIS-supported logistic regression model applied in regional slope stability evaluation 789
X. Li, H. Tang & S. Chen
The stability analysis for FaNai landslide in Lubuge hydropower station 795
K. Li, J. Zhang, S. Zhang & S. He
Numerical analysis of slope stability influenced by varying water conditions in the reservoir
area of the Three Gorges, China 803
S. Li, X. Feng & J.A. Knappett
A numerical study of interaction between rock bolt and rock mass 809
X.P. Li & S.M. He
Macroscopic effects of rock slopes before and after grouting of joint planes 815
H. Lin, P. Cao, J.T. Li & X.L. Jiang
Two- and three-dimensional analysis of a fossil landslide with FLAC 821
X.L. Liu & J.H. Deng
Application of the coupled thin-layer element in forecasting the behaviors of landslide
with weak intercalated layers 827
Y.L. Luo & H. Peng
Numerical modelling of a rock avalanche laboratory experiment in the framework
of the ‘‘Rockslidetec’’ alpine project 835
I. Manzella, M. Pirulli, M. Naaim, J.F. Serratrice & V. Labiouse
Three-dimensional slope stability analysis by means of limit equilibrium method 843
S. Morimasa & K. Miura
Embankment basal stability analysis using shear strength reduction finite element method 851
A. Nakamura, F. Cai & K. Ugai
Back analysis based on SOM-RST system 857
H. Owladeghaffari & H. Aghababaei
Temporal prediction in landslides – Understanding the Saito effect 865
D.N. Petley, D.J. Petley & R.J. Allison
3D landslide run out modelling using the Particle Flow Code PFC3D 873
R. Poisel & A. Preh
Double-row anti-sliding piles: Analysis based on a spatial framework structure 881
T. Qian & H. Tang

X
Centrifuge modeling of rainfall-induced failure process of soil slope 887
J.Y. Qian, A.X. Wang, G. Zhang & J.-M. Zhang
A GIS-based method for predicting the location, magnitude and occurrence time of
landslides using a three-dimensional deterministic model 893
C. Qiu, T. Esaki, Y. Mitani & M. Xie
Application of a rockfall hazard rating system in rock slope cuts along a mountain road of
South Western Saudi Arabia 901
B.H. Sadagah
Model tests of collapse of unsaturated slopes in rainfall 907
N. Sakai & S. Sakajo
Calibration of a rheological model for debris flow hazard mitigation in the Campania region 913
A. Scotto di Santolo & A. Evangelista
Optical fiber sensing technology used in landslide monitoring 921
Y.X. Shi, Q. Zhang & X.W. Meng
Finite element analysis of flow failure of Tailings dam and embankments 927
R. Singh, D. Mitra & D. Roy
Landslide model test to investigate the spreading range of debris according to rainfall intensity 933
Y.S. Song, B.G. Chae, Y.C. Cho & Y.S. Seo
Occurrence mechanism of rockslide at the time of the Chuetsu earthquake in 2004 – A dynamic
response analysis by using a simple cyclic loading model 939
N. Tanaka, S. Abe, A. Wakai, H. Kawabata, M. Genda & H. Yoshimatsu
Analysis for progressive failure of the Senise landslide based on Cosserat continuum model 945
H.X. Tang
Large-scale deformation of the La Clapière landslide and its numerical modelling
(S.-E. de Tinée, France). 951
E. Tric, T. Lebourg & H. Jomard
A novel complex valued neuron model for landslide assessment 957
K. Tyagi, V. Jindal & V. Kumar
Prediction of slope behavior for deforming railway embankments 963
V.V. Vinogradov, Y.K. Frolovsky, A. Al. Zaitsev & I.V. Ivanchenko
Finite element simulation for the collapse of a dip slope during 2004 Mid Niigata Prefecture
earthquake in Japan 971
A. Wakai, K. Ugai, A. Onoue, K. Higuchi & S. Kuroda
Sensitivity of stability parameters for soil slopes: An analysis based on the shear
strength reduction method 979
R. Wang, X.Z. Wang, Q.S. Meng & B. Hu
Back analysis of unsaturated parameters and numerical seepage simulation of the Shuping
landslide in Three Gorges reservoir area 985
S. Wang, H. Zhang, Y. Zhang & J. Zheng
Slope failure criterion: A modification based on strength reduction technique 991
Y.G. Wang, R. Jing, W.Z. Ren & Z.C. Wang
Unsaturated seepage analysis for a reservoir landslide during impounding 999
J.B. Wei, J.H. Deng, L.G. Tham & C.F. Lee
A simple compaction control method for slope construction 1005
L.D. Wesley

XI
Numerical analysis of soil-arch effect of anti-slide piles 1011
Y. Xia, X. Zheng & R. Rui
Determination of the critical slip surface based on stress distributions from FEM 1017
D. Xiao, C. Wu & H. Yang
Effect of drainage facilities using 3D seepage flow analysis reflecting hydro-geological
structure with aspect cracks in a landslide – Example of analysis in OODAIRA Landslide area 1023
M. Yamada & K. Ugai
Back analysis of soil parameters: A case study on monitored displacement of foundation pits 1031
B. Yan, X.T. Peng & X.S. Xu
3D finite element analysis on progressive failure of slope due to rainfall 1035
G.L. Ye, F. Zhang & A. Yashima
Block-group method for rock slope stability analysis 1043
Z. Zhang, Y. Xu & H. Wu
Quantitative study on the classification of unloading zones of high slope 1051
D. Zheng & R.Q. Huang
Investigations on the accuracy of the simplified Bishop method 1055
D.Y. Zhu
Author index 1059

VOLUME 2

Landslide mechanism, monitoring and warning

GIS-based landslide susceptibility mapping in the Three Gorges area – Comparisons of mapping
results obtained by two methods: Analytical hierarchy process and logistic regression 1067
S. Bai, J. Wang, G. Lu, P. Zhou, S. Hou & F. Zhang
Importance of study of creep sliding mechanism to prevention and treatment
of reservoir landslide 1071
J. Bai, S. Lu, J. Han
Stability prediction of landsides before and after impoundment for Lijiaxia hydropower station 1077
J. Bai, S. Lu & J. Han
The technical concept within the Integrative Landslide Early Warning System (ILEWS) 1083
R. Bell, B. Thiebes, T. Glade, R. Becker, H. Kuhlmann, W. Schauerte, S. Burghaus, H. Krummel,
M. Janik & H. Paulsen
The Åknes rockslide: Monitoring, threshold values and early-warning 1089
L.H. Blikra
DInSAR techniques for monitoring slow-moving landslides 1095
D. Calcaterra, M. Ramondini, F. Calò, V. Longobardi, M. Parise, C.M. Galzerano & C. Terranova
Multitemporal DInSAR data and damages to facilities as indicators for the activity of
slow-moving landslides 1103
L. Cascini, S. Ferlisi, D. Peduto, G. Pisciotta, S. Di Nocera & G. Fornaro
The Serre La Voute Landslide (North-West Italy): Results from ten years of monitoring 1111
M. Ceccucci, G. Maranto & G. Mastroviti
Onset of rockslide by the peak-residual strength drop 1119
Q.G. Cheng & G.T. Hu
Analysis of mechanism of the K31 landslide of Changzhi-Jincheng express highway 1127
Y. Cheng

XII
A plane-torsion rockslide with a locked flank: A case study 1133
Q. Cheng
Monitoring of natural thermal strains using hollow cylinder strain cells: The case of a large
rock slope prone to rockfalls 1143
C. Clément, Y. Gunzburger, V. Merrien-Soukatchoff & C. Dünner
Landslide hazards mapping and permafrost slope InSAR monitoring, Mackenzie valley,
Northwest Territories, Canada 1151
R. Couture & S. Riopel
Advanced monitoring criteria for precocious alerting of rainfall-induced flowslides 1157
E. Damiano, L. Olivares, A. Minardo, R. Greco, L. Zeni & L. Picarelli
Investigation of slope failure mechanisms caused by discontinuous large scale geological
structures at the Cadia Hill Open Pit 1165
J. Franz & Y. Cai
Two approaches for public landslide awareness in the United States – U.S. geological survey
warning systems and a landslide film documentary 1173
L.M. Highland & P.L. Gori
Formation and mechanical analysis of Tiantai landslide of Xuanhan county, Sichuan province 1177
R.Q. Huang, S. Zhao & X. Song
Development of wireless sensor node for landslide detection 1183
H.W. Kim
Redox condition and landslide development 1189
Y.H. Lang, S.Y. Liang & G.D. Zheng
Prepa displacement mechanism and its treatment measures for Hancheng landslide 1195
T.F. Li & L.C. Dang
Investigation of the stability of colluvial landslide deposits 1205
X. Li & L.M. Zhang
Choice of surveying methods for landslides monitoring 1211
S.T. Liu & Z.W. Wang
No. 1 landslide on the eastern approach road to ErLang Mountain tunnel: Inference
factors and controlling measures 1217
H.M. Ma & Z.P. Zhang
Estimation of landslide load on multi-tier pile constructions with the help of a combined method 1225
S.I. Matsiy & Ph.N. Derevenets
The use of PSInSAR™ data in landslide detection and monitoring: The example of the
Piemonte region (Northern Italy) 1233
C. Meisina, F. Zucca, D. Notti, A. Colombo, A. Cucchi, G. Savio, C. Giannico & M. Bianchi
Fill slopes: Stability assessment based on monitoring during both heavy rainfall
and earthquake motion 1241
T. Mori, M. Kazama, R. Uzuoka & N. Sento
The mechanism of movement of mud flows in loess soils, successful and unsuccessful
cases of forecast 1247
R.A. Niyazov, Sh.B. Bazarov & A.M. Akhundjanov
Influence of fine soil particles on excess of basal pore-fluid pressure generation in granular
mass flows 1253
Y. Okada & H. Ochiai
An early warning system to predict flowslides in pyroclastic deposits 1259
L. Pagano, G. Rianna, M.C. Zingariello, G. Urciuoli & F. Vinale

XIII
Monitoring and modeling of slope movement on rock cliffs prior to failure 1265
N.J. Rosser & D.N. Petley
Active tectonic control of a large landslide: Results from Panagopoula landslide multi
parametric analyses 1273
S. El Bedoui, T. Lebourg & Y. Guglielmi
A warning system using chemical sensors and telecommunication technologies to
protect railroad operation from landslide disaster 1277
H. Sakai
Distributive monitoring of the slope engineering 1283
B. Shi, H. Sui, D. Zhang, B. Wang, G. Wei & C. Piao
Observational method in the design of high cutting slope around bridge 1289
S. Sun, B. Zhu, B. Zheng & J. Zhang
Ultrasonic monitoring of lab-scaled underwater landslides 1297
Q.H. Truong, C. Lee, H.K. Yoon, Y.H. Eom, J.H. Kim & J.S. Lee
Interaction between landslides and man-made works 1301
G. Urciuoli & L. Picarelli
Desiccation fissuring induced failure mechanisms for clay levees 1309
S. Utili, M. Dyer, M. Redaelli & M. Zielinski
Stability analysis by strength reduction finite element method and monitoring of
unstable slope during reinforcement 1315
Z.Q. Wang, H.F. Li & L.M. Zhang
Displacement monitoring on Shuping landslide in the Three Gorges Dam reservoir area,
China from August 2004 to July 2007 1321
F.W. Wang, G. Wang, Y.M. Zhang, Z.T. Huo, X.M. Peng, K. Araiba & A. Takeuchi
Deformation mechanism and prevention measure for strongly expansive soft-rock slope in
the Yanji basin 1329
X. Wu, N. Xu, H. Tian, Y. Sun & M. He
Twenty years of safety monitoring for the landslide of Hancheng PowerStation 1335
M.J. Wu, Z.C. Li, P.J. Yuan & Y.H. Jiang
A time-spatial deterministic approach to assessment of rainfall-induced shallow landslide 1343
M.W. Xie, C. Qiu & Z.F. Wang
Introduction of web-based remote-monitoring system and its application to landslide
disaster prevention 1349
M. Yamada & S. Tosa
Deformation mechanism for the front slope of the left bank deposits in Xiluodu
hydro-electrical power station, China 1355
M. Yan, Z. Wu, R. Huang, Y. Zhang & S. Wang
Monitoring of soil nailed slopes and dams using innovative technologies 1361
J.-H. Yin, H.-H. Zhu & W. Jin
Application of multi-antenna GPS technique in the stability monitoring of roadside slopes 1367
Q. Zhang, L. Wang, X.Y. Zhang, G.W. Huang, X.L. Ding, W.J. Dai & W.T. Yang

Effects of earthquakes on slopes

Influences of earthquake motion on slopes in a hilly area during the Mid-Niigata
Prefecture Earthquake, 2004 1375
S. Asano & H. Ochiai

XIV
The 1783 Scilla rock-avalanche (Calabria, southern Italy) 1381
F. Bozzano, S. Martino, A. Prestininzi, M. Gaeta, P. Mazzanti & A. Montagna
Self-excitation process due to local seismic amplification and earthquake-induced
reactivations of large landslides 1389
F. Bozzano, S. Martino, G. Scarascia Mugnozza, A. Paciello & L. Lenti
Geological constraints to the urban shape evolution of Ariano Irpino (Avellino province, Italy) 1397
D. Calcaterra, C. Dima & E. Grasso
Landslide zones and their relation with seismoactive fault systems in Azerbaijan, Iran 1405
E. Ghanbari
Ground movements caused by lateral spread during an earthquake 1409
S.C. Hsu, B.L. Chu & C.C. Lin
2-D analysis of slope stability of an infinite slope during earthquake 1415
J. Liu, J. Liu & J. Wang
High-cutting slopes at Qingshuichuan electric power plant in the North of Shaanxi:
Deformation and failure modes and treatment scheme 1421
H. Liu, Z. Liu & Z. Yan
GIS-based real time prediction of Arias intensity and earthquake-induced landslide hazards
in Alborz and Central Iran 1427
M. Mahdavifar, M.K. Jafari & M.R. Zolfaghari
Geomorphology of old earthquake-induced landslides in southeastern Sicily 1433
P.G. Nicoletti & E. Catalano
Coseismic movement of an active landslide resulting from the Mid-Niigata Prefecture
Earthquake, Japan 1439
T. Okamoto, S. Matsuura & S. Asano
Characteristics of large rock avalanches triggered by the November 3, 2002 Denali
Fault earthquake, Alaska, USA 1447
W.H. Schulz, E.L. Harp & R.W. Jibson
FE analysis of performance of the Lower and Upper San Fernando Dams under the 1971
San Fernando earthquake 1455
C. Takahashi, F. Cai & K. Ugai
Reduction of the stability of pre-existing landslides during earthquake 1463
B. Tiwari, I. Dhungana & C.F. Garcia
Probabilistic hazard mapping of earthquake-induced landslides 1469
H.B. Wang, S.R. Wu, G.H. Wang & F.W. Wang
Investigation on stability of landfill slopes in seismically active regions in Central Asia 1475
A.W. Wu, B.G. Tensay, S. Webb, B.T. Doanth, C.M. Ritzkowski, D.Z. Muhidinov & E.M. Anarbaev
Mechanism for loess seismic landslides in Northwest China 1481
L. Yuan, X. Cui, Y. Hu & L. Jiang

Climate, hydrology and landslides

Evaluation of the landslide potential in Chahr Chay dam reservoir slopes 1489
K. Badv & K. Emami
Effect of well pumping on groundwater level and slope stability in the Taiwan
Woo-Wan-Chai landslide area 1493
M. Chang, B.R. Li, Y.S. Zhang, H.S. Wang, Y.H. Chou & H.C. Liu

XV
Case study: Embankment failure of Cable-Ski Lake development in Cairns 1501
K. Chen
Analysis method for slope stability under rainfall action 1507
X.D. Chen, H.X. Guo & E.X. Song
Hydrological modelling of the Vallcebre landslide 1517
J. Corominas, R. Martín & E. Vázquez-Suñé
Landslides in stiff clay slopes along the Adriatic coast (Central Italy) 1525
F. Cotecchia, O. Bottiglieri, L. Monterisi & F. Santaloia
Research on the effect of atomized rain on underground water distribution in Dayantang landslide 1533
J. Ding
Landslide hydrogeological susceptibility in the Crati valley (Italy) 1539
P. Gattinoni
Sustainable landslide stabilisation using deep wells installed with siphon drains
and electro-pneumatic pumps 1547
A. Gillarduzzi
Biological and engineering impacts of climate on slopes – learning from full-scale 1553
S. Glendinning, P.N. Hughes, D.A.B. Hughes, D. Clarke, J. Smethurst, W. Powrie, N. Dixon,
T.A. Dijkstra, D.G. Toll & J. Mendes
Some attributes of road-slope failure caused by typhoons 1559
M.W. Gui, C.H. Chang & S.F. Chen
A small rock avalanche in toppled schist, Lake Wanaka, New Zealand 1565
G.S. Halliday
NRCS-based groundwater level analysis of sloping ground 1571
L.I. Ju, O.T. Suk, M.Y. Il & L.S. Gon
A numerical case study on load developments along soil nails installed in cut slope subjected
to high groundwater table 1575
A.K.L. Kwong & C.F. Lee
Landslides at active construction sites in Hong Kong 1581
T.M.F. Lau, H.W. Sun, H.M. Tsui & K.K.S. Ho
Landslide ‘‘Granice’’ in Zagreb (Croatia) 1587
Z. Mihalinec & Ž. Ortolan
Improvement of subsurface drainage provisions for recompacted soil fill slopes in Hong Kong 1595
K.K. Pang, J.M. Shen, K.K.S. Ho & T.M.F. Lau
Biotechnical slope stabilization and using Spyder Hoe to control steep slope failure 1603
P. Raymond
Rapid landslides threatening roads: Three case histories of risk mitigation in the Umbria
region of Central Italy 1609
D. Salciarini, P. Conversini, E. Martini, P. Tamburi & L. Tortoioli
Assessment of the slope stabilisation measures at the Cadas Pangeran road section,
Sumedang, West Java 1615
D. Sarah, A. Tohari & M.R. Daryono
Analysis of control factors on landslides in the Taiwan area 1621
K. Shou, B. Wu & H. Hsu
Inclined free face riverbank collapse by river scouring 1627
J.C. Sun & G.Q. Wang
Drainage control and slope stability at an open pit mine: A GIS-based hydrological modeling 1633
C. Sunwoo, Y.S. Choi, H.D. Park & Y.B. Jung

