Use of Gsi For Rock Engineering Design

Use of GSI for Rock Engineering Design
T.G. Carter
Golder Associates, Toronto, Ontario, Canada
V. Marinos
Aristotle University of Thessaloniki, Thessaloniki, Greece
ABSTRACT: In rock engineering design significant advances have occurred in recent years in numerical
modelling capability. Better and more advanced insight is now possible of rock-support interaction and rockmass progressive failure processes. However one major drawback to more realistically evaluating these processes has been and still remains the lack of reliable estimates of strength and deformation characteristics.
Use of the GSI rockmass classification system and the associated m, s and a parameter relationships linking
GSI with the Hoek-Brown failure criterion provides a proven, effective and reliable approach for prediction
of rockmass strength for surface and underground excavation design and for rock support selection. Backanalyses of tunnels, slopes and foundation behaviour using the approach attest to its reliability. One of the key
advantages of the index is that the geological reasoning it embodies allows rating adjustments to be made to
cover a wide range of rockmasses and conditions, whilst also allowing some understanding to be gained of
applicability limits. This paper attempts to outline approaches for the application of GSI for the quantitative
characterization of rockmasses for inclusion in the Hoek-Brown failure criteria for rock engineering design,
not just for the original range of applicability, but also for both ends of the rock competence scale.
1 INTRODUCTION
In recent years, significant advances have occurred
within almost every area of geotechnical design. In
rock engineering arguably the greatest developments
have been in numerical modelling capability. Codes
are now available that can not only afford better and
more advanced insight into rock-support interaction
and rockmass progressive failure processes, but they
are now capable enough to allow synthetic rockmasses to be efficiently built so that design layouts
can be more realistically evaluated. The drawback to
maximizing the advantages that this progress allows
is that to date similar levels of improvement have
not occurred in the observational characterization of
geological variability existing in natural rockmasses.
For effective rock engineering design it is necessary that reliable estimates be available of strength
and deformation characteristics of the rockmasses
on which or within which engineering structures are
to be created, be it a tunnel, a foundation or a slope.
Definition of rockmass properties for a particular
design problem usually involves one or more of the
following data acquisition methods: a) laboratory
testing; b) in situ testing, c) use of rockmass classifications and/or d) back analysis. However, there can
be significant differences in scale between results
that each of these approaches may yield. Additional
1
complications in representativeness may also arise

because, for instance, laboratory scale samples may
not be truly representative of the rockmass due to
natural heterogeneity present for most formations.
Cost and time issues additionally, complicate data
acquisition, particularly when it comes to carrying
out in situ tests. To estimate reasonable geotechnical
parameters for the design of many engineering projects, and in particular for design of tunnel support
for long deep tunnels beneath mountain ranges
where drilling is difficult before construction starts,
and where a back analysis approach would not be
possible, there is no option but to rely upon the use
of some form of rockmass classification scheme that
is correlated with the basic parameters needed for
design. Back-analysis, is indisputably the best way
to estimate appropriate geotechnical parameters,
when construction has started, provided the analysis
approach being used for the parameter definition is
valid. For almost all rock engineering problems,
back-analysis evaluation of deformation measurements generally provides the most insight, and oftentimes yields the most credible parameter understanding, such that it can be used to validate or modify
any parameters previously used for design.
The need for design input parameters is satisfied
nowadays mainly through the use of geotechnical
classification systems. When the earliest of the more
well-known and more commonly applied rockmass

