See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/320868051

Shear wave velocity profiles of fill dams

Article  in  Soil Dynamics and Earthquake Engineering · January 2018

DOI: 10.1016/j.soildyn.2017.10.013

CITATIONS READS

2 411

2 authors, including:

Tadahiro Kishida
Khalifa University
70 PUBLICATIONS   480 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Engineering Implications of the 2014 South Napa Earthquake Strong Motions View project

GEER reconnaissance of central Italy earthquake sequence View project

All content following this page was uploaded by Tadahiro Kishida on 21 March 2018.

The user has requested enhancement of the downloaded file.

 Soil Dynamics and Earthquake Engineering 104 (2018) 250–258

Contents lists available at ScienceDirect

Soil Dynamics and Earthquake Engineering

journal homepage: www.elsevier.com/locate/soildyn

Shear wave velocity profiles of fill dams MARK

a,b,⁎ c,d
DongSoon Park , Tadahiro Kishida
a
K-water Research Institute, 1689 beon-gil 125, Yuseong-daero, Yuseong-gu, Daejeon 34045, Republic of Korea
b
California State University Sacramento, 6000 J St., Sacramento, CA 95819, United States
c
Department of Civil Infrastructure and Environmental Engineering, Khalifa University of Science and Technology, United Arab Emirates
d
formerly Pacific Earthquake Engineering Research Center, University of California, Berkeley, CA 94720, United States

A R T I C L E I N F O A B S T R A C T

Keywords: The shear stiffness (Gmax), which is determined from the shear wave velocity (Vs), is an essential parameter in
Fill dam dynamic analyses of fill dams. In this study, Vs profiles were collected and interpreted after comprehensive in
Shear wave velocity situ geophysical seismic surveys of 28 fill dams. The Vs profiles were compared with the empirical formula
Shear stiffness proposed by Sawada and Takahashi, which was found to overestimate the shear stiffness of a core layer and
Geophysical survey
underestimate that of a shell layer. Regression equations for Vs and Gmax profiles as functions of effective stress
Dynamic analysis
were developed for both the core and shell materials. A regression analysis including the mixed effect model was
Seismic survey
Core layer performed to account for the potential bias of data depending on the material types, survey methods, locations,
and repeatability. The presented results will improve the prediction of Vs and the reliability of dynamic analyses
of fill dams.

1. Introduction surveys. These different survey methods often yield different profiles
[3].
The seismic safety of large dams under strong excitations is an im- In recent engineering practice, in situ Vs profile measurements are
portant issue in terms of dam functionality and their social and eco- preferred to estimate Gmax [4–7]. DHT, suspension logging, SCPT, and
nomic impacts. The International Commission on Large Dams (ICOLD) MASW surveying are frequently selected based on cost and time re-
has collected statistics on 50,000 large dams, of which approximately quirements [4–7]. When the field measurements of Vs are not available,
70% are fill dams [1]. Because many fill dams are located in high- the empirical formulas are typically used to determine the shear wave
seismicity areas and new scientific findings have driven seismic design velocity profile [7–12]. However, in dam practice, the use of invasive
standards to higher levels, reliable seismic analyses have become in- methods for field measurements of Vs is limited because of the ex-
creasingly important. cessive cost and the risk associated with borehole drilling resulting in a
The dynamic analysis of fill dams requires an investigation of the potential seepage hazard. Therefore, the available empirical formulas,
dynamic shear stiffness (Gmax), which can be determined from the shear which are helpful in such cases of lack of adequate data, are limited.
wave velocity (Vs) and the density of the soil (ρ) as Gmax = ρ Vs2. Based on this motivation, this study develops empirical formulas for
Geophysical survey methods are used to obtain the Vs profiles when compacted cored fill dams for correlations of Vs (or Gmax) versus ver-
sufficient funds are available, whereas empirical formulas are often tical effective stress (σ’v).
used when field measurements are not available. Several studies have provided empirical models to determine Vs for
Geophysical survey methods can be classified as invasive or non- the central core and shell materials. These studies used surface wave
invasive. Invasive methods, which involve the use of boreholes, include seismic surveys of a limited number of fill dams [10–12]. Sawada and
downhole surveys (DHT), crosshole surveys, seismic cone penetration Takahashi's empirical formula [4] is widely used to estimate the Vs of
tests (SCPT), and suspension logging. Geophysical seismic wave testing fill dams in Japan and Korea when field measurements are unavailable
using wave propagation provides estimates of stiffness, damping, and [13,14]. This formula is developed by a limited number of dams using
layering characteristics [2]. Non-invasive geophysical survey methods borehole seismic surveys and seismic data analyses of recorded accel-
include spectral analysis of surface waves (SASW), multi-channel ana- eration time histories. Therefore, the application of this formula to
lysis of surface waves (MASW), harmonic wavelet analysis of waves different dams requires specific care because the prediction model may
(HWAW), short-array beamforming (SBF), and seismic reflection range depending on dams, which were not addressed by past studies.

⁎
Corresponding author at: K-water Research Institute, 1689 beon-gil 125, Yuseong-daero, Yuseong-gu, Daejeon 34045, Republic of Korea.
E-mail address: dspark210@gmail.com (D. Park).

http://dx.doi.org/10.1016/j.soildyn.2017.10.013
Received 30 January 2016; Received in revised form 3 August 2017; Accepted 17 October 2017
0267-7261/ © 2017 Elsevier Ltd. All rights reserved.
D. Park, T. Kishida Soil Dynamics and Earthquake Engineering 104 (2018) 250–258