XVI
Assessment of regional rainfall-induced landslides using 3S-based hydro-geological model 1639
C.H. Tan, C.Y. Ku, S.Y. Chi, Y.H. Chen, L.Y. Fei, J.F. Lee & T.W. Su
Investigation of a landslide along a natural gas pipeline (Karacabey-Turkey) 1647
T. Topal & M. Akin
Influence of extreme rainfall on the stability of spoil heaps 1653
I. Vanicek & S. Chamra
Behavior of expansive soil slope reinforced with geo-grids 1659
M.Y. Wang, X.N. Gong, M.Y. Wang, J.T. Cai & H. Xu
Geotechnical properties for a rainstorm-triggered landslide in Kisawa village,
Tokushima Prefecture, Japan 1667
G. Wang & A. Suemine
Yigong rock avalanche-flow landslide event, Tibet, China 1675
Q. Xu, S.T. Wang, H.J. Chai, Z.Y. Zhang & S.M. Dong
Key issues of emergency measures and comprehensive remediation projects to control
the Danba landslide, Sichuan province, China 1681
Q. Xu, X.-M. Fan, L.-W. Jiang & P. Liu
Enhanced slope seepage resulting from localized torrential precipitation during a
flood discharge event at the Nuozhadu hydroelectric station 1689
M. Xu, Y. Ma, X.B. Kang & G.P. Lu
An issue in conventional approach for drainage design on slopes in mountainous regions 1697
Z.Q. Yue
Analysis of geo-hazards caused by climate changes 1703
L.M. Zhang

Slope stabilization and protection

Back experience of deep drainage for landslide stabilization through lines of siphon drains and
electro-pneumatics drains: A French railway slope stabilization example 1713
S. Bomont
Experimental geo-synthetic-reinforced segmental wall as bridge abutment 1721
R.M. Faure, D. Rossi, A. Nancey & G. Auray
Rock slope stability analysis for a slope in the vicinity of Take-off Yard of Karun-3 Dam 1725
M. Gharouni-Nik
Stabilization of a large paleo-landslide reactivated because of the works to install
a new ski lift in Formigal skiing resort 1731
J. González-Gallego, J. Moreno Robles, J.L. Garcı́a de la Oliva & F. Pardo de Santayana
A case study on rainfall infiltration effect on the stability of two slopes 1737
M.W. Gui & K.K. Han
Consolidation mechanism of fully grouted anchor bolts 1745
S. He, Y. Wu & X. Li
Stability analysis for cut slopes reinforced by an earth retention system by considering
the reinforcement stages 1751
W.P. Hong, Y.S. Song & T.H. Kim
Landslide stabilization for residential development 1757
I. Jworchan, A. O’Brien & E. Rizakalla
Influence of load transfer on anchored slope stability 1763
S.K. Kim, N.K. Kim, Y.S. Joo, J.S. Park, T.H. Kim & K.S. Cha

XVII
Review of slope surface drainage with reference to landslide studies and current practice
in Hong Kong 1769
T.M.F. Lau, H.W. Sun, T.H.H. Hui & K.K.S. Ho
Analysis of dynamic stability about prestressed anchor retaining structure 1775
H. Li, X. Yang, H. Liu & L. Du
Safety analysis of high engineering slopes along the west approach road of ZheGu
mountain tunnel 1781
T.B. Li, Y. Du & X.B. Wang
Landslide stabilizing piles: A design based on the results of slope failure back analysis 1787
M.E. Popescu & V.R. Schaefer
Landslides on the left abutment and engineering measures for Manwan hydropower project 1795
X. Tang & Q. Gao
Factors resulting in the instability of a 57.5 m high cut slope 1799
J.J. Wang, H.J. Chai, H.P. Li & J.G. Zhu
Orthogonal analysis and applications on anchorage parameters of rock slopes 1805
E.C. Yan, H.G. Li, M.J. Lv & D.L. Li
Waste rock dump slope stability for a gold mine in California 1811
H. Yang, G.C. Rollins & M. Kim
Properties of the high rock slope of Hongjiadu hydropower project and its engineering
treatment measures 1817
Z. Yang, W. Xiao & D. Cai
Typical harbor bank slopes in the Three Gorges reservoir: Landslide and collapse and their
stability control 1825
A. Yao, C. Heng, Z. Zhang & R. Xiang
Weighting predisposing factors for shallow slides susceptibility assessment
on the regional scale 1831
J.L. Zêzere, S.C. Oliveira, R.A.C. Garcia & E. Reis
Analyses of mechanism of landslides in Tongchuan-Huangling highway 1839
L. Zhang & H. He
Treatments of Loess-Bedrock landslides at Chuankou in Tongchuan-Huangling expressway 1847
J.B. Zhao
Types, characteristics and application conditions of anti-slide retaining structures 1855
J. Zheng & G. Wang
The stabilization of the huge alluvial deposit on the left bank and the high rock slope on
the right bank of the XiaoWan Hydropower Project 1863
L. Zou, X. Tang, H. Feng, G. Wang & H. Xu

Risk assessment and management

Malaysian National Slope Master Plan – Challenges to producing an effective plan 1873
C.H. Abdullah & A. Mohamed
Spatial landslide risk assessment in Guantánamo Province, Cuba 1879
E. Castellanos Abella & C.J. van Westen
Landslide risk management: Experiences in the metropolitan area
of Recife – Pernambuco, Brazil 1887
A.P. Nunes Bandeira, & R. Quental Coutinho

XVIII
Societal risk due to landslides in the Campania region (Southern Italy) 1893
L. Cascini, S. Ferlisi & E. Vitolo
Landslide risk in the San Francisco Bay region 1899
J.A. Coe & R.A. Crovelli
A first attempt to extend a subaerial landslide susceptibility analysis to submerged slopes:
The case of the Albano Lake (Rome, Italy) 1905
G.B. Fasani, C. Esposito, F. Bozzano, P. Mazzanti & M. Floris
Landslide susceptibility zonation of the Qazvin-Rasht-Anzali railway track, North Iran 1911
H. Hassani & M. Ghazanfari
Assessment of landslide hazard of a cut-slope using linear regression analysis 1919
S. Jamaludin, B.B.K. Huat & H. Omar
Global monitoring strategy applied to ground failure hazards 1925
E. Klein, C. Nadim, P. Bigarré & C. Dünner
Regional slope stability zonation based on the factor overlapping method 1933
J.F. Liu, G.Q. Ou, Y. You & J.F. Lui
Landslide hazard and risk assessment in the areas of dams and reservoirs of Serbia 1939
P. Lokin & B. Abolmasov
The evaluation of failure probability for rock slope based on fuzzy set theory
and Monte Carlo simulation 1943
H.J. Park, J.G. Um & I. Woo
Macro-zoning of areas susceptible to flowslide in pyroclastic soils in the Campania region 1951
L. Picarelli, A. Santo, G. Di Crescenzo & L. Olivares
Zoning methods for landslide hazard degree 1959
J. Qiao & L.L. Shi
A proposal for a reliability rating system for fluvial flood defence embankments in
the United Kingdom 1965
M. Redaelli, S. Utili & M. Dyer
Simplified risk analysis chart to prevent slope failure of highway embankment on soft
Bangkok clays 1971
A. Sawatparnich & J. Sunitsakul
Determining landslide susceptibility along natural gas pipelines in Northwest Oregon, USA 1979
J.I. Theule, S.F. Burns & H.J. Meyer
Landslide susceptibility assessment using fuzzy logic 1985
Z. Wang, D. Li & Q. Cheng
Prediction of the spatiotemporal distribution of landslides: Integrated landslide susceptibility
zoning techniques and real-time satellite rainfall 1991
H. Yang, R.F. Adler, G.J. Huffman & D. Bach
Entropy based typical landslide hazard degree assessment in Three Gorges 1995
Z. Yang & J. Qiao
The optimal hydraulic cross-section design of the ‘‘Trapezoid-V’’ shaped drainage canal
of debris flow 2001
Y. You, H.L. Pan, J.F. Liu, G.Q. Ou & H.L. Pan
Practice of establishing China’s Geo-Hazard Survey Information System 2005
K. Zhang, Y. Yin & H. Chen
A XML-supported database for landslides and engineered slopes related to China’s water
resources development 2011
Y. Zhao & Z. Chen

XIX
Landslide and engineered slopes in China
Failure and treatment technique of a canal in expansive soil in South to North
Water Diversion project 2019
Y.J. Cai, X.R. Xie, L. Luo, S.F. Chen & M. Zhao
High slope engineering for Three Gorges ship locks 2027
G.J. Cao & H.B. Zhu
Large-scale landslides in China: Case studies 2037
R.Q. Huang
Early warning for Geo-Hazards based on the weather condition in China 2055
C.Z. Liu, Y.H. Liu, M.S. Wen, C. Tang, T.F. Li & J.F. Lian
Slope engineering in railway and highway construction in China 2061
G. Wang, H. Ma, M. Feng & Y. Wang
Mining slope engineering in China 2075
S. Wang, Q. Gao & S. Sun
Structure and failure patterns of engineered slopes at the Three Gorges reservoir 2089
Y.P. Yin
Slope engineering in hydropower projects in China 2101
J.P. Zhou & G.F. Chen
A thunder at the beginning of the 21st century – The giant Yigong Landslide 2111
Z.H. Wang
Author index 2119

XX
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Preface

The city of Xi’an, China is privileged to have the honor of hosting the 10th International Symposium on
Landslides and Engineered Slopes, following its predecessors: Rio de Janeiro, Brazil, 2004; Cardiff, U.K.,
2000; Trondheim, Norway, 1996; Christchurch, New Zealand, 1992; Lausanne, Switzerland, 1988; Toronto,
Canada, 1984; New Delhi, India, 1980; Tokyo, Japan, 1977; and Kyoto, Japan, 1972.
China is one of the countries in the world that suffer severely from landslide hazards. Statistics have shown
that every year 700 to 900 people are killed by landslides. With the large scale infrastructure construction,
failures of engineered slopes are increasing, and have become a serious concern of the government, various
enterprises and technical societies. The Chinese geological and geotechnical communities look forward to this
unique opportunity of exchanging and sharing technical know-how and experience of combating landslides
disasters with our international peers.
Xi’an is a historical city of China. It has been the capital for China’s twelve dynasties, spanning over
1200 years, and was a starting point of the famous Silk Road. It is a nice place for participants from all
over the world to meet and review our past experience, and in the meanwhile, look forward to a productive future
against landslides and slope failures.
From this book, readers will find that there are 7 Symposium Themes, as allocated by the Steering Committee,
including 13 keynote and special papers. At the JTC1 meeting held at the 9th Symposium, it was decided that
a special session entitled ‘China Afternoon’ would be organized, whose papers are arranged under a separate
theme ’Landslides and Engineered Slopes in China’ in this Proceedings.
This symposium was jointly organized by the Chinese Institution of Soil Mechanics and Geotechnical
Engineering-CCES, Chinese National Commission on Engineering Geology, Chinese Society of Rock Mechan-
ics and Engineering, and the Geotechnical Division of the Hong Kong Institution of Engineers. The Organizing
Committee is grateful to the reviewers for reviewing more than 300 submitted papers. It is practically impossible
to list the large number of these volunteers here. However, their contributions must be fully acknowledged,
without which the quality of this book would not have been maintained.
Special thanks also go to the members of JTC1 for their constant attention and useful comments during the
preparation of this symposium.
Zuyu Chen

XXI
Landslide mechanism, monitoring and warning
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

GIS-based landslide susceptibility mapping in the Three Gorges area –

Comparisons of mapping results obtained by two methods:
Analytical hierarchy process and logistic regression

Shibiao Bai, Jian Wang & Guonian Lu

National Education Administration Key Laboratory of Virtual Geographic Environments, Nanjing Normal University,
Nanjing, China

Pinggen Zhou & Shengshan Hou

Chinese Institute of Geological Environment Monitoring, Beijing, China

Fanyu Zhang
School of Civil Engineering and Mechanics, Lanzhou University, Lanzhou, China

ABSTRACT: The study aims at evaluating and comparing the results of GIS based landslide susceptibility
mapping by two methods: the analytical hierarchy process and the logistic regression model. For the purpose
of this study, the Zhongxian segment in the Three Gorges area was selected as a test area because of its high
frequency of historical landslides. This area covers an area of 260.9 km2 , in which 5.3 km2 is occupied with
landslide. A detailed landslide inventory map of the study area was drawn up by interpreting 1:20,000-scale
color aerial photographs and through extensive field surveys. Four types of data were used in this study: remote
sensing products, thematic maps, geological maps, and topographical maps. The size of pixels for all of the
data layers is 25 × 25 m2 . The analytical hierarchy process and logistic regression methods were applied in
the test area, and two separate susceptibility maps were produced. The susceptibility maps produced by these
two methods were validated and compared. This study indicates that the logistic regression map was better in
matching the historical landslide density than the analytical hierarchy process method.

1 INTRODUCTION for each class of an independent parameter (Guzzetti

et al. 1999, Lee & Min, 2001, Dai & Lee, 2002,
Historically, the Three Gorges Region of China is an Ohlmacher & Davis, 2003). In this study, we extended
area of high landslide hazard. This region has attracted the application of logistic regression.
a large amount of attention and research because of the The study area lies between the latitudes 30.11◦ N
Three Gorges Dam and Reservoir Project and its major and 30.37◦ N and longitudes 107.86◦ E and 108.15◦ E,
potential impact on the environment, geo-hazards, and and covers an area of 260.93 km2 . Geological for-
socioeconomy (Zhou et al. 1996, 1997). More than mations of it are primarily Quaternary deposits and a
2500 known localities of slope instability were iden- Jurassic system. Seven geologic formations are reco-
tified in this area (Liu et al. 2004, Bai et al. 2005a, gnized based on the similarity of their engineering
2005b, Bai et al. 2007), and the construction of the dam properties (Du et al. 1991).
is expected to cause a dramatic increases in landslide
hazard in this area.
Applications of bivariate statistical landslide sus- 2 SPATIAL DATABASE
ceptibility assessment methods have been widly
reported (Kienholz, 1978, Malgot & Mahr, 1979, Landslide and landslide causative factor databases
Rupke et al. 1988; Gupta & Joshi, 1990, Pachauri & were constructed. The size of pixels for all of the
Pant, 1992, van Westen, 1993, Soeters & van Westen, data layers is 25 × 25 m2 . The first dataset is the
1996, Hou et al. 2006, Bai et al. 2007). In the geological map. Geological paper maps at 1:10,000-
application of logical regression model for landslide scale were digitized and the geologic formations were
susceptibility mapping, some researchers have tried identified. The second dataset is topographical param-
to exploit the data via using dummy binary variables eters that were collected from 1:50,000-scale paper

1067
topographic maps. Elevation, slope angle, aspect, and statistical values, into the AHP matrixes. For all the
shape of slope parameters were obtained from the models, where AHP was used, the CR (Consistency
DEM using Surfer 8.0 (Wang et al. 2004, Bai et al. Ratio) was calculated (Saaty, 1977). The models with
2005). The third dataset is land cover, which was inter- the CR greater than 0.1 were automatically rejected.
preted from thematic mapping (TM) 5 (Path 127/Row With the AHP method, the values of spatial factors
39, dated 2000) satellite imagery using various image weights were defined. Using a weighted linear sum
processing and enhancement techniques. Eleven veg- procedure (Voogd, 1983) the acquired weights were
etation types were recognized. The fourth dataset used in the calculation for the landslide susceptibility
includes distances from drainage line, roads, and models.
settlements which were digitized from the 1:50,000 The resultant susceptibility map show five suscep-
topographic maps. tibility classes (Figure 1) based on standard deviations
A detailed landslide inventory map of the study of the corresponding histograms.
area was made from interpretation of 1:20,000-scale
color aerial photographs. Extensive field studies were
used to check the size and the shape of landslides; 3.2 Logistic regression approaches
and to identify the type of movements and the materi- In this study, all training data were input to the logis-
als involved. A total of 142 landslides were identified tic regression algorithm within statistical product and
within the study area. The total area of all the landslide service solutions (SPSS) and IDRISI Kilimanjaro to
scars is 5.32 km2 , accounting for 2.04% of the study calculate the correlation of landslide. After the forward
area. Characteristics of the landslides were recorded stepwise logistic regression analysis, variables such
on a standard landslide inventory data sheet, but the as Elevation, distance to drainage line, slope angle,
main purpose herein was to map the boundaries of the aspect, geological formation, soil and vegetation cover
landslides. Because the best undisturbed morpholo- were selected for being statistically significant. The
gical conditions (conditions before a landslide occurs) pseudo R 2 in this study is closer to 0.2 and shows a rel-
would be extracted from the vicinity of the landslide
atively good fit (Clark & Hosking, 1986). The receiver
polygon itself. Suzen and Doyuran (2004) presented
operating characteristic (ROC) value of IDRISI ranges
a new approach in the generation of decision rules
from 0.5 to 1, where 1 indicates a perfect fit and
of landslide occurrence, called Seed Cells. The Seed
0.5 represents a random fit. A value of 0.8701 is
Cells are achieved by adding a buffer zone to the crown
obtained in this study, which can be taken as an indica-
and flanks of the landslide. In this study, 6793 seed
tion of good correlation between the independent and
cells were derived from landslide polygon boundaries.
dependent variables.

3 LANDSLIDE SUSCEPTIBILITY MAPPING

3.1 Analytical hierarchy process approaches

The analytical hierarchy process (AHP) developed by
Saaty (1980) provides a flexible and easily compre-
hensible method of analyzing complicated problems.
It is a multiple criteria decision-making technique that
allows subjective as well as objective factors to be con-
sidered in the decision-making process. AHP allows
the active participation of decision-makers in reach-
ing agreement and gives managers a rational basis on
which to make decisions.For landslide susceptibility
has been shown before (Barredo et al. 2000, Mwasi,
2001, Nie et al. 2001) and it was used to define the
governing factors on landslide occurrence in a trans-
parent manner and to derive their weights. One set
of the models was developed using the values from
statistical to manually define the relationships between
the different factors according to the AHP methodol-
ogy. These values were later imported into the AHP
matrixes. The other set of the models was developed
by automatically importing the calculated relation- Figure 1. The categorized landslide susceptibility map of
ship values between different factors, based on the AHP approach.