classification systems were initially developed
(Terzaghi 1946, Lauffer 1958, Barton et al. 1974,
Bieniawski 1973), they were intended principally for
tunnelling and mainly as an empirical design method
for the purpose of estimating underground support.
The knowledge base for these approaches were the
hundreds of kilometres of tunnels that had to that
date been successfully constructed under moderate
stresses and in good quality rock conditions. With
the rapid growth of improved numerical design
tools, which now allow progressive failure processes
and also sequentially installed support to be analyzed, and synthetic rockmasses to be built, the need
for acquisition of more reliable rockmass parameters
has also grown.
Introduction of the Hoek-Brown failure criterion
in 1980 has over the last three decades provided a
sound basis for the evaluation of rock mass strength
parameters when predicated on reliable engineering
geological input, but only when parameters are carefully selected with good geological judgment. While
the Hoek-Brown criterion has aided the revolution in
rock engineering, it has over the years also suffered
a lack of precision in definition of input constants. It
was recognized that this was only partially initially
addressed through use of the then available rockmass classification systems. As these were considered at the time somewhat restrictive and not sufficiently linked with real geological observation, and
thus were perceived to be over-rigid, almost mechanical in their application, the Geological
Strength Index, GSI, was developed, initially by
Hoek (1994) but then amplified and improved in
subsequent papers published by Hoek, Marinos and
Benissi (1998) and Marinos & Hoek (2000, 2001)
and Marinos, Marinos and Hoek (2007), resulting in
the now familiar basic chart, as per Figure 1.
According to the original Hoek-Brown failure criterion publication of 1980, the criterion is based upon an assessment of the interlocking of rock blocks
and the condition of the surfaces between these
blocks. The geological strength index (GSI) more
than the other available classification systems was
formulated to attempt to characterize rockmasses
from a more geological rather than a typical engineering approach so as to better meet the need for
delivering reliable input data, particularly related to
those rockmass properties required as inputs into
numerical analysis or into closed form solutions for
designing tunnels, slopes or foundations in rocks.
One of its great advantages over other approaches
is that it allows characterization of difficult-todescribe rockmasses.
This paper attempts to outline the approaches for
application of GSI for quantitative characterization
of rockmasses for inclusion in the Hoek-Brown failure criteria for rock engineering design.
Figure 1. Basic GSI Chart for Visual Geologic

Characterization of Rockmasses
2 THE GEOLOGICAL STRENGTH INDEX

2.1 Background to development of GSI
Right from the inception of the Hoek-Brown failure
criterion, these two authors recognized that for it to
have practical value it must be related to geological
observations that could be made quickly and easily
by an engineering geologist or geologist in the field.
Initially, they considered developing a completely
new classification system during the evolution of the
criterion in the late 1970s, but soon gave up the idea
and settled with using the already published RMR
system. In the early days this worked well because
most of the problems being evaluated in the 1970s
were in reasonable quality rockmasses (RMR>30)
under moderate stress conditions. In the context of
the Hoek-Brown criterion the estimation of a rock
quality value was needed so that it could be used as
a primary scaling correction on intact strength for
use in the criterion. Since the RMR classification
value was being solely utilized in the Hoek-Brown
criterion, for estimating rockmass properties that
would then be used in numerical analyses, it was
recognized that many such analysis methods already
accounted for a number of facets of the basic RMR
description parameters.
Fine grained polyminerallic igneous crystalline rocks,

e.g. andesite, dolerite, diabase and rhyolite
Coarse grained polyminerallic igneous and

metamorphic crystalline rocks,
e.g. amphibolite, gabbro, gneiss, granite,
norite and quartzdiorite
Intact rock samples

Laboratory size samples free from
m=7
m=10
pre-existing fractures
s=1
s=1
Bieniawski, 1974b (CSIR)* rating
100
Barton et al, 1974 (NGI)** rating
500
Very good quality rock mass
Tightly interlocking undisturbed
rock with rough unweathered
m=3.5
m=5
joints spaced at 1 to 3 m
s=0.1
s=0.1
85
100
Good quality rock mass
Fresh to slightly weathered rock,
slightly disturbed with joints
m=0.7
m=1
spaced at 1 to 3 m
s=0.004
s=0.004
65
10
Fair quality rock mass
Several sets of moderately weathered joints spaced at 0.3 to 1 m,
m=0.14
m=0.20
disturbed
s=0.0001
s=0.0001
44
1
Poor quality rock mass
Numerous weathered joints
spaced at 30 to 500 mm with some
m=0.04
m=0.05
gouge.
s=0.00001
s=0.00001
Clean, compacted rockfill
23
0.1
Very poor quality rock mass
Numerous heavily weathered
joints spaced at 50 mm with
m=0.007
m=0.010
gouge. Waste rock
s=0
s=0
3
0.01
*CSIR: Commonwealth Scientific and Industrial Research Organization
**NGI: Norway Geotechnical Institute
Arenaceous rocks with strong crystals and

poorly developed crystal cleavage,
e.g. sandstone and quartzite
Empirical failure criterion