For instance, the degree of saturation with depth is one factor that may profiles measured on a dam crest represent a profile of the core layer
range between dams in the empirical formula. Discrepancies have also exclusively? (4) Can newly developed Vs (or Gmax) profiles be proposed
been observed between the empirical formula developed in past studies. for each core and shell layer?
For example, the model by Kim et al. [10] typically produces greater Vs As a result, the regression models of Vs and Gmax profiles were de-
profiles for the shell layer than for the core layer, while Sawada and veloped for both the core and shell layers. The main features of the
Takahashi's [8] formula produces similar trends for both layers. newly developed Vs and Gmax profiles were explored and compared
Therefore, it would be beneficial to compare previous empirical for- using various seismic survey methods and empirical formulas. More
mulas with the newly developed model by using recent field measure- rigorous statistical models (e.g., residual and standard deviation ana-
ments. lyses of effective stress-dependent Vs dataset using fixed-effect and
Correlations between Vs and the penetration resistance have also mixed-effect models) were introduced for identifying potential bias and
been proposed [15]. Brandenberg et al. [16], Wair et al. [7], and uncertainty coming from dams, survey methods, survey locations, and
Kishida and Tsai [17] proposed useful methods for correlating Vs pro- survey repeatability.
files with penetration resistance, using more advanced statistical
models. Brandenberg et al. [16] suggested multiple linear regression
2. Dams and geophysical survey procedures
models by using N and effective overburden stress, and showed that
60% of the total variance of Vs is contributed by the variance between
Geophysical surveys were conducted for 28 existing fill dams in
boreholes, indicating the importance of adjusting the correlation
Korea. All the dams were built using modern types of construction
models to site-specific conditions. Kishida and Tsai [17] proposed the
equipment with fairly good compaction techniques and quality control.
approach to adjust the Vs correlation models to site-specific conditions
Of the 28 dams, 21 were earth-core rock-fill dams (ECRDs) and seven
based on the conditional probability framework. However, these studies
were concrete-faced rock-fill dams (CFRDs). Every ECRD had a central
were not specifically intended for compacted-core fill dams, and the use
earth core and a granular shell layer. The shell layer was mainly
of the correlation requires borehole drilling and penetration tests.
composed of rocks and/or sand and gravel. Such core and shell layers
Therefore, the development of field-measurement-based empirical
are believed to exhibit significant stiffness contrasts. The CFRDs were
correlations of Vs (or Gmax) versus σ’v (vertical effective stress) is useful
composed predominantly of rock-fill and gravel-fill embankments.
in the analysis of compacted cored fill dams.
Table 1 shows the basic dimensions of the dams, their construction
In this study, in situ seismic survey data was collected from 28 ex-
project lengths, the unit weights of the core, and shell layers, and the
isting fill dams. Multiple downhole surveys were conducted on the dam
geophysical survey methods applied. The unit weights of the materials
crests of 13 earth-core fill dams. The downhole test results were com-
were used in the computation of vertical effective stresses.
pared with surface wave velocity profiles and profiles obtained using
Most of the non-invasive geophysical surveys for the ECRDs were
Sawada and Takahashi's empirical formula [8].
conducted on the dam crests to obtain the Vs profiles of the underlying
This study was based on the four fundamental questions: (1) Are
layers. In a few cases, the surveys were performed on the upstream or
measured Vs profiles obtained for the same dam using different
downstream berms/slopes to measure the Vs profiles of the shell layers.
methods consistent or compatible with one another? (2) Are measured
For the CFRD surveys, all the data were used to obtain the Vs profiles of
Vs profiles consistent with empirically derived profiles? (3) Do Vs
the rock-fill and gravel-fill materials, regardless of the survey location.

Table 1
Basic dimensions and survey methods of the fill dams used in the study.

Dam Dam type Project period H (m) L (m) γt (kN/m3) Survey on dam crest Survey on shell

Core Shell DHT SASW SBF MASW HWAW Reflection SASW MASW HWAW

GD ECRD 1985–1989 39.5 292 19 19 O O

DB ECRD 1986–1990 53.5 326 19 19 O O
YCN ECRD 1974–1980 42.0 300 20 20 O O O
AG ECRD 1968–1971 32.5 223.5 20 20 O O
GP ECRD 2002–2006 35.0 108 19 19 O O O O O
WM ECRD 1985–1994 55.0 407 21 21 O O O
SY ECRD 1962–1965 46.0 300 20 19 O O O O O
DA ECRD 1968–1969 27.0 318 19 19 O O
SA ECRD 1962–1964 22.0 331 19 19 O O
YC ECRD 1977–1979 24.5 120.0 20 20 O O
GC ECRD 1984–1987 50.0 234.0 20 19 O O O O O
SO ECRD 1974–1978 67.0 437.0 20 18 O O
PL ECRD 2001–2007 37.3 390.5 19 20 O O
AD ECRD 1971–1977 83 612 21 22 O O O O
IH ECRD 1984–1993 73 515 20 21 O O
BR ECRD 1990–2000 50 291 19 19 O O O O O
DC ECRD 1975–1981 72 495 19 20 O O O
JA ECRD 1984–1992 58 330 18 20 O O O
JAR ECRD 1984–1996 99.9 562.6 19 19 O O O
SYG ECRD 1967–1973 123 530 20 20 O O
HS ECRD 1990–2002 48.5 205 19 20 O O
YD CFRD 1990–2006 70 498 – 20 O O O O
MY CFRD 1990–2002 89 535 – 20 O
DG CFRD 1999–2005 52 192 – 21 O
NG CFRD 1987–2003 34 1126 – 20 O
BA CFRD 1990–1996 50 282 – 20 O O
ZH CFRD 1996–2007 53 403 – 21 O O
GW CFRD 2000–2012 45 390 – 21 O

Note: ECRD = Earth-core rock-fill dam, CFRD = Concrete-face rock-fill dam, γt = total unit weight of soil, H = height of dam, L = length of dam.