1068
 5 DISCUSSIONS AND CONCLUSIONS

The AHP approaches to classify continuous data sets

in order to overcome some of the disadvantages of
previous methods. Logistic regression which exploits
the data via using dummy binary variables for each
class of an independent parameter has increased the
complexity of the data structure and limits the flexi-
bility of the statistical system (Carrara et al. 1991,
Chung, 1995). If there are many parameters, it
will produce a long regression equation and may
even create numerical problems. The new approach
uses the continuous data as it is, in order not to
alter the state and information present in the para-
meter maps.
Susceptibility maps produced using these two meth-
ods were validated with the 6793 landslide seed cells
data. The logistic regression was found to the bet-
ter in reflecting the actual landslide density based on
historical landslide than that desired using the AHP
method.

Figure 2. The categorized landslide susceptibility map of

AHP approach.
REFERENCES

Bai, S.B., Wang, J., Lu, G.N., Zhou, P.G., Hou, S.S. &
Finally, logistic regression values of the factors Xu, S.N. In press. GIS-based and analytical hierarchy pro-
were calculated. The resultant susceptibility map show cess landslide susceptibility mapping in the Three Gorges
five susceptibility classes (Figure 2) based on standard area, China. Pedosphere.
deviations of the corresponding histograms (Ayalew & Bai, S.B., Lu, G.N. & Sheng, Y.H., 2005. Analysis of land-
Yamagishi, 2005). slide causative factors using GIS in the three Gorges
Reservoir area, China (in Chinese). Journal of Mountain
Science. 23(1): 63–70.
4 VERIFICATIONS AND COMPARISONS Bai, S.B., Lu, G.N., Sheng, Y.H. & Yang, Y.P. 2005. Appli-
cation of GIS Technology During the Primary Study on
Landslide Hazards in the three Gorges Reservoir area,
Verification is performed by comparison between the China. Resource and Environment in the Yangtze Basin
two susceptibility maps and the 6793 landslide seed (in Chinese).14(3): 386–391.
cells data. The landslides area covered 2.04% of Barredo, J.I., Benavides, A., Hervas, J. & Van Westen, C.J.
the study area. The very high susceptibility zone is 2000. Comparing heuristic landslide hazard assessment
3.23% and 2.86%, respectively. They are approx- techniques using GIS in the Tirajana basin, Gran Canaria
imately equal. The low and very low susceptibil- Island, Spain. International Journal of Applied Earth
ity zones constitute 49.2% and 53.73% of the area Observation and Geoinformation, 2, 9–23.
respectively with corresponding 5.68% and 6.15% of Carrara, A., Cardinalli, M., Detti, R., Guzzetti, F.,
Pasqui, V. & Reichenbach P. 1991. GIS techniques and
the area overlapped by the landslide seed cells. The statistical models in evaluating landslide hazard. Earth
high and very high susceptibility zones constitute surface processes and landforms. 16(5): 427–445.
21.85% and 19.07% respectively of the area with cor- Chung, 1995. Application of Geographic Information Sys-
responding 63.24% and 69.41% of the area overlapped tems to landslide hazard zonation, ITC Publ 15, Enschede,
by the landslide seed cells. pp 245.
The outcome was that the landslide seed cells zones Clark, W.A. & Hosking, P.L. 1986. Statistical Methods
do not completely fit into the very high susceptibil- for Geographers. John Wiley and Sons, New York.
ity class of both maps. But 69.41% of these landslide pp 518.
zones fall into the high and very high susceptibil- Dai, F.C. & Lee, C.F. 2002. Landslide characteristics and
slope instability modeling using GIS Lantau Island, Hong
ity zones of the logistic regression map while this is Kong. Geomorphology, 42, 213–238.
63.24% in the case of AHP method. This indicates that Du, R.H., Liu, X.M. & Yuan, J.M. 1991. Study on Land-
the logistic regression map was better in reflecting the slides and Debris-flows in the Three Gorges. Chengdou:
actual landslide density based on historical landslide Sichuan Science & Technology Publishing House (in
than the AHP method. Chinese), 207 pp.

1069
Gupta, R.P. & Joshi, B.C. 1990. Landslides hazard zoning Ohlmacher, C.G. & Davis, C.J. 2003. Using multiple regres-
using the GIS approach—A case study from the Ram- sion and GIS technology to predict landslide hazard
ganaga catchment, Himalayas. Engineering Geology. 28: in northeast Kansas, USA. Engineering Geology. 69,
119–131. 331–343.
Guzzetti, F., Carrara, A., Cardinali, M. & Reichenbach, P. Pachauri, A.K. & Pant, M. 1992. Landslide hazard mapping
1999. Landslide hazard evaluation: a review of current based on geological attributes. Engineering Geology. 32:
techniques and their application in a multi-scale study, 81–100.
central Italy. Geomorphology. 31: 181–216. Rupke, J., Cammeraat, E., Seijmonsbergen, A.C. &
Hou, S.S., Li, A., Zhou, P.G. & Wang, R. 2006. Regional van Westen, C.J. 1988. Engineering geomorphology
landslide susceptibility assessment using bivariate statis- of the Widentobel catchment, Apenzell and Sankt
tic method—a case study of Yucheng District, Yaan Gallen, Switzerland. A geomorphological inventory sys-
City, Sichuan Province (in Chinese). Hydrogeology and tem applied to geotechnical appraisal of slope stability.
Engineering Geology. 33(1):1–4. Engineering Geology. 26: 33–68.
Kienholz, H. 1978. Maps of geomorphology and natural haz- Saaty, T.L. 1977. A scaling method for priorities in hierarchi-
ards of Grindelwald, Switzerland, scale 1:10000. Arctic cal structures. Journal of Mathematical Psychology, 15,
and Alpine Research.10: 169–184. 234–281.
Lee & Kyungduck Min, 2001. Statistical analysis of landslide Saaty, T.L. 1980. The Analytical Hierarchy Process. McGraw
susceptibility at Yongin, Korea. Environmental Geology Hill, New York.
40, 1095–1113. Soeters, R. & van Westen, C.J. 1996. Slope instability
Liu, J.G., Mason, P.J. & Clerici. 2004. Landslide hazard recognition analysis and zonation. In: Turner, K.T.,
assessment in the Three Gorges area of the Yangtze river Schuster, R.L. (eds.) Landslides: Investigation and Miti-
using ASTER imagery: Zigui-Badong. Geomorphology. gation. Transportation Research Board National Research
61: 171–187. Council, Special Report, Washington, DC. 247: 129–177.
Lulseged Ayalew & Hiromitsu Yamagishi, 2005. The applica- Suzen, M.L. & Doyuran, V. 2004. Data driven bivariate land-
tion of GIS-based logistic regression for landslide suscep- slide susceptibility assessment using geographical infor-
tibility mapping in the Kakuda-Yahiko Mountains, Central mation systems: a method and application to Asarsuyu
Japan. Geomorphology. 15–31. catchment, Turkey. Engineering Geology. 71: 303–321.
Malgot, J. & Mahr, T. 1979. Engineering geological map- Van Westen, C.J. 1993. Application of Geographic Informa-
ping of the west Carpathian landslide areas. Bulletin of tion System to landslide hazard zonation: a review of
the International Association of Engineering Geology. 19: principles and practice. ITC Publication No. 15, ITC,
116–121. Enschede, The Netherlands. pp. 245–260.
Mwasi, B. 2001. Land use conflicts resolution in a frag- Voogd, H., 1983. Multicriteria Evaluation for Urban and
ile ecosystem using Multi-Criteria Evaluation (MCE) Regional Planning. Pion Ltd., London.
and a GIS-based Decision Support System (DSS). Int. Wang, J., Bai, S.B. & Chen, Y. 2004. Surfer 8 geography
Conf. on Spatial Information for Sustainable Develop- information mapping, The sinomaps press, p176–181.
ment, Nairobi, Kenya, FIG—International Federation of Zhou, P.G. 1996. Environmental engineering geological pre-
Surveyors. 11 pp. diction of the new urban site area of badong city (in
Nie, H.F., Diao, S.J., Liu, J.X. & Huang, H. 2001. The Chinese). Journal of Engineering Geology. 4(4): 86–92.
application of remote sensing technique and AHP-fuzzy Zhou, P.G. & Qu, Z.D. 1997. Stability estimation of the
method in comprehensive analysis and assessment for different types of gentle slope in the Reservoir Area of
regional stability of Chongqing City, China. Proc. of the Three Gorge Project (in Chinese). The Chinese Journal of
22nd Asian Conf. On Remote Sensing, Vol. 1, Centre Geological Hazard and Control. 8(2): 24–32.
for Remote Imaging, Sensing and Processing (CRISP),
National University of Singapore, Singapore Institute of
Sorveyors and Valuers (SISV) and Asian Association on
Remote Sensing (AARS), Singapore, pp. 660–665.

1070
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Importance of study of creep sliding mechanism to prevention

and treatment of reservoir landslide

Junguang Bai, Shengdi Lu & Jianshe Han

Northwest Hydro Consulting Engineers, CHECC, Xi’an, China

ABSTRACT: Creep sliding mechanism of the reservoir landslide and creepage property of the sliding-zone
soil are theoretically studied from the viewpoint of dynamics. The results show that the increase and decrease of
the shearing stress ratio for the sliding plane of a reservoir landslide and the viscosity coefficient of the soil in
the sliding zone are the essential conditions which can define whether the creep sliding of the landslide will turn
to a violent sliding. It has proved that the slide acceleration test by simulating the operational conditions of the
reservoir with a representative sliding block under appropriate environmental conditions is a reliable approach
to verify the creep sliding mechanism of reservoir landslide and an effective method to treat the landslide and
eliminate in advance the danger of landslide surge.

1 GENERAL of the landslide during the later impoundment, at the

same time it proves that the slide acceleration is an
Most of the large-sized hydropower and water conser- effective method to treat the landslide and eliminate in
vancy projects are constructed in mountain valleys. advance the danger of landslide surge.
Deformed or destabilized rock mass or landslide
widely distributed at natural bank slopes in the valley
area, which is caused by the combined actions of var- 2 EXPERIMENTAL RESEARCH
ious geological factors, is defined as one of the most ON LANDSLIDE CREEP SLIDING
severe engineering geological problems encountered CHARACTERISTICS
during the planning, construction and operation of
hydroelectric projects. As the most common geologic 2.1 Experimental research on creep
phenomenon, reservoir landslide may lead to severe characteristics of the sliding-zone soil
damage, for instance, the Vajont reservoir landslide in
Italy occurred on October 9, 1963 killed 1925 people From the viewpoint of dynamics, the magnitude of
and completely destroyed the reservoir. Therefore, the the sliding speed of a sliding mass, with the excep-
stability prediction and study of reservoir landslide as tion of a small part of the slope bodies which behave
well as the prevention and treatment of the destruc- by themselves, is mainly governed by the physi-
tive effect shall be paid with special attention all the cal mechanics characteristics, especially the creep
time. In the case of the Lijiaxia hydropower station, characteristics of the sliding-zone soil. Therefore,
creep sliding mechanism of the reservoir landslide and experimental researches were carried out on the creep
creepage property of the soil in the sliding zone are characteristics of the sliding-zone soil of the reservoir
theoretically studied from the viewpoint of dynamics landslide before and after initial impounding.
and on the basis of systematic analyses of the mon- (1) Direct shear test on sliding-zone soil
itored landslide deformation data.. The results show Represented by the bedding strata sliding-zone soil
that the increase and decrease of the shearing stress of the No.2 landslide, the relation between shearing
ratio for the sliding plane of a reservoir landslide and force (τ) and shear displacement (ε) falls into plastic
the viscosity coefficient of the soil in the sliding zone curve type, and the content of fine grain and clay grain
are the essential conditions which can define whether of less than 2 mm is higher (accounting for over 90%).
the creep sliding of the landslide will turn to a violent The result of direct shear test is as follow:
sliding. On this basis, a representative sliding block
is selected to execute the slide acceleration test by
simulating the operational conditions of the reservoir Yield strength/peak strength (τy/τp) = 0.76∼0.85
and the creep sliding mechanism is proved again. The
test results successfully predict the deformation trends Residual strength/peak strength (τr/τp) ≈ 1

1071
 It shows that the shearing strength of the sliding- Critical strength τ0 = 0.325σ + 0.890
zone soil and the physical parameters (water content,
dry density) remain unchanged, during sliding of Peak strength τp = 0.320σ + 0.767
the slide mass, the shear strength parameters of the
sliding-zone soil have become stable basically, and Long-term strength τ∞ = 0.162σ + 0.0175
reached or closed up to the residual strength in most
cases; the change in micro-structure have little effect
When down sliding shear stress (τ) lies between
on the strength of the sliding-zone soil.(2)
long-term strength (f ∞) and critical strength (f0) of
sliding-zone soil, landslide would not develop to the
(2) Servo shear test on sliding-zone soil
accelerating creepage stage; meanwhile, coefficient
Servo shear test indicates that: under natural
of viscosity (η = τ/ε, of which ε = strain rate) will
environmental conditions, the water content of the
decrease with the increase of shear stress (τ); vis-
sliding-zone soil is basically in saturation state
cosity will increase with the increase of time; when
(ω = 15∼20%), and density varies in the range of
shear stress remains unchanged (or decreases slightly),
2.0∼2.25 g/cm3 in most cases; shearing speed has a
regeneration phenomenon may happen to coefficient
very small impact on the strength of the sliding-zone
of viscosity.
soil when it is larger than 16 mm/h, and both values
Compared with the change amplitude of the land-
of f and c tend to be stable; when shearing speed is
slide coefficient of viscosity (η) of Gepatsch reservoir
0.6∼16 mm/h, friction coefficient K increases at first
in Austria (102 ∼106 MPa·S) (4) , the change amplitude
and then decreases, however c value decreases at first
of the landslide coefficient of viscosity (η) of Lijiaxia
and then increases, with less change of comprehensive
(105 ∼106 MPa · S) is smaller to some extent. The rea-
shear strength.
son is that the normal stress in the Lijiaxia rheology
The test result is basically consistent with that of
test (0.1–0.4 MPa) is on the low side, and the shearing
direct shear test, namely when the water content and
time is shorter.
dry density of the sliding-zone soil remain unchanged,
During impounding, the slide mass has a loose
shear strength parameters of the sliding-zone soil will
structure, and coefficient of viscosity η decreases.
not significantly decrease during the sliding of the
After stabilization of water level, τ can be considered
slide mass, and no governing factors of intensive slid-
unchanged on the whole; as time lasts, η will increase,
ing would be formed.
and the rate will decrease; due to limited increasing
of η, creep sliding would not stop, maintaining at a
(3) Shear rheology test on sliding-zone soil
low rate stable creep status.
In testing, a water content of 15.21% is simu-
lated by use of ‘‘Chen’s loading method’’(3) ; a den-
sity of 2.23 g/cm3 , a normal stress of Grade 4
2.2 Landslide creep sliding mechanism analysis
(0.1∼0.4 MPa), and an increment of shear load of
1/8∼1/11 P (the peak strength in the case of different The tests indicate that when the shear stress acting
grades of normal stress) are adopted; loads with vari- on the sliding-zone soil lies between the long-term
ous grades last 5 days. The result is listed in Table 1. strength and critical strength, shear creepage of the
Relation expressions of shear stress and normal sliding-zone soil is actually deceleration creepage. In
stress are obtained from analysis of the rheology case this deceleration creepage has a long duration and
test result of the sliding-zone soil of No.1 and No.2 the displacement quantity of sliding mass is less than
landslides: the accuracy of observation, it is deemed to turn into

Table 1. Peak strength and long-term strength for a sliding-zone soil test of Lijiaxia.

Peak Long-term Critical

Sliding-zone strength strength strength Cp C∞ C0
soil fp f∞ f0 (MPa) (MPa) (MPa)

Mud
sandwiched
with debris 0.320 0.162 0.325 0.023 0.020 0.025
Debris
sandwiched
with mud 0.415 0.325 0.418 0.040 0.036 0.042

1072
steady creep. Because the actions of external factors landslide can be triggered by increasing the static
such as reservoir level change, rainfall as well as earth- hydraulic pressure in the landslide body, it will be
quake, etc., will finally be reflected by the action of the most cost-effective method.
shear stress, each change of shear stress may result in a 2. Landslide body under the most unfavorable envi-
new adjustment of creepage state of sliding-zone soil, ronment: During initial reservoir impounding,
namely a new deceleration creep sliding would occur. reservoir level rose about 90 m, and apart from the
This phenomenon has been verified by many times front superficial zone of the landslide that had large
of water level rising and sudden lowering. When this displacement (generally, no more than 10 m), the
deceleration creep sliding movement is simulated by middle and deep zones of the slide mass commonly
use of logarithmic curve, the monitored displacement got dislocation of several tens of centimeters along
procedure can be perfectly fitted; when its physical the rear edge cracks and cracks of the sub-sliding-
characteristics is illustrated by coefficient of viscos- body or creep sliding rate increased, however,
ity, the conclusion is that it has a similar ‘‘viscosity it returned to decelerated creepage or constant
coefficient regeneration’’ as that of the sliding-zone creepage when the reservoir level remained at EL.
soil test result for the Gepatsch reservoir, i.e. ‘‘coeffi- 2145 m, which indicates that large displacement
cient of viscosity decreases with the increase of shear for the integral landslide will not necessarily occur
stress; viscosity increases with the increase of time’’. during landslide acceleration test. The most unfa-
A lot of case histories indicate that the reason vorable environment for the slide mass is such that
of state shifting of landslide from creep sliding to the reservoir level abruptly drops to EL. 2,145 m
intensive creep sliding is because the shear stress on from the normal water level, and the corresponding
slide planes suddenly exceeds the critical strength, the case of the landslide acceleration test is such that
positional adjustment caused by landslide displace- the underground water in the sliding mass reaches
ment failed to duly fit the situation of shear load to the normal pool level of EL. 2,180 m or even
increase(5) . This may contribute to the sudden decreas- higher.
ing of strength of slide planes and sudden increase 3. Verification of down-sliding mechanism of land-
of main sliding load. Moreover, it is also related to slide: The stability of landslide of the reservoir
geometrical factors such as the shape of slide plane. in front of the dam has gone through several
The reservoir landslide in front of Lijiaxia Dam is an abrupt environmental changes, e.g., the extreme
ancient landside always in movement, its slide plane is high flood (rainfall) on the Yellow River in 1981,
steep in the upper portion and flat in the lower portion, the Yellow River break off during impounding of
and the sliding-zone soil is of residual strength. Sliding the Longyangxia reservoir in 1986, the river clo-
mass has a loose structure and strong water perme- sure by the Lijiaxia cofferdam in 1991 (water level
ability, the change of reservoir level can be responded increased by 8∼10 m, and the reservoir impound-
by slide mass movement at any time, which makes it ing in 1997 (reservoir level increased by 90 m).
impossible all the time that the strength on slide plane The general rule is that the stability of the slide
exceeds the critical strength. Based on this analysis, mass is closely associated with water action; every
it is impossible that the reservoir landslide in front of environmental change triggers increase of creep
Lijiaxia Dam will come into the stage of acceleration sliding rate of the slide mass and gradual extension
creep, i.e. intensive creep sliding is not likely to occur of surface tensile crack, but gradually returned to
to result in surging hazard. constant creep sliding with small magnitude after
accelerated creep sliding of short duration.(6) If the
landslide acceleration test had the slide mass under-
gone these changes again, it will be of realistic
3 PRODUCTIVE LANDSLIDE
engineering significance to the stability prediction
ACCELERATION TEST TO SIMULATE
and engineering remedial work during subsequent
RESERVOIR OPERATION ENVIRONMENT(1)
reservoir level rising.
3.1 Test Assumption
3.2 In-situ landslide acceleration test
1. Stability improvement alternative by excavation:
Static stability analysis shows that the most unsta- The test zone is located on the No. 5 longitudinal pro-
ble water level on the calculated profile of land- file of the downstream No.2 landslide, the landslide
slide during reservoir impounding is at elevations acceleration plane is 200 m in length and 200∼250 m
between 2,130∼2,145 m, and the stability of land- in width, the sliding plane is the secondary sliding
slide will increase when water level reaches EL. plane on No.2 landslide corresponding to No.3 crack
2,145 m, however, the required design safety on the surface (refer to Figure 1), with a total volume
allowance still cannot be met, and necessary exca- of 1.53 × 106 m3 , of which, the part above water level
vation work shall be done. If the sliding of a 2,145 m is about 780 × 103 m3 .