1=3+ (mc 3+ s c2)1/2
1=major principal stress
3=minor principal stress
c=uniaxial compressive
strength of intact rock
m, s=empirical constants
Lithified argillaceous rocks,

e.g. mudstone, siltstone, shale and slate (tested normal to cleavage)
It specifically was set up to avoid including other

factors, such as intact strength, in situ stresses and/or
groundwater pressures, because these factors normally would be allowed for in any sophisticated
analyses. As GSI was initially introduced into the
transfer equations as a direct replacement for RMR
at the time it was proposed that a new variation of
the previously published Hoek-Brown transfer equations should be used for disturbed and undisturbed
rock conditions with the limit GSI for the equation
set equal to 25.
Carbonate rocks with well developed crystal
cleavage,
e.g. dolomite, limestone and marble
In particular the strength term, the structural orientation term and the ground-water term were thought to
likely be being double counted. Similar issues were
evident for the Q system, due to the parameters for
groundwater and stress. While it was appreciated
that both RMR and Q were basically developed for
estimating underground excavation support, and
therefore included some parameters not required for
the estimation of rockmass properties, it was considered that these specific parameters were better dealt
with by means of full structural and effective stress
analyses approaches. It was therefore felt incorporation of these parameters into the methodology for
rockmass property estimation was inappropriate.
It soon also became obvious that both the RMR
and Q systems were difficult to apply to rockmasses
of very poor quality. It was found in particular that
the relationship between RMR and the constants m
and s of the Hoek-Brown failure criterion began to
break down for severely fractured and/or weak
rockmasses. Other classification systems, such as
that by Palmstrm, 2000, 2005 were reviewed and
also found to suffer many of the same problems at
both the low and high ends of the rock competence
scale. While Palmstrm had made an attempt to look
at classifying a broader range of block size than feasible using the RQD classification introduced by
Deere, 1964, both of the RMR and the Q systems
include and are heavily dependent upon RQD. Since
RQD for most weak rockmasses is essentially zero
or meaningless, and as none of the existing systems
seemed to cope well with this problem, it became
necessary to consider an alternative classification
approach. It was decided that the required system
should not include RQD, but would place greater
emphasis on basic geological observations of rockmass characteristics to better reflect the material, its
structure and its geological history. Further, it would
be developed specifically for the estimation of
rockmass properties rather than for estimating tunnel
reinforcement and support. This new classification,
which was considered more an index of rock competence than a mechanical quality rating was thus
termed GSI (geological strength index) in 1992
when work was started on definition of mi relationships for various rock types (Hoek et al. 1992). The
index and its use in the transfer equations from
rockmass description through to mi and s definitions
for application in the Hoek-Brown failure criterion
were further developed by Hoek (1994), Hoek et al.
(1995) and Hoek & Brown (1997) but remained still
a hard rock system, roughly equivalent to RMR at
this stage.
The initial GSI classification was set up to match
the earliest tables for m and s (as per Table 1) and to
address the two principal factors considered important influences on the mechanical properties of a
rockmass the structure (or blockiness), and the
condition of the joints.
m=15
s=1
m=17
s=1
m=25
s=1
m=7.5
s=0.1
m=8.5
s=0.1
m=12.5
s=0.1
m=1.5
s=0.004
m=1.7
s=0.004
m=2.5
s=0.004
m=0.30
s=0.0001
m=0.34
s=0.0001
m=0.50
s=0.0001
m=0.08
s=0.00001
m=0.09
s=0.00001
m=0.13
s=0.00001
m=0.015
s=0
m=0.017
s=0
m=0.025
s=0
Table 1. Original GSI tabulations for m and s for various