251
D. Park, T. Kishida Soil Dynamics and Earthquake Engineering 104 (2018) 250–258

Downhole survey data are considered the most reliable type of by conducting both right and left strikes for the same depths [19].
survey data for use in obtaining Vs profiles of core layers, because they Downhole surveys have been utilized at potential dam sites to ob-
are obtained from boreholes drilled vertically from the crest to the tain reliable Vs profiles [20–22] but have rarely been used with existing
foundation. In this study, borehole drilling was performed on the crests fill dams [23–25]. Other types of surface wave geophysical surveys are
of the 13 selected ECRDs. Because borehole drilling is difficult for regarded as suitable indirect methods for determining shear wave ve-
CFRDs, downhole surveys were performed only for the ECRDs. The locity profiles that reflect the complex nature of the interaction between
borehole drilling was performed using two types of large-capacity (100- the core and shell layers when elastic waves travel through a dam body.
HP and 175-HP) drilling machines. To prevent any harmful effects from Since there have been few seismic surveys of large fill dams using
rotary wash boring associated with the disturbance of the neighboring boreholes, the downhole Vs data presented in this paper will contribute
soil, no-water boring and continuous single-core barrel sampling tech- to the establishment of in situ Vs profiles for fill dam core layers.
niques were employed. Intentional borehole drilling of fill dam cores is not considered a
When a downhole survey was conducted on a crest, the first profile desirable way of determining dynamic material properties because it
was measured within polyvinyl chloride (PVC) pipes embedded in NX- disturbs the clay core, which is critical for dam safety. Thus, indirect,
sized boreholes (with external diameters of 83.9 mm and internal dia- non-invasive geophysical techniques have frequently been adopted.
meters of 81 mm) to prevent any blockage or deformation from sig- SASW has been employed to determine the shear-wave velocity
nificant lateral compression of the holes. To improve the accuracy of profiles of dams and landfills [9,26–29]. The SASW survey method
the Vs data acquisition, a second profile was then measured im- involves the use of two or more geophones positioned on the ground in
mediately after removal of the PVC pipes from the boreholes. In most a linear array and a dynamic source that generates surface waves
cases, the boreholes collapsed at certain depths because of the sig- (Fig. 1(a)). The layering profile depends on the actual surface wave
nificant lateral pressure of the core layer. Vs data were collected and penetration depth, which is assumed to be approximately one third of
interpreted for both cases—with PVC casing and without PVC casing. the wavelength and is dependent on the frequency content [18].
Comparisons of the two types of Vs profiles thus obtained showed that SASW techniques have some technical limitations—relatively low
they were very similar, although the Vp (P-wave velocity) profiles were resolution, difficulty of interpretation at higher modes, and insufficient
considerably different. On the other hand, although Vs should ideally understanding of near-field effects [30]. Zywicki and Rix [30] devel-
be measured without casing within the core, there is an advantage in oped a cylindrical beamforming technique to overcome the limitations
measuring Vs with casing because it permits Vs measurements to be of plane-wave processing methods. Joh et al. [31] developed the more
obtained at greater depths. advanced short-array beamforming (SBF) technique to overcome
The Vs profiles of the shell layers of the ECRDs were difficult to higher-mode interference in surface-wave testing (Fig. 1(b)). These
obtain because of the lack of accessibility of survey equipment and the techniques were compared with downhole tests, and the results con-
difficulties associated with the installation of the equipment on the ri- firmed their applicability to cored fill dams [31,32].
prap rocks. Therefore, the Vs profiles of the ECRDS were obtained While SASW-type seismic survey techniques have been developed
mainly from surveys conducted on downstream berms. Most of the Vs and have been used mainly in the geotechnical field, MASW surveying
profiles of the CFRDs were obtained on the dam crests. The MASW, has seen widespread use in the field of geophysics because of its sim-
SASW, and HWAW methods were employed in the surface wave surveys plicity and speed [5,6,33,34]. The MASW method evaluates surface
of the CFRDs. The survey data obtained represent the Vs profiles of the waves at relatively low frequencies (typically 1–30 Hz) and a much
granular shell layers of the CFRDs only; they do not reflect the effects of shallower depth range than SASW (i.e., a few meters to a few tens of
the dams’ core layers. meters) [35]. The MASW survey method adopts single-measurement
A downhole test involves the use of a technique for measuring shear surveying of a broad depth range with high levels of redundancy
waves propagated downward toward a geophone within only one cased (Fig. 1(c)) [36]. The use of MASW to determine the dynamic properties
borehole from a stationary surface point (Fig. 1(e)) [18]. To enhance of fill dam materials is growing. The MASW method has also been
the accuracy of the test, mirrored recordings are typically superimposed compared with other geophysical survey methods in other applications

Fig. 1. Conceptual sketch of different geophysical

seismic survey methods.

252
D. Park, T. Kishida Soil Dynamics and Earthquake Engineering 104 (2018) 250–258

[20,24,37–40]. greater depths. Compared to the upper and lower bound profiles by
Another surface-wave seismic survey method that has recently been Sawada and Takahashi's formula, the downhole survey profiles were
developed is the HWAW survey method [41,42]. The HWAW survey close to or lower than the lower bound predictions.
method involves the use of a single pair of receivers and a single array Fig. 3 shows Vs profiles on the dam crest obtained using various
inversion (Fig. 1(d)) [42]. This method primarily utilizes the signal geophysical survey methods. A profile obtained using Sawada and Ta-
portion of the peak signal-to-noise ratio to evaluate the phase velocity kahashi's empirical formula is also plotted for comparison. Fig. 3(a)
and minimize the noise effect [42]. shows that the Vs profiles obtained from DHT are close to or lower than
Seismic reflection is an alternative to surface wave techniques that the lower bound by Sawada and Takahashi's empirical formula [8],
can be used to compute Vs profiles and detect subsurface layers with which is similar to the observation in Fig. 2. Figs. 3(b) and 3(c) show
different impedances [43–47]. High-resolution seismic reflection sur- profiles by MASW and HWAW versus the empirical profiles. The Vs
veying employs P-waves or S-waves generated on the surface. Signals profiles obtained by MASW at the crest are consistent with those from
acquired at multiple receivers are affected by differences in acoustic Sawada and Takahashi's formula, except for a few outliers. The HWAW
impedance, which are the product of the seismic wave velocity and the profiles are much higher than the Sawada and Takahashi line, whereas
density of a layer. the number of sampling profiles (= 2) was limited. The difference
The use of surface-wave seismic survey methods for fill dams has between DHT and MASW may be explained by the fact that the Vs
been debated [36,48]: Two main debatable assumptions concerning the profiles obtained by MASW are affected by the combined response of
applicability of surface wave surveys to earth-core fill dams are pre- 3D core and shell layers, which resulted in higher Vs profiles than those
valent in practice. The first assumption is that Vs profiles measured on a of downhole ones. Therefore, the stiffness of the core layer cannot be
dam crest reflect the properties of the central core layer vertically be- determined from a surface-wave seismic survey at a dam crest. The
neath the crest. The second assumption is that a surface wave seismic trend that Vs profiles from DHT are lower than that those from MASW
survey yields correct and reasonable Vs profiles using an inversion becomes more apparent at greater depths because higher confining
process based on a two-dimensional (2-D) semi-infinite plane-layer pressures may affect the stiffness of the composite layers to a greater
assumption. In reality, three-dimensional (3D) surface topography may degree. However, at relatively shallower depths (less than ~20 m), the
have a significant effect on the energy distribution in a dispersion image difference in Vs by DHT and MASW is not as apparent. Fig. 3(d) pre-
[36,48,49]. Further, ECRDs have 3D geometries and multiple complex sents the Vs profiles by three seismic reflection methods, which does
layers of different stiffnesses. not show a significant increase in Vs at depths greater than approxi-
The objective of this study was not to resolve these technical de- mately 25 m. The seismic reflection survey has low resolution of Vs
bates but rather to investigate Vs profiles obtained using various survey with respect to depth, but the averaged values are similar to the results
methods and to develop useful regression equations for Vs profiles for obtained by other geophysical surveys.
modern types of fill dams [8].
3.2. Vs profiles at the shell
3. Geophysical survey results
Vs profiles measured on the shell of an ECRD and on the crest/shell
3.1. Vs profiles at the crest of a CFRD are compared with empirical profiles in Fig. 4. Fig. 4(a)
compares the MASW Vs profiles with profiles obtained using Sawada
One particular dam (listed as the SY dam in Table 1) was used to and Takahashi's empirical formula [8]. Fig. 4(b) shows a plot of SASW
investigate the similarity of Vs profiles obtained using different survey and SBF survey lines together with the profiles by Sawada and Taka-
methods on the dam crest. Fig. 2 shows a comparison of Vs profiles hashi. Overall, the Vs profiles for the shell layer are considerably higher
measured using three different seismic survey methods (downhole, than Sawada and Takahashi's trend line although there are some dif-
MASW, and SBF). The figure shows that the surface-wave survey data ferences among the Vs profiles obtained using the different survey
are in good agreement with the downhole survey data at depths of methods. This trend seems to be more apparent at greater depths with
approximately 15–25 m, but the Vs values obtained from the surface higher confining pressures.
wave survey exceed the values obtained from the downhole survey at Although there is some degree of scatter in the plots, it is interesting