1073
 Figure 2. Ground water Duration Curve in Test Zone.

Figure 1. Schematic profile of productive environment

simulation test zone.

Test pits, test boreholes and ground water sprin-

kling irrigation are jointly used for the test. A total of
three rows of water pits (32 pits) spacing at 5 m and
a total of 91 boreholes are provided. Bottom eleva-
tions of the boreholes from the first row to the third
row are EL. 2,145 m and 2,170 m respectively; 10
underground water observation holes and 21 surface
displacement monuments are provided in the test zone;
1 monitoring adit (PD No.70) consisting of 3 sub-
adits is provided parallel to the sliding plane of the
test zone at the downstream side EL. 2,163 m, all 3
sub-adits run through the sliding plane and the sec-
ondary sliding plane in the test zone. The test and
monitoring system provides complete monitoring with
respect to hydraulic pressure, discharge, underground Figure 3. Isoline of underground water level in test zone
water, creep behavior, etc., and some samples are taken (1998.09.30).
from the sub-adits to check the physical and mechani-
cal parameters of the soil in the sliding zone for further
study and verification of the creep sliding property of higher than 800 m3 /h, and the isoline of groundwater
the slide mass and the variation behavior of the strength in the test zone is shown in Figure 3. The behavior of
of the soil in the sliding zone. Landslide acceleration groundwater dynamic is as follows:
test was carried out in afternoon on 9th June, 1998 ① The groundwater in the test zone has an obvious
and ended on 30th September, 1998, water was sup- hump curve at the third row of the boreholes, form-
plied during consecutive 114 days. Average daily water ing groundwater dividing line, and the water level is
injected from 9th June to 16th September is about largely above El. 2,180 m and the overall groundwa-
5.5 × 103 m3 /d (200 m3 /h), and the water supplied ter level in the test zone is not less than 2,175 m. It
from 17th September to 30th September increased to can be seen from Figure 6 that the groundwater seeps
20 × 103 m3 /d (about 800 m3 /h). toward the outer side of the reservoir bank and ground-
water level gradually drops to about EL. 2,150 m,
and the average hydraulic gradient is 40∼50%; for
3.3 Analysis of landslide acceleration test
groundwater seeps toward No. 3 crack at the inner
(1) Groundwater behavior side of the reservoir bank, the average hydraulic gra-
The groundwater in the test zone before water was dient is 60∼80%; and the groundwater, controlled by
injected is almost as same as reservoir water level, non-linear property of the sliding mass, distributes
and it began to rise 3 days after water was injected irregularly.
(duration curves of typical groundwater observation ② Figures 2 and 3 show that although the under-
holes during water filling refers to Figure 2). The ground water level in the test zone didn’t fully reach
discharge of the water injected at early stage of the the expected EL. 2,180 m, the effect of saturation of
test is about 200 m3 /h, and that during later stage is slide mass and underground water seepage behavior in

1074
the test zone on the stability of reservoir banks at least Some fresh radial tension cracks occurred along No. 3
equals to the case that the reservoir level lows to EL. crack and the outside platform, further reflecting the
2145 m from EL. 2,175 m at a very short duration, characteristics of the behavior of the slide mass during
almost the most unfavorable operation environment landslide, such as pervasive crack, loose disintegration
for the reservoir in the near future. and displacement and concurrently reflecting the creep
③ Since water is injected through pits and boreholes sliding rate of the slide mass undergone acceleration
and sprinkling irrigation, for the superficial zone of the and deceleration at a second time when underground
slide mass, the intensity of water injection is far higher water level rises with increase of quantity of the
than the infiltration intensity of future heavy rainfall. injected water, which coincides with the displacement
Calculated based on the injected water and the applied during initial impounding.
area, the water injection intensity at early stage equals
to the rainfall (11.6 mm/d) within 100 consecutive (3) Stability calculation check
days; and the intensity at later stage equals to rain- The result of the sample test carried out in the adit
fall (42.8 mm/d) within 14 consecutive days. Hence it (PD70) for landslide acceleration test shows that the
is deemed that saturation and instant pore water pres- strength parameter of the soil in the sliding zone was
sure of the slide mass is far higher than the capacity of reduced at a certain extent during the test, basically
long term rainfall. reflecting the real strength of the soil in the sliding
zone during subsequent reservoir impounding. Since
(2) Displacement behavior the strength value of the soil in the sliding zone is
Figure 4 shows that the displacement rate mea- almost the same as the value verified before impound-
sured by instruments during landslide acceleration test ing, the verified parameters used in the test is adopted
apparently increased. The displacement behavior is as in the stability calculation check for the slide mass for
follows: easy comparison. The result of calculation check is
① When small quantity of water injected at early shown as follows:
stage of the test, the creep sliding rate is only 2∼3 The landslide acceleration test environment has
times the initial value; when water quantity was lowered the safety factor of the testing slide mass by
increased, creep sliding rate may increase by 10 times 3.6∼11%; the safety factor of the secondary sliding
approximately, and the value is about 1/2∼1/3 of mass with No. 2 crack as its bottom plane reduced
the maximum rate measured during initial reservoir by 3.5∼4.86%, its maximum sliding rate was approx-
impounding. imately equal to 1/2∼1/3 of the value during initial
② It is found during the test that some localized dis- impounding and didn’t result in abrupt landslide. Cal-
integrated loose slide mass on the surface of the bank culation result shows that the stability factor of the
slope failed, and the displacement measured by the slide mass during further increase of reservoir water
instruments provided at the platform outside of No. level above EL. 2,145 m has a tendency of increase
3 crack in the test zone increased to 220∼270 mm, at a certain extent, which is in compliance with the
and that at the inside of No. 3 crack is about 150 mm. calculation results obtained before reservoir impound-
ing. Therefore, analyzed from perspective of stability,
surge wave will not occur in the future at other
locations having similar stability conditions as the
reservoir landslide acceleration test zone.

4 CONCLUSIONS

The landslide dynamic analysis should incorporate

mechanics theory of creep sliding mass. Study and
monitoring data of the Lijiaxia reservoir give evi-
dence that relaxation and displacement of landslide
adjust the stability state of sliding mass. Landslide
in this area experiencing creep sliding acceleration
and deceleration before and after reservoir impound-
ing proves that the viscosity of the sliding zone soil
increases during creep, which restrains the shear stress
acting on the sliding plane within critical state so
that accelerating creep and abrupt sliding will not
Figure 4. Duration Curve of Displacement Measured at occur and accordingly, surge wave hazard will not be
Typical Monitoring Points in Test Zone. incurred.

1075
 At an appropriate environment, a typical sliding Liu Xiong, 1994. An Introduction to Rock Rheology, Geo-
mass can be selected for landslide acceleration test logical Publishing House.
under simulated reservoir operation conditions, which Henderlun, A.J. & Patton, Jr.F. 1988. Geotechnical Analy-
is a reliable solution for verification of landslide creep sis of New Geological Observation Data on the Failure
sliding mechanism and also one of the effective ways Plane of Vajont Landslide (Final Report, June, 1985),
reprinted in the ‘‘Translation Collection of Reservoir
to improve the stability of landslide and eliminate Landslide Articles’’ (for internal exchange and reference
landslide surge wave risk. only), Northwest Hydropower Investigation, Design and
Research Institute.
Wang Lansheng, et al., 1988. Preliminary Study on Develop-
REFERENCES ment Characteristics and Mechanism of Initiation, Sliding
and Braking of Xintan Landslide, Typical Landslides in
Northwest Hydro Consulting Engineers, CHECC, Stability China, Science Press.
Forecast and Study Report for No.1 and No.2 Land- Jin Delian & Wang Gengfu, 1988. Tangyanguang Landslide
slides of Lijiaxia Hydropower Station after Preliminary of Zhexi Reservoir, Typical Landslides in China, Science
Impoundment, December, 1999. Press.
Xiao Shufang & K. Akilov, 1991. Structural Fabric and
Strength Creep Characteristics of Mudded Intercalation,
Jilin Science & Technology Press.

1076
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Stability prediction of landsides before and after impoundment

for Lijiaxia hydropower station

Junguang Bai, Shengdi Lu, Jianshe Han

Northwest Hydro Consulting Engineers, CHECC, Xi’an, China

ABSTRACT: On the basis of the landslide deformation monitoring data in decades of years and taking the
actual creep sliding trends of landslide during the initial impoundment as a model, the stability of reservoir
landslide before and after impoundment is predicted and evaluated for the Lijiaxia hydropower project on the
Yellow River in accordance with the systematic analyses of the deformation monitoring data obtained before
and after the initial impoundment and the test results of sliding-zone soil. The predicted results are verified in
the process of initial impoundment, site sliding acceleration test and subsequent reservoir level rising. At the
same time it is concluded that the investigation and research of large landslide of reservoir bank near the dam
is rather important, while the calculated surge height of landslide by calculation of landslide stability and slide
speed with conventional methods may be quite different from the actual height. Therefore, it is necessary to
study the creepage characteristics of sliding-zone soil in ancient landslide and the creep sliding mechanism of
landslide from the viewpoint of dynamics.

1 GENERAL and accelerated sliding were occurred from outside to

inside of both landslides. Displacement of the land-
With a total installed capacity of 5×400 MW, Lijiaxia slides was mostly within the range of 0.5∼3.5 m;
hydropower station on the Yellow River is the third however, severe sliding and obvious surge did not hap-
largest cascade hydropower station early planned on pen. The reservoir level was kept at the level of 2145 m
the Upper Yellow River. It has a double curvature arch for 21 months, during this period both landslides slid
dam with a maximum height of 155 m, a crest eleva- continuously and slowly at a speed rate equivalent to
tion of 2185 m, a normal storage level of 2180 m and a several times to dozens of times of the speed rate
total reservoir storage of 1.65 × 109 m3 . The construc- before the impoundment. Therefore, it is necessary
tion of this project was started in April 1988, reservoir to predict and study the stability of the landslides
impoundment was begun on December 26, 1996, the and the deformation trend in the course of continu-
initial operating level of 2145 m was reached on ous rising of water level so as to prevent the danger of
January 26, 1997 and the first unit was commissioned surge, eliminate the limitation on storage level and
in February of the same year. With the limitation of give play to the benefits of the project as soon as
construction progress of dam and foundation treatment possible.
at that time, especially controlled by the stability of On the basis of the systematic analyses of creep
upstream landslides No.1 and No.2, a 15 m-high free- sliding data of the two landslides measured before
board is provided to prevent surge danger induced by and after initial impoundment as well as the creep-
landslide in reservoir. age characteristics of sliding-zone soil, the stability
Landslide No.1 on the right bank is 90∼630 m of primary and secondary sliding planes of the two
upstream of the dam, with a volume of 6.92 × 106 m3 , landslides under different water levels are checked and
and Landslide No.2 on the left bank is 830∼2090 m computed, and the variation regulations of the safety
upstream of the dam, with a volume of 18.45×106 m3 , factor and storage level are analyzed and predicted. It
(see Figure 1). Before impoundment, both landslides is found that when the level is between 2130∼2145 m
were in a critically stable state of uniform creep slid- the safety factor is minimal, i.e. the extent of creep
ing. The speed of creep sliding was increased and sliding is maximal; but when the level is greater than
decreased due to the variation of rainfall and river 2145 m, the safety factor tends to slightly increase
level. Water level was raised for about 90 m from along with the rising of storage level. This conclusion
the initial river level to the initial operating level of is consistent with the prediction and analysis results
2145 m. Within this period, cracking, disintegration obtained before impoundment. Only the study results

1077
 The longitudinal landform of the Landslide No.2 is
gentle at top (30◦ ∼40◦ ) and steep at base (40◦ ∼50◦ )
with obvious slope step. While in lateral direction,
the landslide is divided into the upstream zone and
the downstream zone by the cone-shaped massive
bedrock at the middle of front edge (refer to Fig.1).
During normal impoundment, 91% of the upstream
zone is submerged and is far away from the dam
with less impact on the project; 66% of the down-
stream zone is located underwater with a 45 m-high
semicircular fractured wall at the rear edge and a
slope gradient of 60∼75◦ . The downstream zone
is large in volume and close to the dam, so it is
Figure 1. Plan of landslides upstream of the dam of Lijiaxia predicted that the project would be endangered if
Hydropower Station. destabilization occurs. Therefore, deformation mon-
itoring and study as well as disaster prevention and
control shall be carried out mainly for the downstream
zone.

of the Landslide No.2 are introduced in this paper due 2.2 Deformation characteristic of landslide
to paper length limitation.
Signs of renewal of the landslides were already found
in the early geological investigation in 1978. To further
understand and study the deformation and movement
2 DEFORMATION CHARACTERISTIC
properties of the landslides, monitoring works com-
AND STABILITY PREDICTION
prising geological visit and surface and underground
OF LANDSLIDE BEFORE IMPOUNDMENT
survey were started in the second half of 1983.
The observed results show that the deformation and
2.1 Development characteristics of landslide
displacement of the landslides are characterized by
The Landslide No.2 is a large, deep-seated landslide creepage. The rate of ground displacement in the direc-
influenced by buckling failure. Its total volume is tion of primary sliding plane is maximum at the top
18.45 × 106 m3 , in which the upstream zone (II- 1) and minimum at the middle, having the property of
is 4.40 × 106 m3 and the downstream zone (II-2) is combined traction and draught. The reason of this phe-
14.05×106 m3 . Developed in the stratified chorismite nomenon is that the landform is comparatively steep
of the primary Sinian period, the landslide belongs to a at the top and base and gentle at the middle. The dis-
consequent one with rock strata dipping to the strike of placement rate of ground is much greater than that
slope. The upper and medium rock strata are in a state of the primary sliding plane, about 2.5∼22 times the
of consequent dipping with dip angle greater than the displacement rate of the primary sliding plane. The sta-
slope gradient of landform; the lower rock strata are bility of the secondary sliding plane is lower than that
in a state of reversed dipping characterized by typical of the primary sliding plane and the displacement rate
bedding sliding∼bending (buckling). is slightly bigger, the deformation of sliding mass itself

Table 1. Comparison of deformation rate of Landslide No.2 before impoundment.

Deformation rate in construction (mm/day)

(1988.3∼1996.12)

Water retaining
Deformation rate by cofferdam and After top
before construction Water retaining top unloading of unloading of
(mm/day) by cofferdam Landslide No.2 Landslide No.2
Zone 1984.11∼1991.10 1991.10∼1991.11 1991.12∼1993.7 1993.8∼1996.5

Upstream
zone (II-1) 0.02∼0.09 0.16∼0.42 0.11∼0.27 0.04∼0.16
Downstream
zone (II-2) 0.07∼0.19 0.15∼0.59 0.12∼0.47 0.06∼0.15

1078
accounts for a considerable proportion of the ground that Landslides No.1 and No.2 will destabilize and
displacement. Under the complex influence of ground slide downwards in the course of impoundment.
landform, various rock types and structures and dif- Furthermore, it is predicted that the possible maximum
ferent stability of the primary and secondary sliding volume of a single sliding is 3.0 × 106 m3 (Landslide
planes, the sliding mass tends to disintegrate in the No.2) and 4.0 × 106 m3 (Landslide No.1), the mini-
process of creep sliding and deformation. The results mum water level of destabilization is 2070∼2080 m,
of displacement monitoring in different stages before the possible maximum sliding speed is 9∼10 m/s and
impoundment are shown in Table 1. the height of surge in front of the dam is respectively
15 m and 10 m for the No.1 and No.2 landslides under
the level of 2145 m. The surge height is controlled by
3 PREDICTION AND ANALYSIS Landside No.1.
OF LANDSLIDE STABILITY BEFORE At the same time it is concluded that the stability of
INITIAL IMPOUNDMENT the landslides is sensitive to the action of water. Dur-
ing impoundment the sliding mass slides downwards
3.1 Stability analysis of landslides along with the rising of water level and achieves a state
of balance; after that with the new rising of water level
Various parameter combinations have been applied to the sliding mass slides continuously. This process is
calculate the rigid body equilibrium and stability of the so slow that the probability for surge damage is quite
primary and secondary sliding masses. According to small. Considering the shorter distance between the
calculated results, except for No.3 longitudinal profile, landslides and the dam, the above predictions shall
the safety factors of other profiles are minimal under be further proved through practice. For the sake of
the water level of 2130∼2145 m, and increase slightly safety, before initial impoundment the calculated slid-
along with the rising of water level. The calculated ing speed and the test results of surge are adopted as
results are as shown in Table 2. the foundation for risk evaluation of surge, and corre-
sponding protective measures are taken. It is suggested
that the protective height of the dam during the initial
3.2 Prediction and study of surge induced impoundment shall be preserved as 15 m. Measures
by landslide for protection and treatment in operating period are to
Before impoundment, basic kinematic formula con- be defined after the deformation trend of the landslides
verting conventional potential energy and kinetic are observed and studied under the level of 2145 m for
energy and energy method (recommended by ASCE) a certain time.
and Panjiazheng method are applied to calculate slide
speed, and the result is 9∼12 m/s.
Model tests with model scale of 1:22 are carried out 4 CREEP CHARACTERISTICS OF LANDSLIDE
for the surge induced by landslide on the basis of slide DURING INITIAL IMPOUNDMENT
speed and surge calculation. In these tests, the height
of surge in front of the dam under combined condition Reservoir impoundment began at 11 o’clock on
of different storage levels, slide speeds and volumes 26 December 1996, and reservoir level rose from
are simulated. EL.2055.95 m to EL.2145 m by January 26, 1997.
In accordance with the creepage property of the Compared with the normal river level, the reservoir
landslide before impoundment and the stability cal- level rose by 89.05 m, the rising hydrograph is shown
culation results of different positions, it is predicted in Figure 2.