geological conditions
The most recent major revision of the HoekBrown criterion was published by Hoek, CarranzaTorres and Corkum in 2002, which resulted in the
following suite of equations:
where
and
with
One of the most important goals of this revision
had been to remove the switch at GSI = 25 which
had been required in the version published by Hoek,
Kaiser and Bawden (1995). In place of disturbed

rock, defined by GSI < 25 and undisturbed rock
for GSI > 25, a more general Blast Damage factor D
was introduced to handle the transition. While also
not perfect this new D factor has proved workable,
although some clarification and refinements of definition have over the years also been necessary, as
discussed subsequently in this paper.
With continuing use worldwide, the GSI system
has continued to evolve. Additional publications,
from 1998 onwards mainly have been aimed towards
providing clarification of application methodology,
primarily aimed at addressing difficulties with usage, particularly for weak inhomogeneous rockmasses. Many of the most notable improvements
have come from work that Hoek and Marinos had
undertaken during tunnelling in difficult ground in
Northern Greece (Hoek et al. 1998, Marinos and
Hoek 2000), (Figure. 2).
the rockmass as the basis for ascribing a GSI value.

Needless to say this unfortunately led to confusion
and mis-definition of GSI values, especially when
other classifications had been in use on the same
project or scheme. Accordingly, last year, rather
reluctantly a set of quantitative codified rules were
introduced on the side of the GSI chart to attempt to
benchmark the ranges so some global consistency
could be achieved between different observers, as
discussed in the next few paragraphs. A two pronged
approach to characterization use was also proposed
(Hoek et al. 2013) purely observational, as per the
original intent using the various already published
charts, and quantitative, defined per specifically
selected scales.
Figure 3 shows a flow chart to aid selection of
approach for arriving at appropriate Hoek-Brown m,
s and a parameters based on either entry point.
Figure 2. Difficult Ground Conditions for undertaking

Rockmass Classification (Photo from the Egnatia Highway).
An extension of the original GSI application

charts for heterogeneous and structurally complex
rockmasses, such as flysch, was initially introduced
by Marinos and Hoek (2001) and recently updated
and extended by Marinos et al. (2007), Marinos et
al. (2012) and Marinos (2014). Specific GSI charts
for molassic formations (Hoek et al. 2005), ophiolites (Marinos et al. 2005), gneiss (in its disturbed
form), and particular cases of limestones (Marinos
2007, 2010) and under particularly difficult geological conditions have been developed from experience
gained during excavation of 62 tunnels as part of the
Egnatia project in Northern Greece,
In its current form the index is based first upon an
assessment of lithology and secondly on structure
and condition of discontinuity surfaces within the
rockmass. As outlined above, historically it was
considered that GSI should be divorced from any
quantitative, numerical coding approach, like RMR
or Q or most other classification systems. Indeed the
focus had been to only allow visual description of
4
Figure 3. Data entry stream for using the Hoek-Brown system

for estimating rock mass parameters for numerical analysis
(Hoek et al., 2013).
2.2 Observational characterization

When a rockmass is exposed in outcrops, in surface
excavations, such as road cuts and/or in tunnel faces
and if sufficient core is available even in borehole
cores, visual assessment of its competence is feasible and assigning a zone of appropriate character is
straightforwardly done directly on the standard GSI
chart shown in Figure 1. In this chart the two basic
parameters of GSI, the blockiness of the mass and
the conditions of discontinuities, are plotted on the
axes respecting the main geological constraints that
govern a formation, thus geologically defining a
sound index that almost all geologists find simple to
assess in the field.
2.3 Quantitative characterization