Fig. 2. Comparison of Vs profiles measured

Vs profiles of SY dam (SY-1) Vs profiles of SY dam (SY-2)
by three different seismic survey methods
100 300 500 700 Vs(m/s) 100 300 500 700 Vs(m/s) used for the SY dam.
0 0
MASW (2013) MASW (2013)
5 5
SBF (2013), SY-1 SBF (2013), SY-2
10 DHT (2013), SY-1 10 DHT (2013), SY-2
Depth(m)

Depth(m)

15 Sawada & Takahashi Sawada & Takahashi

15
1975 1975

20 20

25 25

30 30

35 35

40 40

253
D. Park, T. Kishida Soil Dynamics and Earthquake Engineering 104 (2018) 250–258

Fig. 3. Comparison of Vs profiles measured

(a) (b)
at the crest by different geophysical surveys
Vs profiles on crest (DHT) Vs profiles on crest (MASW) with empirical profiles for core-fill dams.
100 300 500 700 100 300 500 700
Vs(m/s) 0 Vs(m/s)
0

5 Sawada & Takahashi 5

(1975)
10 10
Depth(m)

Depth(m)
15 15

20 20

25 25

30 30

35 35

40 40

(c) (d)
Vs profiles on crest (HWAW) Vs profiles on crest (Reflection)
100 300 500 700 Vs(m/s) 100 300 500 700
0 Vs(m/s)
0

5 5

10 10
Depth(m)

Depth(m)

15 15

20 20

25 25

30 30

35 35

40 40

Fig. 4. Comparison of Vs profiles measured on the

(a) (b) shell by different geophysical surveys with empirical
Vs profiles of shell (MASW) Vs profiles of shell (SASW, SBF) profiles for earth-core fill dams.

100 300 500 700 900 Vs(m/s) 100 300 500 700 900 Vs(m/s)
0 0

5 5

10 10
Depth(m)
Depth(m)

15 15

20 20

25 25

30 30

35 35

40 40

254
D. Park, T. Kishida Soil Dynamics and Earthquake Engineering 104 (2018) 250–258

to note that there are considerable differences overall between the Table 3
MASW and the SASW-SBF surveys. The SASW and SBF profiles of the Resulted regression parameters for lnVs against lnσv’ model.
rock-fill layer are higher than the MASW profiles (Fig. 4). Another
Regression Model:: ln(Vs)i = b0 + b1 lnσv’ + b2 + εi
observation is that the predicted Vs by Sawada and Takahashi for the
shell layer is much lower than most of the profiles measured by MASW, b0 b1 b2 σ
SASW, and SBF methods (Fig. 4). Fig. 4(b) shows that considerable 4.47 0.293 −0.136 (Clay core) 0.176
0.0 (Clay core/granular shell)
uncertainty prevails in determining the Vs profiles of rock-fill shell
0.301(Granular shell)
layers.