Table 2. Calculated results of stability of Landslide No.2 at different water levels.

Safety factors
Profile Sliding
No. plane 2058 m 2080 m 2110 m 2130 m 2145 m 2160 m 2180 m

No.3 longitudinal Primary

profile 1.06 1.00 0.93 0.90 0.89 0.88 0.88
No.4 longitudinal Primary 1.03 0.99 0.97 0.97 0.98 1.00 1.02
profile Secondary 1.03 1.00 0.98 0.97 0.98 1.00 1.03
No.5 longitudinal Primary 1.04 0.99 0.95 0.94 0.95 0.97 1.00
profile Secondary 1.01 0.97 0.94 1.02 0.94 0.97 1.00
No.7 longitudinal Primary 1.14 1.08 1.03 1.02 1.03 1.05 1.08
profile Secondary 1.12 1.08 1.04 1.02 1.03 1.07

1079
4.1 Displacement characteristics for ground millimeters. The maximum displacement speed for
measuring points measuring points was recorded at water level extent
of 2128∼2130 m, after which the displacement speed
From the next day of reservoir impounding, the ten-
gradually decreased with the rising of reservoir level.
sion cracks on sliding mass surface began to expand,
The maximum speed is decreased from 20∼60 mm/d
the speed of displacement increased rapidly at all mea-
to 0.5∼0.8 mm/d at the end of 1998, however, it is
suring points; however, the deformation magnitude
still several times to tens of times of the speed before
at different locations varies due to the difference in
impoundment.
structures of various sliding masses, and the general
feature is that the deformation at lower portion is larger
than that at upper portion, and tends to decrease from 4.2 Tention cracking and slide block
the reservoir bank inward. Within 30 days of initial movement features
impounding, all the displacements at measuring points
of the front edge zone are larger than 1000 mm; the dis- Over 40 tension cracks of different sizes and zonal
placements at measuring points of the mid-front area distribution occurred successively, with an obvious
are generally 600∼1300 mm, and horizontal displace- feature of disintegration. Of which, the No.1, No.2
ment is greater than vertical displacement; the rear and No.3 large cracks, resulted from expanding along
edge zone has smaller deformations with a magnitude the original primary and secondary sliding planes at
ranging from dozens of millimeters to over a hundred the mid and rear parts, basically cut through the down-
stream area, and stretch in parallel with the reservoir
bank, with a stable direction and a great width, gov-
erning mainly the area downstream of Landslide No.2.
The characteristic value of their cracking time with the
rising of water level is shown in Table 3.
In summary, the deformation of Landslide No.2
after impoundment is of a trend to block disintegration
in both longitudinal and transverse directions. Some
block masses of the slope body on the front edge sur-
face layer have slipped into the reservoir, with a lesser
volume and sliding distance of 3–7 m. From the bank
inward, cracks successively opened, with displacement
decreased, mainly presenting the creep along the pri-
mary and secondary slide planes. Its most important
change is that the original downstream No.1 crack cor-
responding to the primary slide plane has expanded
toward upstream area and almost cut through, with a
Figure 2. Time-water level relation curve at measuring newly increased length of about 160 m and cumulative
points of typical section. displacement over 1000 mm

Table 3. Crack deformation along the primary and secondary slide planes at the mid and rear parts of Landslide
No.2 and time characteristic value of reservoir level
Characteristic value of cracking from acceleration
Initial cracking
Sample measuring

to deceleration
Crack Number

(year/month/day)
(year/month/day)

with three days

turning points
speed (mm/h)

delayed (m/d)
turning point
increment at

increment at
Crack width
points

Water level
Occurrence

Water level
Delay time
Reservoir

Reservoir

Cracking
level (m)

level (m)

(mm)

(m/d)
time
Time

(h)

3 5-2 96.12.30 2088.20 84 97.01.05 2111.40 1.78 0.993 3.94 2.45~4.12

2 5-4 97.01.01 2097.47 128 97.01.13 2129.50 176 0.481 1.02 1.01~1.04
1 4-4 97.01.13 2102.41 179 97.01.16 2132.50 73 0.272 1.08 1.02~1.16
1. Delayed time refers to the time later than the beginning of reservoir level rising;
Notes 2. Simple measuring points refer to the points that are temporarily arranged on the two sides
of a crack during impounding, the relative change between two points is measured in practice.

1080
 According to the deformation trend of the land- at middle part varies from 600 mm∼1300 mm), and
slide from the initial impoundment to the subsequent only a small part of front edge superficial sliding mass
reservoir level raising, the reservoir level (when sliding slipped into the reservoir. Afterward, speed rate grad-
masses disintegrated), disintegration way, sequence ually decreased with increase of reservoir water level,
and size are almost as same as what was predicted the maximum speed ranging is decreased to about
before reservoir impounding. The surge wave height 0.5∼0.8 mm/d at the end of 1998 (several times to tens
of landslide before dam (15 m) determined based on of times of the values before reservoir impounding).
normal sliding rate calculation method and surge wave
model test has sufficient allowance. (2) Analyses of creep characteristics of sliding-zone
soil
A number of geotechnical tests carried out before
5 LANDSLIDE STABILITY PREDICTION and after reservoir impounding show that due to high
AND ENGINEERING TREATMENT water content of the soil in sliding zone under nat-
ural conditions, natural shear strength to saturation
Based on study of creep sliding mechanism of land- shear strength ratio is about 0.89∼0.95. Therefore, the
slide before impoundment and surge prediction as impoundment has small influence on the changing of
well as the dynamic observation for prototype of the shearing strength of sliding plane, i.e., the difference
landslide before and after the initial impoundment, between the peak strength and the residual strength
creep sliding mechanism of the reservoir landslide and is small and severe sliding of the landslides will not
creepage property of the soil in sliding zone are the- happen. At the same time, shear rheology tests on
oretically studied. The results show that the increase sliding-zone soil show that the coefficient of viscos-
and decrease of the shearing stress ratio for the slid- ity will decrease with the increment of shear stress
ing plane of a reservoir landslide and the viscosity and viscosity will increase with the increment of time,
coefficient of the soil in sliding zone are the essential which proves the phenomenon of ‘‘reproduction of
conditions which can define whether the creep sliding viscosity coefficient’’.
of the landslide will turn to a violent sliding. At the Creep rate acceleration and deceleration during ini-
same time, the landslide stability is re-analyzed and tial reservoir impounding obviously reflects that the
evaluated by taking the productive sliding accelera- strain hardening of the sliding mass is mainly due to
tion test as model and corresponding prevention and gradually increased viscosity coefficient. The reser-
treatment measures are put forward. The prediction voir landslide in front of Lijiaxia Dam is an ancient
and study results are proved correct in the subsequent landside always in movement, its slide plane is steep
raising of reservoir level. in the upper portion and flat in the lower portion, and
the sliding-zone soil is of residual strength. Sliding
mass has a loose structure and strong water perme-
5.1 Study of prediction of landslide stability
ability, the change of reservoir level can be responded
(1) Analyses of characteristics of landslide displace- by slide mass movement at any time, which makes it
ment trend impossible all the time that the strength on slide plane
According to the monitoring data for more than ten exceeds the critical strength. Based on this analysis,
years (from 1983 to the time just before impound- it is impossible that the reservoir landslide in front of
ment), the landslides are deformed and moved in a Lijiaxia Dam will come into the stage of acceleration
close relationship with environmental factors. Under creep, i.e., intensive creep is not likely to occur to
the influences of factors such as rainfall, rising and result in surging hazard.
falling of river level, project construction and so on,
the displacement speed of the landslide undergoes a (3) Stability calculation and analyses of sliding
process of sliding acceleration and deceleration. From mass
the next day of reservoir impounding, the speed of dis- ① Rigid body equilibrium and stability calculation
placement increased rapidly at all measuring points, and elastic-plastic finite element analysis show that
the maximum speed rate occurred during mid and late the variation of safety factor (K) and shear stress ratio
stages of reservoir lever rising, e.g. EL. 2, 130 m, (τ/σ), and the stress isoline distribution as well as
about tens or hundreds of times of the speed rate the scope of tensile stress zone in relation to the slid-
before impoundment. The maximum speed ranging is ing mass during the period when reservoir water level
between 20 mm/d to 60 mm/d. Besides the expansion changed to EL. 2,145 m from 2,057 m are in compli-
of down dislocation of previous large cracks, some ance with the sliding mass dynamic and displacement
40 fresh cracks with various sizes occurred on the behaviors during initial reservoir impounding. Except
ground surface, and the sliding mass shows tendency for scope of sliding mass defined by No.3 longitudinal
of further disintegration. The total displacement of all profile, safety factor (K) of other profiles (above reser-
slide blocks is not very high (displacement measured voir water level EL. 2,145 m) slightly increased; stress

1081
(acting on sliding plane) ratio (τ/σ) remains almost such tendency that safety factor K decreases, stress
unchanged (even slightly reduced). Considering the ratio increases with reservoir water level rising. The
effects of continuous rainfall and water level abrupt possibility of potential surge wave hazard during reser-
drop on slide mass stability, the maximum decrease of voir water level rising can’t be ruled out for the power
K value is commonly less than 0.1 (about 10%), lower station. Two treatment measures are proposed as solu-
than the K value of the sliding mass on the same pro- tion to eliminating the above mentioned engineering
file during initial impounding (reservoir water level: risk, one measure is called ‘‘Water Injection Land-
EL. 2,057 m∼2,145 m). slide Acceleration’’, and the other one is called ‘‘Top
② Non-linear finite element calculation shows that Unloading’’. Water injection landslide acceleration
sliding mass stress decreases with the increase of reser- is the most cost-effective solution; top unloading is
voir water level, and tensile stress acting in the tensile the safest engineering solution. The Employer of the
zone on the slope top as well as the scope of the tensile power station decided to adopt top unloading solution.
zone reduced; normal stress and shear stress acting A total of 6.5 × 105 m3 sliding mass located at the
on the main and secondary sliding planes of slide upper part of the slope was removed and discarded
masses decreases with the increase of reservoir water into the reservoir.
level, and shear stress ratio also reduced or reduced at Although landslide acceleration zone and other slid-
low extent, indicating the slope stability will improve ing mass for which no engineering treatment work
during reservoir water level rising in the future. was done re-experienced acceleration to deceleration
③ To verify the analysis results of creep sliding creep, no heavy landslide and surge wave occurred
mechanism and stability of the landslides, predict during subsequent reservoir water level rising; it has
the displacement trend in the continuous raising of been nearly 6 years since reservoir water level reached
reservoir level and minimize the risk of surge, a to the normal water level (EL.2,180 m) in November,
representative sliding block with a total volume of 2001, during this period the landslide before dam has
1.53 × 106 m3 is selected to perform the produc- reached stability state and will no longer be a threat to
tive landslide acceleration test by simulating reservoir the normal operation of the power station.
operation environment. The test result shows that the
geological environment formed through water injec-
tion has satisfied and even exceeded the requirement 6 CONCLUSION
that the decrease of the K value is about 6∼11%
when the reservoir water level abruptly drops from To carry out investigation, design and research activ-
EL.2180 m (or EL.2160 m) to EL.2160 m (or EL.2145 ities for heavy landslide adjacent to reservoir bank is
m) in normal operation conditions plus long term rain- essential. The normal methodology applied for analy-
fall. Both the main and secondary sliding masses are sis of stability and sliding speed of landslide, through
found having new cracks during test and new acceler- which landslide surge wave height is determined, can’t
ated creep was found at various monitoring points on objectively reflect the actual displacement mechanism
sliding mass surface, basically duplicating such creep of landslide, because results of sliding speed calcula-
route as constant speed → transient acceleration → tion may go far from the actual situation. It is necessary
deceleration → constant speed. to study the creep characteristics of sliding-zone soil
in ancient landslide and the creep sliding mechanism
On the basis of the above analyses and test results, of landslide from the viewpoint of dynamics.
it is considered that except for No.3 longitudinal
profile, landslide acceleration test zone and simi-
lar sliding mass will less likely experience landslide REFERENCES
surge wave under future operation conditions, and the
landslide acceleration test result may provide experi- Northwest Hydro Consulting Engineers, CHECC, Stability
mental reference for prediction of landslide stability Forecast and Study Report for No.1 and No.2 Land-
and treatment works. slides of Lijiaxia Hydropower Station after Preliminary
Impoundment, December, 1999.

5.2 Study on landslide stability improvement

The aforesaid analysis of stability prediction shows
that the sliding mass on No.3 longitudinal profile has

1082
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

The technical concept within the Integrative Landslide Early Warning

System (ILEWS)

R. Bell, B. Thiebes & T. Glade

Department of Geography and Regional Research, University of Vienna, Austria

R. Becker
IMKO Micromodeultechnik GmbH, Ettlingen, Germany

H. Kuhlmann, W. Schauerte & S. Burghaus

Department of Geodesy and Geoinformation, University of Bonn, Germany

H. Krummel & M. Janik

GeoFact GmbH, Bonn, Germany

H. Paulsen
Terrestris GmbH & Co KG, Bonn, Germany

ABSTRACT: This paper describes the technical instrumentation used by the ILEWS Project to build up a
landslide early warning system. Measurements of landslide movement and soil-water-status are conducted and
integrated into a physically-based slope stability model for continuous calculation of slope stability and into
surface-analysis early warning model. Beside the technical early warning system the ILEWS Project follows
an integrative approach, also including social scientists, spatial planners and historians. The overall goal is to
provide a complete warning chain, including in-situ measurements described in this contribution, data handling,
automated but supervised analysis, warning and consequent communication and decision making.

1 INTRODUCTION local structures of different countries, and it can be

customized to other natural processes (e.g. rockfalls,
Landslides cause fatalities and economic damage all debris flows).
over the world. Installing early warning systems is one The project ILEWS is funded by the German Min-
option to reduce severe consequences. However, early istry for Education and Research (BMBF) for a period
warning of landslides is a challenging topic because of three years.
the possibilities of prediction vary significantly. A
distinction is not only limited to different types of
landslides it also comprises the differentiation between 2 INTEGRATIVE EARLY WARNING SYSTEM
reactivated and new movements. Due to their already (ILEWS)
moved landslide body, reactivated landslides are more
easily to identify and if required, appropriate monitor- Recent natural disasters (amongst others: tsunami in
ing instruments can be installed. The locality of new December 2004; hurricanes in 2005) clearly demon-
landslides is by contrast rather difficult to predict. strate that effective early warning must not end with
The aim of the project is to design and implement providing a warning, but has to integrate social
a prototype of an integrative early warning system for science, humanities and decision making as well.
known and reactivated landslides, so that information These current demands on integrative early warning
on possible future, possibly catastrophic failures can systems were stated by some of the founders of hazard
be collected in advance. The methodical configura- research in the 70s: ‘‘It is necessary to know how tech-
tion of the early warning system is developed to be nological advances in development of detection and
transferable and modular, i.e. it can be adapted to warning systems for hurricanes, floods, earthquakes

1083
and other hazards can be more effectively translated the project. The paper presented here focuses on the
into official and public actions to protect life and technical part of the project.
property’’ (White & Haas 1975). The ILEWS project
addresses these issues.
Within this project the early warning concept
3 STUDY AREA
is based on three parts: Monitoring (measurement,
data transmission), Modelling (historical and cur-
3.1 The Swabian Alb
rent frequency-magnitude-determination, continuous
slope stability modelling, threshold determination, The study area is located in the Swabian Alb,
early warning modeling) and Implementation (infor- SW-Germany and is part of the southern German
mation management, cooperative risk communication, cuesta landscape (Figure 1).
coordination of early warning with other elements of It consists of Jurassic sediments of limestone over-
an integrated risk management). lying marls and clays. The region shows significant
The concept of the research project requires an landslide activity comprising deep-seated rotational
interdisciplinary approach. Therefore, social scien- and shallow translational landslides as well as creep-
tists, spatial planners and historians participate in ing and flowing processes. Human settlements and

Figure 1. Location of the study area.