Experience has shown over the years that quite a
number of mining engineers and equally many civil
engineers without strong geological backgrounds are
less comfortable assigning a quality range based
solely on visual assessment, and as a consequence
some quantified scales were included on the standard GSI chart to aid application in a more quantitative manner. (Hoek. et al, 2013). This modified chart
(which is shown in Figure 4), has been drawn up to
include one of the parameters most wanted to be
avoided RQD, but this was with good reason. It
may have its faults, but these are now well known,
and can be worked around. Moreover, despite these
issues, RQD concept is generally well understood
and accepted to give reasonable estimates of brokenness, provided that one remains within blocky
rockmass conditions, such as are defined in Figure 4.
paper it is thought reliable enough in most circumstances to provide a good starting point for characterization.
The quantification processes suggested for use in
these cases are related to frequency and orientation
of discontinuities and are limited to rockmasses for
which these numbers can quite easily be measured.
In consequence the quantifications do not work well
in tectonically disturbed rockmasses in which the
structural fabric has been destroyed. In such types of
rockmasses the authors recommend the use of the
original qualitative approach based on careful visual
observations.
t must be noted here though, that is meaningless
to attempt to assign a precise unique GSI number for
a rockmass. In almost all situations, rockmasses
have variability, thus GSI is best assigned as a range.
For analytical purposes this range may best be defined by a normal distribution with the mean and
standard deviation values assigned on the basis of
common sense.
2.4 Transfer equations
As is evident from the suite of standard equations
for using the Hoek-Brown criteria, the GSI term is
used in conjunction with appropriate values for the
unconfined compressive strength of the intact rock
ci and the petrographic constant mi, and a measure
of the brokenness as defined by the s and a parameters to allow estimation of the mechanical properties
of a rockmass, in particular, compressive strength
(cm). For defining the deformation modulus (Erm) of
the rockmass, the following relationship, proposed
by Hoek and Diederichs (2006) is recommended:
Figure 4. Quantification of GSI by Joint Condition, as graded

by RMR, and RQD only for the central section of the basic
chart (Hoek et al., 2013).
In the above expression and in the previous equations for m and s the Disturbance Factor, D can be
assessed from Table 2 or from the graphic table in
Hoek et al, 2002. It is noted that D should only be
applied to the blast damage zone (typically up to a
few metres into the wall of a tunnel and for some
proportion of depth into each bench face based on
the slope height for a slope) and definitely should
not be applied to the entire rockmass.
Location
The chart in Figure 4 was also introduced because

the lack of quantified scales on the original GSI
chart had already prompted several other authors to
attempt to quantify the original chart by adding
scales derived from RMR, Q, RMi or other rockmass classifications. Various attempts by Cai et al.
(2004), Russo (2007) and Sonmez & Ulusay (1999),
each individually have merit, but also some limitations. Similarly the quantified chart proposed by
Hoek at al. (2013), shown in Figure 4, has some limitations, but as explained in more detail in the 2013
5
Underground
Excavations
(confined
conditions)
Open Cuts and
Open Pits
(de-stressed
conditions)
D-Factor
0
0.5
0.8
0.7
1
0.7
1
Disturbance Characteristics
High Quality Perimeter Blasting (100% half barrel traces)
or Mechanical Excavation with TBM or Roadheader
NATM excavation in weak rock with mechanical excavation
As above (but with invert heave issues)
Poor Quality Blasting (<50% half barrel traces)
Controlled Blasting (>80% half barrel traces)
Poor Quality Blasting (<50% half barrel traces)
Mechanical Excavation in Weak Rock with Face Shovel etc
Typical Open Pit Production Blasting/Quarry Blasting
Table 2. Guidelines for the selection of the Disturbance

(Blast Damage) Factor D (after Hoek et al. 2002)
In examining these transfer equations between

GSI and the Hoek-Brown criteria constants m, s and
a and also the rockmass modulus expression Erm it
should be noted these equations have been set up to