4. Statistical analysis

4.1. Statistical processing of Vs profiles

This study develops regression equations for Vs profiles for both the
core and shell layers of fill dams by conducting extensive statistical
analyses of the measured datasets. Since different Vs profiles were ob-
served by different survey methods described in the previous sections,
regression analyses were performed using the individual and total
survey datasets. In the regression analysis, the near-surface Vs values
(for depths less than 4 m) were excluded from the dataset because of its
difficulty in measurements and heterogeneous upper fill layers.
Regression analyses were performed for the following three dataset
(1) downhole data for the core layer, (2) other surface-wave dataset for
the composite (core and shell) layer on the dam crest, and (3) surface-
wave dataset for the shell layer. The corresponding material types are
classified into three groups (i.e., earth core, granular shell, and clay
core/granular shell) depending on dam types, survey locations, and
survey methods. The material type of “clay core/granular shell” was
assigned for surface wave surveys on the crest at the ECRD dam type.
Fig. 5. Variation of regression models with different material types.
Table 2 summarizes the combination of survey locations and methods
with the assigned material types.
Table 3 shows the analyzed regression model between Vs and σ’v for The obtained regression model in Table 3 was further analyzed by
different corresponding materials, where εi denotes the residual of ith adding random effects between dams (ηι) and survey locations (γj) as
measurements. Regression parameters and results are presented in shown in Table 4. ηi and γj denote the residuals depending on ith dam
Table 3 and Fig. 5. The results show that the prediction model ranges site and jth survey location. εijk denotes the residual of the kth mea-
widely, depending on the material type. The predicted Vs increases surement from the ith dam site and jth survey location. Similar analyses
from clay core, clay core/granular shells, and granular shell, which is a were previously performed by Brandenburg et al. [16]. Table 4 shows
reasonable order, based on the geotechnical characteristics. The stan- the regression parameters, where ϕ and ψ denote the standard devia-
dard deviation (σ) was obtained as 0.176. tions between dam sites and survey locations, respectively. The re-
The residual analyses were also performed for the regression model sulting parameters are approximately constant compared to those in
in Table 3 between survey locations and survey methods. We observed Table 3, which indicates that the analyzed dataset is relatively stable for
no significant difference in mean values depending on these differences. the analyzed mixed-effect models. The standard deviations show that
However, there is a difference in standard deviation between geophy- 46% of the total variance is contributed by the variance between survey
sical survey methods. The standard deviations were obtained as 0.160 locations, where 13% is by that between dam sites. This indicates that
and 0.209 for invasive and noninvasive methods, for which the var- the Vs prediction model ranges ± 15% depending on the dam sites and
iance equal tests (e.g. [50]) show that the probability is very small if survey locations. This observation is consistent with those by Bran-
these have the same variance. Therefore, the analyzed data shows that denburg et al. [16] in which the variation between sites contributes to
the noninvasive methods have larger uncertainties than invasive ones. 60% of the total variance, and the predictions can range ± 25% be-
The residuals were also analyzed against σ’v. There was no clear de- tween sites. Based on these observations, the regression models in
pendency of the standard deviation on σ’v by regressing the square of Tables 3, 4 provide reasonable trends based on geotechnical char-
residuals against this variable. Kishida and Tsai [17] similarly men- acteristics and the comparison with past studies.
tioned that the standard deviation was constant against σ’v when Vs was We also analyzed the regression model of ln Vs against ln σ’v by
predicted based on σ’v and standard penetration test blow count, where adding the fixed effects for each survey method. The standard devia-
Brandenberg et al. [16] showed that the standard deviation decreases as tions of residuals were then reviewed based on the geophysical survey
σ’v increases. methods, and are presented in the third column of Table 6. The

Table 2
Material types for measured Vs based on the combination of dam type, survey location, Table 4
and geophysical survey method. Regression parameters by mixed-effect model for lnVs against lnσv’.

Dam type Survey location Survey type Corresponding material Regression Model: ln(Vs)ijk = b0 + b1lnσv’ + b2 + ηi + γj + εijk

ECRD Crest Invasive Clay core b0 b1 b2 ϕ ψ σ

ECRD Crest Noninvasive Clay core/granular shell 4.34 0.319 −0.118 (Clay Core) 0.129 0.123 0.069
CFRD Crest Noninvasive Granular shell 0.00 (Clay core/granular shell)
ECRD, CFRD Berm/Slope Noninvasive Granular shell 0.284 (Granular shell)

255
D. Park, T. Kishida Soil Dynamics and Earthquake Engineering 104 (2018) 250–258

Table 5 Table 6
Regression parameters by mixed-effect model for lnGmax against lnσv’. Measurement uncertainties in Vs depending on the geophysical methods.

Regression model: ln(Gmax)ijk = b0 + b1lnσv’ + b2 + ηi + γj + εijk Geophysical Method category σ from σ from paired σ from repeated
method regression measurements measurements
b0 b1 b2 ϕ ψ σ residuals
9.36 0.638 −0.227 (Clay Core) 0.257 0.246 0.129
0.00 (Clay core/granular shell) DHT Invasive 0.108 0.100 0.105
0.596 (Granular shell) MASW Noninvasive 0.134 0.141 0.127
SASW Noninvasive 0.137 0.105 –