1084
infrastructure are frequently affected by these pro- probes. Thus, movements can only be continuously
cesses (Terhorst 1997, Bibus 1999, Bell 2007). Large monitored in two to three depths per inclinometer.
events that occurred in historic times include e.g. the In addition, periodical measurements are carried
landslide at the Bronner Mühle (1960, 3 fatalities) and out using a mobile inclinometer device to improve the
the Mössingen landslide (1983) in which 6 million m3 resolution in depth (measurements are taken in 0.5 m
of material was moved (Bell 2007, Bibus 1986). increments). Periodic measurements started in 2004
and were performed at a more or less monthly basis.
Three inclinometers (produced by Glötzl GmbH)
3.2 Lichtenstein-Unterhausen are used and installed at depths between 8.5 m and
The study area is a settlement area on a historically 15.5 m.
active complex rotational slide. The landslide body
had already been under investigation within the pre- 4.2 Continuous geoelectric measurements
ceding research project InterRISK which documented
an extremely slow movement of 0.8 mm in 2.5 years A geoelectrical soil moisture monitoring system run-
by inclinometer measurements (Bell 2007). The land- ning on a monthly basis showed that monitoring of
slide under investigation shows a flowing movement soil moisture is possible using geoelectrics. However,
after strong summer rainfalls down to depths of 8.5 m. deep insights into the moisture variation were hin-
Sliding is initiated in spring after snow melting and dered by effects caused by the high portion of clay
reaches down to depths of 15 m. While the flowing in the landslide material and due to vegetation effects
movement provides considerable stress on at least one (Bell 2007). New studies of the Geological Survey
house in the study area which suffers cracking dur- of Austria show that even temperature variations have
ing these times, the deeper sliding movement does not an influence on resistivity measurements and must be
cause any damage. investigated. Using the old monitoring set up, all these
effects could not be quantified since the time resolu-
tion was too poor. Thus, a new continuously running
geoelectrical monitoring system is developed.
4 TECHNICAL EARLY WARNING SYSTEM The intention is to permanently install a 2D/3D
OF ILEWS geoelectrical survey system on the landslide. An auto-
matic procedure will be developed for data collection
The time variant soil water dynamics is the key factor and for transferring the data via modem to a central
ruling current geomechanical stability of the land- processing unit. The calibration of the geoelectrical
slide body. This issue is reflected in the kind of early data is done by singular in situ soil moisture mea-
warning system to be set up. surements using TDR-Probes and Tensiometers (see
The early warning modelling concept uses a physi- subchapter 4.3). The processed results of the survey
cally based ‘‘Near Real-Time’’ Early Warning Model will be implemented into a central database and evalu-
as well as a Surface Movement Analysis Early Warning ated and interpreted with respect to the common goal
Model. The system includes sensors for determin- of the development of an early warning system for
ing the load (meteorological data), the inner status landslides.
of the vadose zone (water content, soil suction head, Prior to the installation the landslide body was
and pore water pressure) and the system response examined with geophysical methods (e.g. seismic,
(assessed by monitoring kinematics using inclinome- geoelectric) to determine the geometry of the body
ters and high precision geodetic methods like tacheom- and critical ‘‘hot spots’’ for the setup of the permanent
etry and GPS) for the landslide body. Conventional array. A device developed by the Geological Survey of
sensors as well as novel measurement procedures Austria is used. The survey system will automatically
are used. The monitoring system includes particular perform several surveys each day.
measuring procedures and integrated data recording Other objectives include the optimisation of the
mechanisms, which are specifically adapted to the 2D/3D-geoelectrical survey arrays (e.g. linear, star
early warning of rotational slides. shaped or quadratic arrays, electrode separation) for
The concept and the installations are explained in continuous surveying of soil moisture/flow potential
detail in the following. An overview on the planned at ‘‘hot spots’’, e.g. sliding surface of a potential land-
installations is given in Figure 2. slide and the deployment of an automatic procedure for
periodic data collection and data transfer. Also an auto-
matic procedure to analyse the geoelectrical data for
4.1 Inclinometer and inclinometer chains
interpretation of soil moisture conditions will be devel-
Subsurface landslide movements are monitored con- oped which includes an inversion procedure to create
tinuously with two permanent installed inclinometer geoelectrical resistivity models from apparent resistiv-
chains, each consisting of two to three inclinometer ity data. Scale transformation for soil moisture from

1085
Figure 2. Location of installations (Source: Digital Orthophoto © Landesvermessungsamt Baden-Württemberg (www.lv-
bw.de) AZ: 2851.9–1/11).

singular in situ measurement to spatial information institutes allows determination of water content
by using 2D/3D-geoelectrical survey systems will be profiles along elongated sensor cables of several
incorporated. Ultimately a prototype of an automated meters length. However, this procedure requires a
2D/3D-geoelectrical survey system for monitoring large mathematical effort to analyse signals and locally
the soil moisture/flow potential at potential landslide does not achieve the same accuracy as conventional
locations will be compiled. TDR sensors. As part of the project it will be tested
whether a combination of Spatial TDR and standard
TDR sensors provides a significant information gain
4.3 TDR probes and tensiometers
concerning infiltration, perched ground water tables,
Determining soil moisture in clay or highly elec- and hanging slippage.
trically conductive soils is a technological chal- Beside the TDR probes tensiometers will be
lenge due to energy dissipation during the measuring installed to measure suction capability of the soil.
procedure. The measuring principle Time-Domain- A piezometer gives details on pore water pres-
Reflectometry (TDR) is less prone to these effects and sure. TDR probes and tensiometers will be mainly
thus especially suited for the particular application. installed along the two profiles (indicated in Figure
TRIME-TDR sensors by IMKO supply reliable mea- 2) in three different depths (approximately 2 m, 5
surements even in difficult soils and are therefore used m and 10 m) to better understand water-movement
as the solid back bone of ground water measurements in the soils. Some sensors will also be installed
related to landslides. adjacent to the geoelectric profiles for testing the
The ‘‘Spatial TDR’’ method currently being interpolation of soil moisture based on the point
developed at several German universities and research measurements.

1086
 Another forward-looking aspect of the system inte- advantages with continuous monitoring methods over
gration is the fusion of IMKO’s well established sensor several weeks provided they are being carried out as
technology for recording environmental variables with static measurements. The data achieved in the local
novel self-organising wireless networks. A small num- network comes together in a central control and eval-
ber of the sensors for vadose zone monitoring will uation unit, by that it is possible to calculate the basic
be taken out of the previously built cabled field bus lines between the network points automatically and
system and will be integrated into a wireless network continuously in order to derive station movements
(together with the project partner Terrestris). The wire- from that.
less network nodes from ScatterWeb have a variety The observation deviations with the GPS method
of interfaces to connect the sensors. Here too, hard- show a certain auto-correlation in the range of a few
and software adjustments will probably be required to minutes up to a few hours. Reasons for this are e.g.
match the different interfaces. Robustness, prevention multipath and extension effects of the electromagnetic
of downtime, and energy supply for the planned sensor waves. The dimension of the deviations lies above
network are important aspects of the investigations. the point movements to be expected. A reduction of
These issues are decisive criteria for future applica- those observation deviations can be achieved during
tions of wireless sensor networks for environmental long observation periods and an analysis using the
monitoring. post-processing method. Due to the planned early
warning system a real-time process is wanted here
which arranges the analysis in such a way that a separa-
4.4 Tacheometry
tion of measurement deviations and point movements
Geodetic monitoring measurements are used to get a will be done in a filter approach.
confirmation of predictable changes (e.g. subsidence To get a relative precision of tacheometric measure-
behaviour of buildings) or the proof for a non-expected ment clearly less than 1 ppm a regular examination
or non-predictable change of an object (e.g. land- of the measuring instrument will be necessary. It is
slide). Information is generally being supplied through also not sufficient to introduce the gained meteoro-
selected measuring points. The behaviour of the object logical parameters of the endpoint as representative
can be quantified by analysing the movements over factors for the entire measuring distance. Hence the
time. Closely related with the determination of move- determination of the refractive index plays an impor-
ments is also the question of reasons in order to derive tant role and finally is the precision limiting factor
a causal connection. for the distance measuring especially in mountainous
By means of a specially created geodetic point net- areas.
work that spreads over stable as well as critical slope A successful way for high-precision distance mea-
areas those areas should be identified whose move- suring was taken in the 1990s. Based upon the light
ment intervals differ significantly from other areas dispersion in a turbulent medium, the fluctuations due
due to certain other effects (e.g. increase of humid- to atmospheric exchange processes are described in a
ity, change of pore water pressure, etc.). Absolute model by means of statistic factors. Suitable commer-
movements of ground points in slope areas are being cial systems to measure these atmospheric fluctuations
recorded and compared to reference points via measur- have been developed by Scintec/Tübingen in form of
ing methods such as electronic precision tacheometry the scintillometer measuring systems. Regarding the
and GPS. The inclinometer locations must be linked chosen study areas a scintillometer will be used, which
to the geodetic measuring points to get best redundant can do measurements up to 4–5 km. Its usage will lead
but assignable measuring information on movements. to a significant improvement with the modelling of the
It is the intention to use both geodetic measur- refractive index and can therefore cover huge parts of
ing methods one after another. If we assume that the refractive components which have previously been
the movement intervals are about 0.3 mm/month, the difficult to determine.
measuring resolution of the precision tacheometer of
about 0.2–0.3 mm will be sufficient to carry out
4.5 Early warning modelling
episodic measurements and repeated measurements
about every 2–3 months. The integrated inclinometer Within the physically based ‘‘Near Real-Time’’ Early
measurements not only deliver redundant information Warning Model the continuously measured data, espe-
on areas close to each other but the automatic relative cially soil moisture and rainfall, will be integrated in
measurements can serve as indicators for beginning equations to calculate slope stability. Thus, a contin-
movements in order to possibly initiate monitor- uous safety factor can be calculated and a respective
ing measurements outside the scheduled measuring WebGIS application will be programmed. If the safety
epochs. factor gets lower than a specified threshold value,
As already indicated GPS measurements will be preliminary warning messages are provided in the
carried out in the same measuring epochs. They have WebGIS and SMS are sent to the scientists to check and

1087
validate the warning using sophisticated slope stability The integrative approach of the whole ILEWS
models as well as current subsurface movements. project will help to create user-optimised warning
The slope stability software will also be applied messages and increase the probability that the right
to calculate highly likely sliding circles for rotational actions are taken by all involved parties and the
slides and to identify the best suited slope stability affected people.
equations. The results are then again input for the phys-
ically based ‘‘Near Real-Time’’ Early Warning Model.
6 OUTLOOK
At the end of the optimisation period an autarkic run-
ning early warning model will be set up, which controls
First field installations have been made in autumn and
to some degree itself and must only be supervised by
2007 and others will follow shortly so that the technical
experts.
early-warning system will be operating soon.
Within the Surface Movement Analysis Early
Warning Model all measured movement rates (incli-
nometer, inclinometer chain, tacheometry and GPS)
REFERENCES
are analysed using the approach of ‘‘progressive fail-
ures’’ (Petley et al. 2005). Depending on the way how Bell, R. 2007. Lokale und regionale Gefahren- und
the movement rates change, it is assumed that the Risikoanalyse gravitativer Massenbewegungen an der
catastrophic failure of a slope can be predicted. In a last Schwäbischen Alb. Dissertation. Universität Bonn,
step, it is investigated if both models can automatically 305 S. (Online abrufbar unter: http://hss.ulb.uni-
support each other. bonn.de/diss_online/math_nat_fak/2007/bell_rainer).
Local Early Warning Modelling will also provide Bibus, E. 1986. Die Rutschung am Hirschkopf bei Mössingen
information on the magnitude of potential events. (Schwäbische Alb). Geowissenschaftliche Rahmenbedin-
To better define the endangered areas depending on the gungen—Geoökologische Folgen: Darmstadt, Geoöko-
type and magnitude of the process empirical or physi- Verlag, 333–360.
Bibus, E. 1999. Vorzeitige, rezente & potentielle Massenbe-
cally based run out models will be applied. Historical wegungen in SW-Deutschland—Synthese des Tübinger
frequency-magnitude analyses will help to estimate Beitrags zum MABIS-Projekt. In: E. Bibus and
the recurrence interval of landslides of various mag- B. Terhorst. Angewandte Studien zu Massenbewegun-
nitudes more reliably and to better interpret local gen. Tübinger Geowissenschaftliche Arbeiten. Reihe D2:
measurement results. 1–58.
Petley, D.N., Higuchi, T., Petley, D.J., Bulmer, M.H. & Carey,
J. 2005. Development of progressive landslide failure in
5 CONCLUSION cohesive materials. Geology 33 (3): 201–204.
Terhorst, B. 1997. Formenschatz, Alter und Ursachenkom-
plexe von Massenverlagerungen an der schwäbis-
The technical concept of the Integrative Early Warn- chen Juraschichtstufe unter besonderer Berücksichtigung
ing System (ILEWS) comprises various innovative von Boden- und Deckschichtenentwicklung. Tübinger
approaches which will lead to a better understanding of Geowissenschaftliche Arbeiten, Reihe D 2, 212 S.
the complex landslide mechanisms of the investigated White, G.F. & Haas, J.E. 1975. Assessment of Research
slope and in general. Especially the role of the rainfall on Natural Hazards. The MIT Press, Cambridge Mas-
and snow melting water in relation to the two different sachusetts.
movement types will lead to a significant improve-
ment in the determination of critical situations and the
reliable release of early warning messages.

1088
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

The Åknes rockslide: Monitoring, threshold values and early-warning

Lars Harald Blikra

Geological Survey of Norway/International Centre for Geohazards, Trondheim, Norway
Department of Geology and Mineral Resources Engineering, Norwegian University of Science and Technology,
Trondheim, Norway

ABSTRACT: The Åknes/Tafjord project is a large investigation, monitoring and early-warning project related
to some large unstable rockslopes in western Norway. The risk is due to the generation of large tsunamis when
rockslides plunge into the fjords. The Åknes rockslide is a large rockslide of possible 30–40 million m3 , moving
with a velocity of 3–10 cm/year. The monitoring systems is today based on surface extensometers, single lasers,
GPS, total station, geophones, climate station and borehole instrumentation (inclinometers and piezometers). The
data is implemented in an integrated web-based system. Major challenges are linked to the steep terrain, remote
setting and problems with rockfalls and snow avalanches. Major effort has been put on to get reliable operational
power and communications systems. The movement data so far demonstrates a continuous movement during
the entire year, but with significant seasonal changes. During snow melt in the spring and heavy precipitation
events, the movement rate can increase up to 10 times the yearly mean. This means a movement of up to
1 mm/day. Based on the historical data from the Åknes rockslide and information from historical rockslide
events elsewhere, preliminary early-warning levels have been proposed (green, blue, yellow, orange and red
levels).

1 INTRODUCTION

The Åknes/Tafjord project, lead by the involved

municipalities and the county, is going to take care of
investigations, monitoring and early warning related
to large unstable rockslopes in Storfjorden in west-
ern Norway. The risk is due to the generation of
large tsunamis when rockslides plunge into the fjords
(Blikra et al., 2005). The project has established sev-
eral monitoring systems in the unstable areas of Åknes
and Hegguraksla. The project also includes a regional
hazard analysis of the inner Storfjorden area, includ-
ing Tafjorden, Norddalsfjorden, Sunnylvsfjorden and
Geirangerfjorden (Figure 1). The work is now being
focused on implementing the different installations Figure 1. Locality map showing the monitored sites at the
into an integrated monitoring system. It is then a need Åknes rockslide and Hegguraksla.
for establishing threshold values for alarm functions
on several levels.
Based on the nature of failures in rock slopes the other active measures. One of the most challenging
most, and normally the only method of managing tasks is therefore to develop realistic monitoring
the risks associated with a rock slope movement, is thresholds that are not too conservative resulting in
the development of an effective warning system. As excessive false alarms, but provides adequate warning
pointed out by Crosta and Agliardi (2003), remedial should an event occur (e.g. Froese, 2006, Moreno &
countermeasures are normally not useful when deal- Froese, 2006).
ing with a large rockslide due to the extremely high This paper gives an introduction to the geology,
kinetic energy involved. Suitable emergency planning geometry, volumes, movement pattern and slide sce-
is then the only effective tool to reduce the conse- narios for the Åknes slide. It presents further the first
quences by means of evacuation, road closure and attempt of defining thresholds values, based on the

1089
historical data at Åknes, and gives the outline of the
early-warning system.

2 THE ÅKNES ROCKSLIDE: GEOLOGY,

MOVEMENT PATTERN AND SLIDE
SCENARIOS

The rockslide at Åknes is located on the western side

of Sunnylvsfjorden, stretching from about 900 m at
the upper tension cracks to about 100 m above sea
level at the base (Figure 3). Geological and geophysi-
cal investigations at Åknes indicates that the unstable
area covers almost 0.8 km2 .

2.1 Geology
The Åknes rockslide is situated within the Western
Gneiss Region, dominated by gneiss of Proterozoic
age (Braathen et al., 2004a). The bedrock at Åknes
composes different types of gneisses, but the domi-
nating types is medium grained granitic gneiss to a
dark grey biotite bearing granodiorite gneiss (Ganerød
et al., in press). There also occur biotite-rich schists
up to 20 cm thick.
The morphological investigations shows several
characteristic features, listed below (Figure 3). Figure 3. A shaded relief map of the Åknes rockslide
showing some of the morphological characteristics.
1. An about 500 m more or less continuous back
crack (Upper tension fracture). The tension crack
demonstrates a movement of 1 to 2 m. 4. Prominent slide scars along the southwestern
2. A large depression in the upper western corner of canyon. Historical data indicates a slide in the upper
the rockslide, developed into a graben structure. part in the late 1800, and slides also in 1940 and
The total vertical displacement is from 20–30 m, 1960.
while the real horizontal movement is estimated to 5. Small slide scars in the lower part of the rockslide.
be less than 10 m. 6. Large blocks or parts of the rocks is coming out of
3. A series of tension fractures from the upper to the slope at two particular areas, one in the middle
the middle part of the slope. They are oriented in part and one area in the lowermost part.
a WNW to ESE direction and shows up to 1 m 7. Distinct water springs at the lowermost part of the
openings in bedrock and depressions and collapse slope at about 100 masl. However, there are also
structures in the blocky colluvial debris. smaller springs in the middle part of the slide area.

2.2 Movement pattern

The movement of the Åknes area has been measured
by a series of different methods, including GPS, total
station, ground-based radar, extensometers and single
lasers (Figure 4). The large tension cracks in the upper
part are more than 500 m long and limit the total vol-
ume which is moving. The movement data measured
by periodic GPS and total station demonstrates that
there is a movement of between 2 and 4 cm/year. The
south-western flank has a much larger movement with
Figure 2. The upper back fractures at Åknes. The movement a general trend of 5–10 cm a year, but locally up to
data is indicated. 15 cm/year. The same trend can also be seen on the

1090
 Figure 5. A profile showing the 2D resistivity data at Åknes
with indications of the deformation pattern and interpreted
sliding surfaces. See location in Figure 4.

Based on the geology, the rock structure and the

movement pattern at Åknes, the following main slide
scenarios are postulated during an event:
1. Increased movement in the southern flank, devel-
oping into a major flank collapse. This can further
destabilize the entire unstable area at Åknes.
2. Increased deformation and movement in the lower
part, leading to a toe collapse, developing into a
retrogressive fashion ending in a large failure of
Figure 4. Shaded relief map of the Åknes area showing
movement data measured by periodic GPS and total station.
the entire unstable area.
The extent of the possible unstable area is shown in black line. It is important that the monitoring systems manage
The area with largest movement is indicated by the dashed
black line. Note the location of the profile in Figure 5.
to include the two most probable release areas in the
southern flank and at the toe of the unstable area.

measurements done by the ground-based radar placed 3 MONITORING SYSTEMS

in Oaldsbygda on the other side of the fjord.
The movement data from the Åknes site shows that In order to establish a reliable continuous monitoring
this represent the largest movements observed from network, a series of methods are used, both on the
such instabilities in Norway. It is thus reasonable to surface and in boreholes. The array of sensors was
suggest that the Åknes slope is one of the first large chosen to provide the most complete information as
rock avalanches to happen in Norway. possible in order to cover the entire slope, especially
the upper flank. However there are a series of practical
limitations in terms of distance from measured points
2.3 Slide scenarios to the monitoring instrument, local slope conditions,
rockfall and snow-avalanche hazard and problematic
The depth of the moving mass is still somewhat atmospheric conditions. It is below given an overview
unclear, and two alternative models are considered of the established monitoring systems.
(Figure 5). Seven deep drill holes have been performed Surface monitoring:
at three different locations (Figure 4). The drilling
show that slide planes is situated between 25 and 55 m • Permanent GPS network with 8 antennas
depth, but there are also large fracture zones deeper • Total station with 30 prisms
than 100 m (Rønning et al., 2006). The investigations • Ground-based radar with 8 reflectors
show that the moving part is sliding parallel with the • Five large surface extensometers
foliation planes dipping more or less parallel to the • Surface crackmeters
slope, towards the fjord. The drill cores also show • Surface tiltmeters
several breccias and crushed zones along some of the • Two single lasers (Figure 6)
foliation planes, demonstrating the sliding planes. • Eight 3-component geophones

1091
 larger due to the direction and slope of the individual
extensometers.

4.2 Single lasers

The two single lasers in the upper flank area have been
operational since the beginning of December 2005.
Selected data is documented in appendix 2. The sum-
mary of the data is given in Table 2, showing that
the overall yearly movement is between 5 and 6 cm,
however the movement velocity shows quite large fluc-
tuations. The daily fluctuations can reach up to 1.5 mm
in range, and the threshold values need to account for
this. However, there are peaks with much higher noise
Figure 6. The upper single laser at Åknes. The laser to the which need to be filtered out.
right and the heated reflector plate to the left.

4.3 Water level

Borehole monitoring
There is data on water level in three borehole locations
• Two 50 m long DMS systems (50 inclinometers and from the period 2005 to 2007 (Table 3). The data from
2 pietzometers and 50 temperature sensors in each the borehole show rapid fluctuations with an increase
column)
• Pietzometers, conductivity and temperature sensors
in 3 boreholes Table 1. Summary of the data from the 5 extensometers.

Climate station All data Max. movement

• Temperature mm/ mm/ mm/ mm/ mm/

• Precipitation Extensometers year week day week day
• Two snow-depth sensors
• Wind speed E1 20 0,38 0,055 0,8 0,25
• Ground temperature E2 22 0,42 0,06 1,3 0,25
E3 18 0,35 0,049 0,8 0,12
• Insolation
E4 17 0,33 0,047 1,0 0,23
E5 11 0,21 0.03

4 MOVEMENT DATA AND WATER LEVEL

Table 2. Summary of the monitoring data from 2 single
There is still a limited amount of continuous move- lasers.
ment data from Åknes. The extensometer data from
the tree uppermost extensometers have been operating All data Max. movement
for more than 10 years, and is important for the estab-
Lasers mm/year mm/week mm/day mm/week mm/day
lishment of threshold values for different alarm levels.
There is also more than one year of data from two sin- L1 50 0,96 0,14 3,5 1,3
gle laser systems in the upper flank area. The GPS and L2 41 0,79 0,11 2,7 0,9
total station were established in November/December
2006, and is still in an early monitoring stage. How-
ever, there is sufficient data in order to evaluate the
Table 3. A summary of maximum changes in the water level
noise level. It is at the moment too early to evaluate from the three borehole locations.
the threshold values for the ground-based radar.
Maximum water level changes (m)
4.1 Extensometers Borehole 1 day 2 days 1 week 2 weeks
An overview of the data from the extensometers is Upper 3,8 4,5 4,5 4,5
given in Table 1 and an overview of the long-term Middle 1 1,5 2,7∗ 4
trend is given in appendix 1. The overall velocity of Lower 1 1,3 1,3 1,3
the extensometers is from 11 to 22 mm/year (0.2–
0.4 mm/week). The real total movement is somewhat ∗ 4 days.

1092
of up to 4,5 m/day at the upper location. An increase
of 4 m in late april and beginning of May in 2005
is happening at the same time we see a well-defined
increase in movement both in the single lasers and
extensometers.

5 THRESHOLD VALUES

The operative early-warning system needs different

threshold types and values for the different sensors,
Figure 7. Example of movement data from one of the single
including multisensor thresholds. Alarm-thresholds lasers from Åknes during snowmelt in 2006.
should be defined according to following criteria:

1. Total displacement (absolute readings)

2. Velocity in defined time periods
3. Acceleration

We normally need at least a full annual cycle of

data in order to define thresholds values (site-specific
baseline data). The summary given in tables 1 to 3
has been used as a background for defining thresh-
olds for single lasers, extensometers and water level
in boreholes. One main issue is also to find the dif-
ferent types of noise in the data and finally to define Figure 8. Example of water-level data from the middle
borehole at Åknes during snowmelt in 2006.
thresholds that:

1. Not give to many false alarms.

2. Able to catch real events
3. Provide adequate warnings

The alarm system should be able to pick up some of

the largest and distinct seasonal fluctuations, in addi-
tion to more serious events. The review by Crosta and
Agliardi (2003) indicated further that the prediction
provided by threshold values should be supported by
expert judgment. This will include factors such as the
reliability of the monitoring network, the complex-
ity of the displacement patterns, and the short-time Figure 9. A schematic diagram showing a possible develop-
evolution of meteorological and loading conditions. ment of an event at Åknes, exemplified by surface movement
The Åknes/Tafjord project has as a start chosen to at the upper flank area. The different alarm levels are
use velocity-based threshold values. We need to define indicated.
the period which the velocity trend is going to be based
on. This will be different from instrument to instru-
ment. For the extensometer, which is very stable, we 6 EARLY WARNING
can go down to one day or one hour, while the total sta-
tion needs much longer time due to larger fluctuations. The project is in the process of planning the next
Velocity thresholds for large seasonal fluctuations are important step, which is the total handling of the mon-
shown in the schematic diagram in Figure 9. The nor- itoring data in an early-warning centre. It is the plan
mal situation is the green level, were the movement is to have an operational monitoring and early-warning
in the order of 0,1 to 0,2 mm/day, but this increases system established during the winter 2008. This will
5–10 times during the observed seasonal fluctuations include both technical and geoscientific monitoring.
(Figure 7). The water level has shown to increase up All the monitoring data achieved form the unstable
to 5 m during these episodes (Figure 8). The next areas need to be processed and analysed in order to find
level (yellow) is when the velocity trend from the large movement behaviour and velocity trends. This will
seasonal changes continues to develop. be the basis for establishing threshold values for the

1093
different levels of warning. The early-warning system REFERENCES
also includes the implementation of warning methods
and the project focus on two lines: Blikra, L.H., Longva, O., Harbitz, C. & Løvholt, F.
(2005): Quantification of rock-avalanche and tsunami
• Sirens in all the villages situated in the tsunami hazard in Storfjorden, western Norway, 57–64. I: Sen-
hazard zones neset, K., Flaate, K. & Larsen, J.O. (eds.) Landslides and
• Phone messages based on a continuous updated Avalanches ICFL 2005 Norway.Taylor & Francis Group,
database London.
Blikra, L.H., Jogerud, K., Bergeng, T. & Hole, J. 2007:
In addition, there will be a large effort in planning Åknes/Tafjord prosjektet Status og framdrift for over-
and establish procedures and evacuation routes for all vaking og beredskap. Report Åknes/Tafjord prosjektet
the inhabitants in the tsunami hazard zones. 01–2007 (in Norwegian).
Braathen, A., Blikra, L.H., Berg, S.S. & Karlsen, F. 2004:
Rock-slope failures of Norway; type, geometry, defor-
ACKNOWLEDGEMENTS mation mechanisms and stability. Norwegian Journal of
Geology (NGT) 84, 67–88.
Crosta, G.C. & Agliardi, F. 2003: Failure forecast for
The Åknes/Tafjord project, which is lead by the munic- large rock slides by surface displacement measurements.
ipalities and county, has a corporation with a large Canadian Goetechnical Journal 40, 176–191.
number of national and international institutions, com- Froese, C.R., Murray, C., Cavers, D.S., Anderson, A.K. &
panies and partners: Stranda, Norddal, Stordal and Bidwell, A.K. 2006: Development and implementation of
Ørskog municipalities; Møre & Romsdal county; Geo- a warning system for the South Peak of Turtle Mountain.
logical Survey of Norway (NGU); Directorate for Civil ICRLM.
Protection and Emergensy Planning (DSB); Norwe- Ganerød, G.V., Grøneng, G., Rønning, J.S., Dalsegg, E.,
gian Geotechnical Institute (NGI); Norwegian Uni- Elvebakk, E., Tønnesen, J.F., Kveldsvik, V., Eiken, T.,
versity of technology and Natural Sciences (NTNU); Blikra, L.H. & Braathen, A. submitted: Geological model
of the Åknes rodkslide, western Norway. Submitted to
University of Bergen, University of Oslo; University Engineering Geology.
Centre on Svalbard (UNIS); NORSAR; International Moreno, F. & Froese, C. 2006: Turtle Mountain Field Lab-
Centre for Geohazards (ICG); Fugro Survey A/S; oratory Monitoring and Research Sumary Report 2005.
ISPAS; Unified Messaging Systems AS (UMS); Blom Alberta Energy and Utilities Board, Alberta Geological
Geomatics; ITAS; Norut IT; EL24; Umoe IKT; ETH Survey, Earth Sciences Report 2006–07.
Zürich; University of Milano-Bicocca; University of Rønning, J.S., Dalsegg, E., Elvebakk, H., Ganerød, G.
Lausanne; Universitety of Grenoble, LGIT; CSG, & Tønnesen, J.F. 2006: Geofysiske målinger Åknes og
Italy; LISALab, Italy. A large thanks to all con- Tafjord, Stranda og Nordal kommuner, Møre og Romsdal.
tributors to the Åknes/Tafjord project which have Norges geologiske undersøkelse Rapport 2006.002 (in
Norwegian).
been the basis for this overview paper. This paper is
contribution number 187 of the International Centre
for Geohazards.

1094
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

DInSAR techniques for monitoring slow-moving landslides

D. Calcaterra, M. Ramondini, F. Calò & V. Longobardi

Department of Geotechnical Engineering, Federico II University of Naples, Naples, Italy

M. Parise
National Research Council, IRPI, Bari, Italy

C.M. Galzerano
Moio della Civitella, Italy

ABSTRACT: Remote sensing is being frequently applied in recent years to monitor slope instability
phenomena. Since satellite images cover wide territories with frequent passages, their use seems to be extremely
promising in terms of landslide monitoring. Among the different techniques proposed, the Permanent Scatterers
(PS) technique is a multi-interferogram approach for processing SAR data that allows to overcome the limita-
tions of the traditional DInSAR. In the present paper, some preliminary considerations are developed, deriving
from the use of the PS technique in a small town of Campania region, Italy, where some slow-moving complex
landslides affect the urban area. The SAR data have been relieved from the ERS-1 and ERS-2 sensors of the
European Space Agency in the 1992–2001 time span. The SAR data are being compared with field evidence of
structural damages to urban buildings and infrastructures, in view of further field-based observations, integrated
with classical site monitoring data.

1 INTRODUCTION developed by T.R.E. (a spin-off company from the

Politecnico of Milan, Italy), overcoming the limita-
The space-borne SAR interferometry is a powerful tions of conventional interferometric techniques. The
remote sensing tool, which recently has demonstrated PS technique focuses on the use of long temporal
its ability in detecting and monitoring several natural series of SAR data, extracted from the European
phenomena, including slope instabilities (Metternicht Space Agency archive gathering satellite acquisitions
et al. 2005). The interferometric technique able to since 1991, and on phase stable targets, the so-called
assess ground displacements is the Differential InSAR permanent scatterers. These latter are identified by
(DInSAR) that analyses the differences in phase values means of a statistical study on the amplitude of their
between two SAR images of the same area gathered at electromagnetic signal returns (Colesanti et al. 2003).
different times. Compared to the conventional meth- PS are pointwise scatterers only slightly affected
ods for monitoring a landslide, DInSAR allows to by decorrelation problems, which makes possible at
obtain information about surface deformation over these sites to measure deformations with a very high
wide unstable areas in relatively short times and accuracy (in the order of the millimetre).
at lower costs. Notwithstanding its high potential- Worldwide several applications are demonstrating
ity, DInSAR has not become a full operative tool the high potentiality of the PS technique in differ-
for geoscientists and engineers yet, because of some ent geological fields, including subsidence, mining
limitations. In fact, although successfully applied in related ground instabilities, seismic faults monitor-
several case-studies (Fruneau et al. 1996; Singhroy & ing and landslides (Ferretti et al. 2001; Hilley et al.
Molch 2004; Strozzi et al. 2005), DInSAR is at present 2004; Colesanti at al. 2005; Burgmann et al. 2006;
affected by problems such as the decorrelation noise Colesanti & Wasowsky 2006; Meisina et al. 2006).
induced by random variations of terrain reflectivity Landslides are among the most costly natural haz-
and the atmospheric noise generated by random fluctu- ards in Italy, and are responsible for many fatalities
ations of the atmospheric refraction index (Berardino (Guzzetti 2000). The Campania region of southern
et al. 2003). Italy is one of the sectors in the country where the high-
The Permanent Scatterers (PS) technique is a multi- est frequency of landslides with casualties has been
interferogram approach for processing SAR images, recorded, with several tragic episodes even in recent

1095
years, mostly related to fast-moving soil slide-debris In this paper, we present the preliminary results
flows in pyroclastic terrains: these were the cases of obtained for a small urban centre in southern Campa-
Sarno and Quindici in May 1998; Cervinara in Decem- nia (Fig. 1), selected as case study due to widespread
ber 1999, Ischia in April 2006. At the same time, presence of slow-moving slope instabilities in struc-
structurally complex formations crop out in wide sec- turally complex formations. The main goal of our
tors of Campania region, where slow-moving complex research is to test the use of a satellite-based monitor-
landslides often occur, induced by earthquakes (e.g. ing system framed in the landslide risk assessment, by
Calitri, Senerchia and Buoninventre landslides, all integrating the PS-InSAR results with those deriving
reactivated by the November 23, 1980 Irpinia earth- from traditional methods (i.e. multi-year aerial photo-
quake—Cotecchia 1986; or reactivated by excavations interpretation and field surveys, site investigation and
and embankment constructions or water table rises monitoring campaigns).
(Di Nocera et al. 1996). Therefore, the complex geo-
logical setting of the region, together with an heavy
development of the built-up environment, even in 2 METHODOLOGY
landslide-prone areas, make Campania one of the Ital-
ian areas with higher landslides costs, both direct and The present work is part of a project started with the
indirect (Schuster 1996). analysis of the SAR images available from TELLUS
In order to monitor by satellite the urban areas project, that have been processed by means of PS-
heavily affected by landslides in Campania, the TEL- InSAR technique, initially considering the whole PS
LUS (TELerilevamento Laboratori Unità di Supporto) dataset distributed over the regional territory. A time-
project has been implemented, in the framework of consuming and accurate phase of selection was then
the PODIS project (Progetto Operativo Difesa Suolo), carried out in order to choose, within the boundary
financed by the European Union and aimed at provid- of the Campania region, some landslide areas which
ing technical support to the Soil Defense Department could fulfil the following requirements: presence of
of the Campania Region. To this purpose, 1078 SAR slope movements involving inhabited areas; good den-
images (ascending and descending datasets relieved sity of PS; availability of historical information on
from the ERS-1 and ERS-2 sensors of the European landslide activity. Within the population of case stud-
Space Agency in the time span 1992–2001) have been ies so selected, Moio della Civitella has been chosen
processed by means of PS-InSAR analysis, and more for the preliminary analysis here presented.
than 400.000 permanent scatterers on the regional Following the phase of work above, it is necessary to
territory have been identified (T.R.E. 2006). proceed with the integration of the PS-InSAR results
with those deriving from traditional methods (i.e.
geomorphological mapping, multi-year aerial photo-
interpretation, field surveys, analysis of the landslide
damage to buildings and their comparison with the
observed landslide distribution). This allows to deter-
mine the landside state of activity in the area, which,
in turn, is a fundamental step in the evaluation of the
related hazard. In addition, the aforementioned inte-
gration also permits to recognize and map areas with
relevant surface displacements, even with the possi-
bility within landslide areas to identify zones with
different type and/or amount of deformation. Even-
tually, there is also the possibility to detect unknown
unstable sectors in the area, which were unrecognized
before. These data, implemented in a GIS environ-
ment, will be used to integrate and update the landslide
inventory map; at the same time, performing a tempo-
ral analysis of the slope movements, their connection
with triggering factors (i.e. rainfall, man’s actions,
etc.) will be looked for.