standard deviations were obtained as 0.108 and 0.140 for invasive and
noninvasive methods, respectively, where noninvasive methods have in-depth dam safety inspection and diagnosis program. By using these
0.134 and 0.137 for MASW and SASW in further details, respectively. datasets, the differences between these are calculated as follows:
Therefore, the results show that noninvasive methods have larger un- ΔVs (D1 − D2) = ln Vs (D1) − ln Vs (D2) (7)
certainties compared to invasive ones, after removing the biases for
each survey location. ΔVs (M1 − M 2) = ln Vs (M1) − ln Vs (M 2) (8)
Table 5 lists the regression results for Gmax instead of using Vs in
Table 4. The unit of Gmax is kPa. The parameters for stress scaling and where, the subscripts 1 and 2 denote the first and second paired mea-
standard deviations (i.e. b1, ϕ, ψ, σ) becomes approximately twice surements, respectively. The standard deviations of Eqs. (7) and (8) are
compared to the results in Table 4. This means that a variation in unit obtained as 0.148 and 0.179, respectively. If these residuals are in-
weight does not contribute to variations in these parameters in the dependent, the standard deviations of measurement uncertainty are
dataset. obtained by dividing the standard deviations in Eqs. (7) and (8) with
the square root of 2 (e.g., [51]). The results show that σVs(D) and σVs(M)
are 0.105 and 0.127 for DHT and MASW, respectively, showing that the
4.2. Measurement uncertainty
invasive methods have smaller uncertainties than noninvasive methods.
This observation is consistent with the results in the previous sections
The measurement uncertainties in Vs are further investigated be-
based on the residual analyses from the regression models and those by
tween invasive and noninvasive methods. Some sites have paired Vs
paired Vs measurements. Table 6 summarizes the observed measure-
measurements by different geophysical survey methods at the same
ment uncertainties depending on the geophysical survey methods.
locations (Table 1). The differences in Vs by different geophysical
Based on these results, we can conclude that the standard deviation in
survey methods are calculated as follows.
Vs measurements with the invasive method of DHT is approximately
ΔVs (D − S ) = ln Vs (D) − ln Vs (S ) (1) 0.104, which is 14–23% smaller than the noninvasive method of MASW
and SASW.
ΔVs (D − M ) = ln Vs (D) − ln Vs (M ) (2) Moss [52] summarized the past studies related to the uncertainties
in Vs measurements. Several studies presented the uncertainty in Vs for
ΔVs (S − M ) = ln Vs (S ) − ln Vs (M ) (3) repeated measurements (e.g., Marosi and Hiltunen [53]) or combined
where, the subscripts D, S, and M represent DHT, SASW, and MASW, uncertainties from different survey methods (e.g., Xia et al. [54]).
respectively. The means and biases in Eqs. (1)–(3) were calculated from However, there is no statistical approaches in the past studies, which
the dataset, which shows that the biases between these geophysical separates the measurement uncertainties depending on the methods
survey methods are less than 2%, indicating that all these methods from paired measurements, based on our knowledge (i.e. Eqs. (4)–(6)).
provide similar mean Vs in the entire dataset. The standard deviations The presented results will be useful in quantifying the uncertainty in
are obtained as 0.145, 0.176, and 0.173 for Eqs. (1)–(3), respectively. dam response due to the uncertainty in Vs profiles, which is a critical
These results indicate that the Vs uncertainties increase if MASW is used issue for seismic response (e.g. Pehlivan et al. [55]) and deformation
as one of these pairs. If these residuals are independent, the standard characteristics (e.g. Montgomery and Boulanger [56]).
deviations of these geophysical survey methods are obtained by solving
Eqs. (1)–(3) for each variance as follows. 5. Discussion
2 2 2
σΔV + σΔV − σΔV
σV2S (D) = S (D − S ) S (M − D) S (S − M ) Since the Vs and Gmax correlations with σ’v (and σ’m) proposed in
2 (4) this paper were developed based on measurements obtained in Korea, it
2 2 2 is valuable to review the similarity of Korean construction procedures
σΔVS (S − M )
+ σΔVS (D − S )
− σΔVS (D − M )
σV2S (S ) = and specifications with other international standard practices for
2 (5)
compacted cored fill dams.
2
σΔV 2
+ σΔV 2
− σΔV According to their design documents and construction history re-
S (D − M ) S (S − M ) S (D − S )
σV2S (M ) = cords, all 28 fill dams evaluated in this study were built within the last
2 (6)
50 years using compacted fill and modern types of construction
By substituting the resulting variances in Eqs. (4)–(6), the standard equipment. For the dams for which construction records were available
deviations are obtained as 0.100, 0.105, and 0.141 for DHT, SASW and or geotechnical investigations were recently conducted, the gradations
MASW, respectively, and listed in Table 6. The results show that the of the impervious core materials were typically found to contain
uncertainties with the invasive method appear to be smaller than those 15–20% fines passing the 0.05-mm sieve and to be mostly materials
with noninvasive methods, which is consistent with the previous sec- classified as CL, SC, and SM, with plasticity index values in the range of
tion. 9–28% [57,58]. The impervious cores of the dams listed in Table 1 were
The uncertainties in Vs are also analyzed using repeated measure- constructed with relative compaction levels (= in situ maximum dry
ments with the same geophysical survey method at the same dam sites. density divided by laboratory-determined maximum dry density) of
At the several dam sites, the Vs were measured multiple times for DHT 95–99% (mostly 98%) and field water contents between the OMC
or MASW. For DHT, repeated measurements were obtained with and (optimum moisture content) and OMC + 2%. In addition, the dams
without casing to confirm its effect on the measured Vs. For MASW, were constructed using modern types of compaction equipment (e.g.,
repeated measurements were obtained for every 4–6 years during the sheepsfoot rollers, rubber-tired rollers, and vibratory rollers) and

256
D. Park, T. Kishida Soil Dynamics and Earthquake Engineering 104 (2018) 250–258

standardized layer thicknesses and number of passes. remediation of deteriorated fill dams using a smart grouting and ef-
Based on USBR [59], USACE [60], and Fell et al. [61], the specifi- fectiveness verification system” (Project Award No. 15AWMP-
cations for the studied dams are in good agreement with internationally C104687-01-000000), sponsored by the Korea Agency for
accepted standards for the important compaction variables (e.g., re- Infrastructure Technology Advancement. The SASW and SBF seismic
lative compaction, water content limits, layer thicknesses, compaction surveys were conducted by Prof. Joh SungHo at ChungAng University.
equipment, and number of passes). The consistency of the gradations, The majority of the MASW survey work was performed by HSGEO, Inc.
degrees of compaction, and construction procedures with international of Korea. All this support is greatly appreciated.
standards suggest that the results of this study are applicable to com-
pacted fill dams constructed by modern methods in other countries. References

6. Conclusions [1] Simão P, Pinto Se. Lessons learned from dams behavior under earthquakes.
Springer; 2015. p. 187–246. [In: Perspectives on Earthquake Geotechnical
Engineering].
Dynamic analyses of fill dams require the shear stiffness (Gmax) as an [2] Mayne PW, Christopher BR, DeJong JT. Manual on subsurface investigations.
input, which is determined from the shear wave velocity (Vs) profile. To Washington, DC: National Highway Institute, Federal Highway Administration;
develop empirical Vs and stress-state-dependent Gmax profiles for 2001.
[3] Tran KT, Hiltunen DR. A comparison of shear wave velocity profiles from SASW,
compacted fill dams, comprehensive geophysical surveys (downhole, MASW, and ReMi techniques. ASCE; 2008. p. 1–9. [In: Geotechnical Earthquake
SASW, SBF, MASW, HWAW, and seismic reflection surveys) were Engineering and SoilDynamics IV].
conducted for 28 existing earth-core rock-fill dams (ECRD) and con- [4] Armstrong R. Personal communication; 2016.
[5] Engemoen W. Personal communication; 2016.
crete-faced rock-fill dams (CFRD). After the interpretation of the data [6] Mejia L. Personal communication; 2016.
and statistical regression of Vs and Gmax profiles, the following ob- [7] Wair BR, DeJong JT, Shantz T. Guidelines for estimation of shear wave velocity
servations can be made. profiles, Pacific Earthquake Engineering Research Center. University of California;
2012.

• Based on fixed- and mixed-effect model statistical analyses of the

[8] Sawada Y, Takahashi T. Study on the material properties and the earthquake be-
haviors of rock-fill dam. In: Proceedings of the 4th Japan earthquake engineering
geophysical survey data, newly developed empirical Vs and Gmax symposium; 1975.
profiles are proposed as a function of vertical effective stress for the [9] Kim JT, Kim DS, Park HJ, Bang ES, Kim SW. Evaluation of the applicability of the
surface wave method to rock fill dams. Explor Geophys 2010;41:9–23.
earth-core layer, granular shell layer, and composite layer of zoned [10] Kim J-T, Kim D-S, Park H-J, Kwon H-K. Estimation of dynamic material properties
fill dams. for fill dam: I. In-situ shear wave velocity profiles. J Korean Geotech Soc

• The surface-wave survey data were in good agreement overall with 2009;25:69–85.
[11] Ha I-S, Oh B-H. Study on comparison of methods for estimation of shear wave
the downhole survey data up to depths of approximately 15–25 m, velocity in core layer of existing dam. J Korean Geotech Soc 2008;24:33–43.
but the Vs values obtained from the surface wave survey methods [12] Lee J-W. Case study on estimation of shear wave velocity in core layer of rock-fill
generally exceeded the values obtained from the downhole survey dam using MASW. J Korean Geo-Environ Soc 2008;9:53–60.
[13] Matsumoto N. Personal communication; 2013.
method at greater depths. Thus, the shear stiffness of the core layer [14] KWRA. Dam design guideline. Korea Water Resources Association; 2011.
may be somewhat overestimated by surface wave seismic survey [15] Kramer SL. Geotechnical earthquake engineering. Prentice Hall; 1996.
data. Profiles obtained from surface wave surveys on dam crests are [16] Brandenberg SJ, Bellana N, Shantz T. Shear wave velocity as function of standard
penetration test resistance and vertical effective stress at California bridge sites. Soil
believed to be more representative of the equivalent composite Dyn Earthq Eng 2010;30:1026–35.
stiffnesses of the core and shell layers. Thus, it may be necessary to [17] Kishida T, Tsai C-C. Prediction model of shear wave velocity by using SPT blow
separate the dynamic stiffnesses of the core and shell layers to ob- counts based on the conditional probability framework. J Geotech Geoenviron Eng
2016:04016108.
tain more accurate dynamic analysis results for cored fill dams. On
[18] Mayne P, Christopher B, DeJong J. Manual on subsurface investigation [Publication
the other hand, surface wave survey data obtained on dam crests can No. FHWA NHI-01-031, Federal Highway Administration]. Washington, DC:
be used for the simplified single zoning of a dam with equivalent National Highway Institute; 2001.
shear-stiffness profiles. [19] Campanella R. Field methods for dynamic geotechnical testing: Overview of cap-

•
abilities and needs. In: International symposium on dynamic geotechnical testing II,
A comparison between empirically derived profiles and the down- San Francisco, California, USA; 1994.
hole profiles obtained for core layers in this study revealed that [20] Oh Y-C, Jeong H-S, Lee Y-K, Shon H. Safety evaluation of rock-fill dam by seismic
Sawada and Takahashi's empirical formula overestimates the shear (MASW) and resistivity methods. In: Proceedings SAGEEP, San Antonio, Texas;
2003.
stiffness of the core. For the shell layer (a rock-fill/gravel-fill layer), [21] Hunter LE, Powers MH, Haines S, Asch T, Burton BL. Geophysical investigation of
the Vs profiles were located higher than those obtained using the success dam foundation: An overview. In: Proceedings of the 20th EEGS sym-
Sawada and Takahashi's formula. posium on the application of geophysics to engineering and environmental pro-

• Owing to additional rigorous regression analysis to account for po-

blems; 2007.
[22] Kaneshiro J, Harding R, Johannesson P, Korbin G. Comparison of modulus values
tential bias of datasets, the analyzed dataset is relatively stable for obtained from dilatometer testing with downhole seismic surveys and unconfined
mixed-effect models. It was found that the invasive Vs measurement compressive tests at McKays Point Dam Site, California. In: Proceedings of the 28th
US symposium on rock mechanics (USRMS), American Rock Mechanics Association;
shows less uncertainty than noninvasive surveys, and the corre-
1987.
sponding material type plays an important role in the prediction [23] Llopis JL. In situ seismic investigation of Folsom dam and reservoir project. In: DTIC
model range. Document; 1989.
[24] Miller RD, Ivanov J, Markiewicz RD, O’Connell D. Estimating vibration response of
east Canyon dam, Utah, from P-, S-, and Surface-Wave measurements. In:
The results are expected to be useful in improving the reliability of Proceedings of the 18th EEGS symposium on the application of geophysics to en-
dynamic analyses of fill dams. However, it does not mean that the site- gineering and environmental problems; 2005.
specific shear stiffness profile can be substituted or used in parallel with [25] Olson M, Prashar Y, Inel S. Seismic refraction and downhole velocity surveys for
investigation of the Tinemaha Dam Inyo County, California. In: Proceedings of the
the empirical profiles given here because of the wide variation of Vs 10th EEGS symposium on the application of geophysics to engineering and en-
prediction models depending on the site. If the site-specific shear vironmental problems; 1997.
stiffness profile is measured with good quality, it should be preferred [26] Olson LD, Sack DA. Nondestructive evaluation of concrete dams and other struc-
tures. In: Nondestructive Evaluation of Aging Infrastructure, International Society
over empirical statistical formulas. for Optics and Photonics; 113-124; 1995.
[27] Matasovic N, Kavazanjian Jr E. Seismic response of a composite landfill cover. J
Acknowledgements Geotech Geoenviron Eng 2006;132:448–55.
[28] Park DS. Seismic evaluation of existing earth-cored rockfill dams based on FE and
Newmark type analyses. In: ICOLD Proceedings of the 25th congress and 83rd
The data analyses and interpretation were performed as a part of the annual meeting, symposium hydropower, Stavanger, Norway; 2015.
research project “Development of initiative technologies for [29] Hickey CJ, Römkens MJ, Wells RR, Wodajo L. Geophysical methods for the