3 THE CASE STUDY

The study area, Moio della Civitella, located in the

Figure 1. Location of the study area. Cilento and Vallo di Diano National Park (Salerno

1096
Figure 2. Panoramic view of Moio della Civitella (to the right) and of Pellare (to the left).

province), was selected among the urban centres

presenting higher values of PS density and velocity
in Campania. At the same time, the different types
of landslides affecting the area, and their interaction
with the anthropogenic structures and infrastructures
(as demonstrated by the observed damage to build-
ings; see farther on), make the village one of the most
interesting as regards analysis of the landslide hazard
and mitigation of the related risk.
Moio della Civitella, whose first settlement dates
back to VI–IV century BC, is a small village of
some 1700 inhabitants and is composed of two urban
centres, Moio and Pellare (Fig. 2). The area is geo-
logically characterized by the presence of the Crete
Nere-Saraceno Fmn. (Bonardi et al. 1988), a struc-
turally complex formation, Cenozoic in age, belonging
to the North Calabrian Unit. The formation is mainly
made up of argillites with intercalated carbonate and
silicoclastic arenites, often weathered at the outcrop.
In the study area, this terrain, showing the typi-
cal features of the highly tectonized rocks of the
southern Italian Apennines (diffuse and pervasive tec-
tonic discontinuities, intense fracturing, extremely
variable bedding, etc.) is overlain by a Quaternary
cover consisting of heterogeneous debris in a silty-
clayey matrix. The differences in lithology, and in
Figure 3. Sketch of landslide inventory map at Moio
the hydrogeological behaviour of the terrains as well,
della Civitella (after IFFI project, available at http://www.
are among the main factors predisposing instability at sinanet.apat.it/progettoiffi). Scale: about 1:22.000.
Moio della Civitella.
The area, characterized by hilly morphologies with
low-gradient slopes at elevations between 600 and
200 m a.s.l., is intensely affected by erosional and grav- classification, the prevailing typologies are complex
itational processes. Among them, landslides occupy a landslides and rotational slides evolving to transla-
top-rank position, as well demonstrated by Figure 3. tional slides (Fig. 3). The main slope movements
Such map is an extract from the IFFI project (Inven- derive from ancient phenomena which have involved
tario dei Fenomeni Franosi in Italia—Inventory of large portions of the slopes, if not their whole exten-
Landslides in Italy) a national project dedicated to sion. This is also the case for the landslides that directly
landslide inventory and mapping in Italy (Amanti affect the urban centres. As common in many sim-
et al. 2001). Following the Cruden & Varnes (1996) ilar settings of the southern Italian Apennines, the

1097
rotational component of the complex slope movements In the first step of the damage assessment, the two
is generally limited to the source areas, whilst most building typologies mainly used in Southern Italy were
of the landslide body consists of the flow compo- distinguished: reinforced concrete and masonry. This
nent, which may reach overall lengths on the order distinction is important to provide a correct interpreta-
of several hundreds of metres. Multiple, or superim- tion of the damage, because apparently similar cracks
posed, phenomena are also frequent. As regards state in the walls influence the stability of typologically dif-
of activity, landslides are generally dormant, whilst ferent structures in different ways (Iovine & Parise
rainfall-triggered reactivations are common during the 2002).
rainy seasons, generally involving limited parts of the The classification of the urban damage at Moio
landslides. della Civitella is based on site observations, from
As shown in the landslide map, the largest slope the presence of cracks in the outer walls, to distor-
movements directly involve the inhabited areas, the tion of structures and walls out of perpendicular, to
lifelines and the main communication routes. Some partially or total collapse. As a consequence, one
phases of partial re-mobilizations have been observed of the main difficulties during the survey was the
in December 2005, in at least one case (Corso frequent and recent renovation works, interesting an
Garibaldi landslide, at Moio della Civitella). high percentage of buildings, which may hide the real
A detailed geomorphological mapping is currently damage.
being carried out, aimed at updating the available land- The preliminary damage analysis and zonation
slide inventory map, and at creating the morphological revealed a predominance of buildings showing a mod-
database information necessary for the ongoing mon- erate damage (class 3): they are characterized by open
itoring campaigns. On the basis of the preliminary cracks in the walls, wall disjunction and badly working
field surveys and interpretation of past monitoring data casings (Fig. 4). Several observed buildings have walls
(boreholes, inclinometers, piezometers), the following presenting cracks of few centimetres, often interesting
elements have to be remarked: rigid furring (light damage indicated by the class 2),
while some buildings are affected by very serious dam-
– below the topsoil horizon, an almost continuous ages (class 5) and interested by the collapse of partition
debris cover is present, about 3 to 25 m thick, made and outer walls together with seriously damaged floor
up of arenaceous and conglomeratic blocks in a and lintels. In terms of percentage (Fig. 5), over a sam-
silty-clayey matrix; ple of 93 examined buildings (74 in masonry, and 19 in
– between the debris cover and the Crete Nere- reinforced concrete), 46.2% of the buildings have been
Saraceno Fmn., a clayey horizon deriving from the classified as class 3, whilst 23.7% belong to class 2.
weathering of the bedrock can be found, whose Even though with decreasing percentages, all the most
thickness varies from 0.5 to 5 m; severe classes have also been documented. 17.2 % of
– the available piezometers revealed groundwater lev- the buildings present, on the other hand, negligible
els regularly hosted by the debris cover, located very damage (class 1).
close to the ground surface (depth <5 m);
– with the exception of some minor reactivations, the
main slope movements seem to be dormant at the
present.

4 SURVEYING THE LANDSLIDE DAMAGE

IN THE URBAN AREA

Detailed field surveys aimed at evaluating distribution

and degree of the landslide damage to buildings at
Moio della Civitella have been carried out, according
to the classification proposed by Alexander (1986),
and later modified and applied to several villages in
southern Italy (Crescenzi et al. 1994; Chiocchio et al.
1997; Iovine & Parise 2002).
The scheme to classify the landslide damage to
buildings is based on a intensity scale divided into
7 classes, from 1 corresponding to negligible dam- Figure 4. Example of buildings affected by damage at
age to 7 indicating a total collapse of the building, Moio della Civitella: a) Class 1: negligible damage; b) Class
minimizing in this way the subjectivity in the survey. 3: moderate damage; c) Class 5: very serious damage.

1098
 total collapse
partial collapse
1,1%
very serious 4,3%
2,2% negligible
17,2%
serious
5,4%

light
23,7%
moderate
46,2%

Figure 5. Pie-chart showing classification of the damage

to buildings at Moio della Civitella (survey carried out in
2006–2007).

5 THE PS TECHNIQUE

The PS-InSAR analysis resulted in the identification

of 345 permanent scatterers (belonging to ascend-
ing and descending datasets) in the urban area of
Moio della Civitella (Fig. 6). The predominant fac-
ing direction of the hillslope (approximately south)
makes suitable for interferometry both the ascending
and descending datasets.
Notwithstanding the PS technique is an operational
tool for processing SAR data since over ten years, the Figure 6. Permanent scatterers superimposed on an
physical nature of permanent scatterers is still object ortophoto. (Image available from Progetto TELLUS—
of studies and researches. To this aim, the first step PODIS—Unità di Supporto Locale no 6 Campania).
in the study has been the evaluation of their location
and typology throughout the investigated area. Good
knowledge of PS characteristics is in fact a fundamen-
tal step in the correct interpretation of the kinematics. 10 cm/year. In the case of Moio, in fact, average val-
The identified PS correspond generally to man-made ues of velocity up to 16.2 mm/year were detected in
objects characterized by various geometry and mate- the time span 1992–2001, thus suggesting to select this
rial, such as statues, lamps, building roofs, belfries, site for further in-depth studies.
sewage covers, etc. The velocity range is quite confirmed by the first
PS can be seen as the benchmarks of a geodetic net- inclinometer data recorded in the four instrumented
work, already available on the territory, that present boreholes installed in the area; in two months the
average velocities in the direction of the Line of Sight measured displacements varied in fact up to 0.3 cm.
(LOS), varying between 0 and 16 mm/year. For subsets With the aim of preliminarily verifying the cor-
of PS characterized by coherence values higher than respondence between PS information and site evi-
0.82 for ascending datasets, and 0.62 for descending dence, significant PS pairs (very close ascending and
datasets, full time series of displacements are available descending PS; high displacement component; PS
(Fig. 7), allowing to explore the relationship between location referred to or correlated to structures) were
precipitation and slope movements, and to analyse analyzed and compared to damaged buildings belong-
the seasonal variations of slow—moving landslides ing to class 4 or 5. Only two PS pairs of such kind
(Hilley et al. 2004). The deformation data are relative were identified, being located near masonry struc-
measures in time, since they refer to a master image, tures with relevant damages. However, for both of
and in space as well, because they are relative to stable them the crack direction is compatible with the dis-
points of reference. placement vectors detected by components of LOS in
The PS technique can be successfully applied to horizontal and vertical planes. This promising, even
areas, as Moio della Civitella, affected by slow mass if preliminary, result, needs to be necessarily con-
movements, because it is particularly difficult to firmed by further observations during the ongoing
detect changes in the satellite-to-ground distance over research.

1099
 this result derives from a limitation of the technique,
or from any other reasons, represents a matter that is
presently being investigated.
The availability of new satellite images of the area
will allow to validate the PS-based monitoring sys-
tem by means of a critical comparison with traditional
methods. To this aim, data on ground movements in
the investigated area is already being collected through
a continuous monitoring campaign (started in March
2007) consisting in inclinometer soundings and topo-
graphic network constituted by over 100 benchmarks.
In addition, a site investigation campaign is being
planned at the present, that includes inclinometer bore-
holes to monitor selected slope movements in the study
area, and allow to compare the data deriving from the
traditional monitoring system with those provided by
the PS technique.

ACKNOWLEDGMENTS

Thanks are due to Regione Campania, Settore

Difesa Suolo (I. Giulivo, F. Galluccio) and Ufficio
Figure 7. Temporal analysis of displacements for PS dell’Autorità Ambientale (A. Risi), and to M.A.T.T.M.
with high coherence and characterized by higher velocities. PODIS Unità di Supporto Locale no. 6 Campania. The
A) Ascending dataset; B) Descending dataset. (Data avail-
interferometric data have been made available thanks
able from Progetto TELLUS—PODIS—Unità di Supporto
Locale no. 6 Campania). to Progetto TELLUS—Unità di Supporto Locale
no. 6 Campania Progetto Operativo Difesa Suolo
(PODIS)—PON ATAS QCS 2000–2006—Ministero
dell’Ambiente e della Tutela del Territorio e del Mare.
6 FUTURE PERSPECTIVES

The built-up area at Moio della Civitella resulted in a

high density of PS, and their spatial distribution and REFERENCES
velocity allow to obtain a synoptic view of the whole
study area. Analysis of the landslide damage to build- Alexander, D. 1986. Landslide damage to buildings. Environ.
ings seems to confirm the PS data, notwithstanding the Geol. Water Sci. 8 (3): 147–151.
above-mentioned problem of renovation works; how- Amanti, M., Bertolini, G. & Ramasco, M. 2001. The Italian
ever, it is necessary to integrate the survey of damage landslides inventory—IFFI Project. In: E. Jorge et al. (eds)
to buildings with conventional geological and geomor- Proc. 3rd Panamerican Symp. on Landslides, Cartagena,
phological methods aimed at identifying the state and Colombia, 29 July–3 August 2001, Sociedad Colombiana
distribution in the activity of landslides. In this way, de Geotecnica, Bogotà, vol. 2: 841–846.
Berardino, P., Costantini, M., Franceschetti, G., Iodice, A.,
the damages induced by slope movements will be dis- Pietranera, L. & Rizzo, V. 2003. Use of differential SAR
criminated from those caused by other causes such interferometry in monitoring and modelling large slope
as thermic effects, bad plans, and inadequate foun- instability at Maratea (Basilicata, Italy). Engineering
dations. Attribution of the observed damage to slope Geology 68: 31–51.
movements is in fact one of the main difficulties in Bonardi, G., D’Argenio, B. & Perrone, V. 1988. Carta geo-
the analysis of landslide damage to buildings in urban logica dell’Appennino Meridionale in scala 1:250.000.
areas. This phase of validation of the observed dam- Memorie Società Geologica Italiana 41.
age to buildings, and their correct attribution to slope Burgmann, R., Hilley, G., Ferretti, A. & Novali, F. 2006.
instability phenomena, is currently in progress. Resolving vertical tectonics in the San Francisco Bay
area from Permanent Scatteres InSAR and GPS analysis.
Some problems have also to be better studied and Geology 34 (3): 221–224.
solved: for example, the slope movement at Pellare, Chiocchio, C., Iovine, G. & Parise, M. 1997. A proposal for
where some reactivations have been recorded, is char- surveying and classifying landslide damage to buildings
acterized by very low displacements for the few in urban areas. Proc. Int. Symp. on Engineering Geology
permanent scatterers located on the landslide. Whether and the Environment, Athens, 23–27 June 1997: 553–558.

1100
Colesanti, C., Ferretti, A., Prati, C. & Rocca, F. 2003. Moni- Guzzetti, F. 2000. Landslide fatalities and the evaluation of
toring landslides and tectonic motion with the Permanent landslide risk in Italy. Engineering Geology 58: 89–107.
Scatterers technique. Engineering Geology 68: 3–14. Hilley, G.E., Burgmann, R., Ferretti, A., Novali, F. &
Colesanti, C., Le Mouelic, S., Bennani, M., Raucoules, D., Rocca, F. 2004. Dynamics of slow-moving landslides from
Carnec, C. & Ferretti, A. 2005. Detection of mining Permanent Scatterer analysis. Science 304: 1952–1955.
related ground instabilities using the Permanent Scat- Iovine, G. & Parise, M. 2002. Schema illustrato per la clas-
terers technique—a case study in the east of France. sificazione ed il rilievo dei danni da frana in aree urbane.
International Journal of Remote Sensing 26 (1): 201–207. Memorie Società Geologica Italiana 57: 595–603.
Colesanti, C. & Wasowski, J. 2006. Investigating land- Meisina, C., Zucca, F., Fossati, D., Ceriani, M. & Allievi, J.
slides with space-borne Synthetic Aperture Radar (SAR) 2006. Ground deformation monitoring by using the Per-
interferometry. Engineering Geology 88: 173–199. manent Scatterers Technique: the example of the Oltrepo
Cotecchia, V. 1986. Ground deformations and slope instabil- Pavese (Lombardia, Italy). Engineering Geology 88:
ity produced by the earthquake of 23 November 1980 in 240–259.
Campania and Basilicata. Geologia Applicata e Idroge- Metternicht, G., Hurni, L. & Gogu, R. 2005. Remote sens-
ologia 21 (5): 31–100. ing of landslides: an analysis of the potential contribution
Crescenzi, E., Iovine, G. & Parise, M. 1994. Analysis of to geo-spatial system for hazard assessment in mountain-
landslide damage in a village in Southern Italy: a prelim- ous environments. Remote Sensing of Environment, 98:
inary report. Proc. Renc. Int. Jeun. Cher. Geol. Appl., 284–303.
Lausanne: 68–72. Schuster, R.L. 1996. Socioeconomic significance of land-
Cruden, D.M. & Varnes, D.J. 1996. Landslide types and pro- slides. In: A.K. Turner & R.L. Schuster (eds), Landslides.
cesses. In A.K. Turner & R.L. Schuster (eds), Landslides: Investigation and mitigation. Transp. Res. Board, Nat.
investigation and mitigation. Nat. Res. Counc., Transp. Res. Council, Spec. Rep., 247: 12–35.
Res. Board Sp. Rep. 247: 36–75. Singhroy, V. & Molch, K. 2004. Characterizing and monitor-
Di Nocera, S., Fenelli, G.B., Pellegrino, A. & Ramondini, ing rockslides from SAR techniques. Advances in Space
M. 1996. Geological and geotechnical problems in slopes Research 33: 290–295.
involved in large old landslides. Proc. of 71st Int. Symp. Strozzi, T., Farina, P., Corsini, A., Ambrosi, C., Thüring, M.,
on Landslides, Trondheim 17–21 June 1996: 1195–1200. Zilger, J., Wiesmann, A., Wegmüller, U. & Werner, C.
Ferretti, A., Prati, C. & Rocca, F. 2001. Permanent Scatterers 2005. Survey and monitoring of landslide displacements
in SAR Interferometry. IEEE Transactions on Geoscience by means of L-band satellite SAR interferometry. Land-
and Remote Sensing 39 (1): 8–20. slides 2 (3): 193–201.
Fruneau, B., Achache, J. & Delacourt, C. 1996. Observa- T.R.E.—TeleRilevamento Europa 2006. Progetto di elab-
tion and modeling of the Saint-Etienne-de-Tinée landslide orazione dati SAR con Tecnica PS, Rapporto Finale.
using SAR Interferometry. Tectonophysics 265: 181–190. Regione Campania.

1101
 Landslides and Engineered Slopes – Chen et al. (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-41196-7

Multitemporal DInSAR data and damages to facilities as indicators

for the activity of slow-moving landslides

L. Cascini, S. Ferlisi, D. Peduto & G. Pisciotta

Department of Civil Engineering, University of Salerno, Italy

S. Di Nocera
Department of Earth Science, University of Naples ‘‘Federico II’’, Italy

G. Fornaro
IREA-CNR, Naples, Italy

ABSTRACT: This work deals with a preliminary test for the definition of slow-moving landslide activity
indicators by the joint use of DInSAR data and landslide-induced damage to properties. In the test area, extending
for almost 489 km2 in the Central-Southern Italy, landslides and creeping zones are classified by taking in account
their typology, state of activity and maximum expected velocity on the basis of geological and geomorphological
criteria. The ground displacements, detected via low-resolution DInSAR data over 5 years, and the available
damage database point out that the DInSAR technique, if adequately supported by in-situ surveys, seems to be
a valuable perspective for the landslide activity assessment.

1 INTRODUCTION Basin Authority of Liri—Garigliano and Volturno

rivers (NBA-LGV).
It is well known that urban planning and risk mitiga- In the study area a landslide inventory is available
tion strategies may obtain a significant enhancement at 1.25,000 scale including location, classification and
from landslide susceptibility and hazard zoning maps state of activity defined on the basis of geomorpholog-
(Fell et al. 2005; Cascini et al. 2005). However, the ical criteria. A dataset on landslide-induced damages
reliability of the zoning maps depends on several fac- to properties is also available.
tors; among these the accuracy of landslide inventory
plays a fundamental role.
A well developed landslide inventory involves the
location, classification, volume, travel distance, state
of activity and data of occurrence of landsliding in a
given area. Focusing the attention on the state of activ-
ity, its definition can be challenging when large areas
are detected and no extensive monitoring systems are
available.
To this aim, the use of advanced satellite tech-
niques, which involve data achieved by Synthetic
Aperture Radar (SAR) (Gabriel et al. 1989), can be
extremely useful. In particular, the differential SAR
interferometry (DInSAR) can complement with tradi-
tional topographic techniques to obtain measurements
of ground surface displacements, providing compa-
rable accuracy while being less expensive and time
consuming (Cascini et al. 2007).
The present work addresses this issue by testing
the usefulness of DInSAR technique inside a study
area (Fig. 1) belonging to the territory of the National Figure 1. The study area.

1103
 In the following, after a brief description of the
test area, the data obtained by using a DInSAR algo-
rithm, are discussed. Then, the attention focuses on the
dataset on damage to facilities and the criteria adopted
for the necessary simplification in terms of elements
at risk aggregation, considering the scale of the study
(1:25,000).
Finally, the ground displacements detected via low-
resolution DInSAR data and the damage dataset are
compared in order to test the reliability of DInSAR
data in checking/updating available landslide inven-
tory maps based on geomorphological criteria.

2 THE STUDY AREA

The study area, extending for 489 km2 , covers the

upper part of the territory of the NBA-LGV and Figure 2. Geolithological map of the study area with
includes eleven Municipalities, belonging to two mapped landslides.
Regions (Lazio and Abruzzo) (Fig. 1 and Table 1).
The geology of t