257
D. Park, T. Kishida Soil Dynamics and Earthquake Engineering 104 (2018) 250–258

assessment of earthen dams [In: Advances in Water Resources Engineering]. della Società Geologica Italiana, Volume speciale; 213-236; 2005.
Springer; 2015. p. 297–359. [46] Dobrin MB, Savit CH. Introduction to geophysical prospecting. New York: McGraw-
[30] Zywicki DJ, Rix GJ. Mitigation of near-field effects for seismic surface wave velocity hill; 1960.
estimation with cylindrical beamformers. J Geotech Geoenviron Eng [47] Waters KH. Reflection seismology: A tool for energy resource exploration; 1987.
2005;131:970–7. [48] Karl L, Fechner T, Schevenels M, François S, Degrande G. Geotechnical character-
[31] Joh S-H, Ismail NN, Ramli B. Two-dimensional imaging of soil–bedrock interface by ization of a river dyke by surface waves surface waves. Surf Geophys
short-array beamforming technique. Soil Dyn Earthq Eng 2013;53:37–48. 2011;9:515–27.
[32] Joh S-H, Norfarah NI. Short-array beamforming technique for the investigation of [49] Min D-J, Kim H-S. Feasibility of the surface-wave method for the assessment of
shear-wave velocity at large rock-fill dams. J Korean Soc Civil Eng 2013;33:207–18. physical properties of a dam using numerical analysis. J Appl Geophys
[33] Park CB, Carnevale M. Optimum MASW survey—Revisit after a decade of use.In: 2006;59:236–43.
Proceedings of GeoFlorida; 2010. [50] Park DS, Oh JH. Statistical analysis on in-situ material properties for aged cores of
[34] Tian G, Steeples DW, Xia J, Miller RD, Spikes KT, Ralston MD. Multichannel ana- earth-cored fill dams. J Korean Soc Hazard Mitig 2016;16:243–56.
lysis of surface wave method with the autojuggie. Soil Dyn Earthq Eng [51] Park DS, Oh JH. Potential hazard classification of aged cored fill dams. Korean J
2003;23:61–5. Eng Geol 2016;26:207–21.
[35] Park CB, Miller RD, Xia J. Multichannel analysis of surface waves. Geophysics [52] USBR. Design Standards No. 13 Embankment Dams - Chapter 10: Embankment
1999;64:800–8. construction. U.S. Bureau of Reclamation, U.S. Department of the Interior; 2012.
[36] Zeng C, Xia J, Miller RD, Tsoflias GP, Wang Z. Numerical investigation of MASW [53] USACE. Construction control for earth and rock-fill dams. U.S. Army Corps of
applications in presence of surface topography. J Appl Geophys 2012;84:52–60. Engineers, Washington, DC.
[37] Cardarelli E, Cercato M, De Donno G. Characterization of an earth-filled dam [54] Fell R, MacGregor P, Stapledon D, Bell G, Foster M. Geotechnical Engineering of
through the combined use of electrical resistivity tomography, P-and SH-wave Dams. 2nd edition London, UK: CRC Press Taylor & Francis Group; 2015.
seismic tomography and surface wave data. J Appl Geophys 2014;106:87–95. [55] O'brien RG. A general ANOVA method for robust tests of additive models for var-
[38] Carnevale M, Hager J. Combined ground penetrating radar and MASW surveys to iances. J Am Stat Assoc 1979;74:877–80.
locate dam seeps. In: Proceedings of the 21st EEGS symposium on the application of [56] Kagan YY. Modern California earthquake catalogs and their comparison. Seismol
geophysics to engineering and environmental problems; 2008. Res Lett 2002;73:921–9.
[39] Suto K. An application of multi-channel analysis of surface waves (MASW) to hy- [57] Moss RE. Reduced uncertainty of ground motion prediction equations through
drological study: a case history. ASEG Ext Abstr 2012:1–4. Bayesian variance analysis [In: PEER Report 2009/105]. Berkeley: Pacific
[40] Tsai PH, Chen KN. Application of MASW method for evaluating dynamic properties Earthquake Engineering Research Center, College of Engineering University of
of Lu-Liao-His Earth Dam. Trans Tech Publ; 2012. p. 216–9. [In: Applied Mechanics California; 2009. p. 175.
and Materials]. [58] Marosi KT, Hiltunen DR. Characterization of spectral analysis of surface waves
[41] Park H-C, Kim D-S. Development of seismic site characterization method using shear wave velocity measurement uncertainty. J Geotech Geoenviron Eng
HWAW (harmonic wavelet analysis of wave) method (I)-determination of disper- 2004;130:1034–41.
sion curve. J Korean Soc Civil Eng 2004;24:105–15. [59] Xia J, Miller RD, Park CB, Hunter JA, Harris JB, Ivanov J. Comparing shear-wave
[42] Park H-C, Kim D-S. Development of seismic site characterization method based on velocity profiles inverted from multichannel surface wave with borehole mea-
HWAW (harmonic wavelet analysis of wave) method (II)-Experimental setup and surements. Soil Dyn Earthq Eng 2002;22:181–90.
inversion process. J Korean Soc Civil Eng 2004;24:117–24. [60] Pehlivan M, Hashash Y, Harmon J, Rathje E, Stewart J, Silva S, Campbell K,
[43] Steeples DW, Miller RD. Seismic reflection methods applied to engineering, en- Nikolaou S. Influence of shear wave velocity reversals on one-dimensional site re-
vironmental, and groundwater problems. Geotech Environ Geophys 1990;1:1–30. sponse of spatially varied profiles. In: Proceedings of the 6th international con-
[44] Miller RD, Steeples DW, Myers PB. Shallow seismic reflection survey across the ference on earthquake geotechnical engineering, 1-4, Christchurch, New Zealand;
Meers fault, Oklahoma. Geol Soc Am Bull 1990;102:18–25. November 2015.
[45] Brogi A, Lazzarotto A, Liotta D. Structural features of southern Tuscany and geo- [61] Montgomery J, Boulanger RW. Effects of spatial variability on liquefaction-induced
logical interpretation of the CROP 18 Seismic Reflection Survey (Italy). Bollettino settlement and lateral spreading. J Geotech Geoenviron Eng 2016;91:04016086.

258

View publication stats