account for any basic differences in characteristics
between different parent rock fabrics and rockmass
competence. As such they allow the prediction of an
appropriate mb value properly normalized by the intact mi for that rock type or rock fabric while s and
the exponent a are scale invariant, reflecting simply
the degree of brokenness of the rockmass. Defining
an appropriate rockmass modulus value for use in
any modelling however, like m/mi also requires full
normalization by the intact modulus so that it also
appropriately takes into account the inherent stiffness of the parent intact material.
2.5 GSI and specific rockmasses
Figure 5 shows a compilation of all of the charts that
have been published in papers in terms of probable
mi and ci for the parent rock material. Most of the
common GSI ranges for typical gneisses, granites,
ophiolites, limestones, schists, siltstones / mudstones
/ shales, molassic and flysch formations have been
illustrated here, highlighting how minor geological
differentiations affect the geotechnical properties of
the different formations. For more details the reader
is referred to the original publications presenting
these charts for each specific formation as published
by Marinos and Hoek (2000), Hoek et al. (2005),
Marinos et al. (2005) and Marinos et al. (2011).
In this overall chart it should also be appreciated
that differences in assigned GSI commonly occur
due to different tectonism, weathering and alteration
effects on either brittle or soft rock materials. These
effects due to influence of different geological processes are discussed in subsequent paragraphs.
These charts cover a wide range of rockmass
competence from extremely competent hard rockmasses with non-degradable fabrics, typical of the
deep mines in high-strength/high stress conditions
where spalling and bursts characterize behaviour in
the top left corner of the chart through to the lower
right corner of the matrix diagram where rockmass
competence is low and where squeezing and often
significant closure problems are the characteristic
hallmarks (e.g. flysch, shales etc.).
For any rock engineering design the first issue is
geological characterization. This basically defines
where you are within Figure 5 and this then allows
one to establish the potential range of characteristics
for the rockmass, based on the competence of the
parent rock material. For example if one were working in the flysch conditions of central and southern
Europe then one would be looking at a point plotring
in the lower right third of the diagram, while for
rock conditions within a hard rock mine in Canada
or South Africa at significant depth, one might then
be looking more towards the top left corner.
Himalayan, Andean or Alpine mountain belt conditions would cover the complete suite top left to
bottom right.
The size and spread of one's GSI chart on a sitespecific basis could therefore be expected to vary
depending on parent rock type and mineralogy and
thence on macrofabric and overall competence.
An overprint to this whole matrix would be created by the changes in parent rock competence created
by natural processes such as weathering or mineralogical alteration or solution effects. Indeed, for
tropically weathered rockmasses, such as exist in
many parts of the world, competence ratings could
be expected to change several orders of magnitude
in the vertical plane within the matrix in Figure 5.
For Hong Kong granites, for example, one might
move from intact, high-strength, fine-grained grey
granites with high mi and high ci which plot towards the top left of the diagram, progressively
downwards and to the right through the weathering
grades to completely degraded granite essentially a
soil. However mi for this range of rock characteristics would vary less severely (as is shown by many
triaxial tests undertaken on the granites from Hong
Kong for all different weathering grades).
2.1 Intact rock properties
Right from its original formulation in 1980, through
to the current generalized expression:
1'
3'
'
ci mb 3 s
ci
the concept behind the development of the HoekBrown criterion has remained consistent with the
aim being to allow definition of the strength of any
given rockmass based on its composition, considered as a matrix of blocks of intact material set within a discontinuity framework, thus replicating a typical blocky rockmass in hard competent rock
conditions. In the expression the value of ci defines
the uniaxial compressive strength of the intact rock
fragments making up the blocks within the rockmass, with the dimensionless parameters 3/ci , m
and s modifying the contribution of ci to reflect the
brokenness of the rockmass due to the extent of fracturing present within the rockmass. Depending on
the confining stress (3/ci), the degree of block interlocking (mb) and the condition of the inter-block
surfaces (s), these dimensionless components act to
de-grade the intact strength of the matrix blocks to a
strength considered representative for the overall
rockmass, when considered as an equivalent new
material.
Figure 6 shows Hoek-Brown strength envelopes
for a range of commonly encountered rock types.
Approximate values of the ratio between ci and ti

are listed for each rock type in the Figure 6.
Ideally, a representative ci should be determined
by direct laboratory tests under carefully controlled
conditions thereby generating curves similar to those
shown in Figure 6 and in many earlier publications
on the Hoek-Brown criterion. However, in many
cases, this is not possible in practice because of time
or budget constraints or because it is not feasible to
recover samples for laboratory testing (particularly
the case in weak, schistose or tectonically disturbed
rockmasses where the influence of discontinuities
cannot be avoided within laboratory samples).
Under such circumstances estimates of the value

of ci are quite often made on the basis of published
information, simple index type tests (such as point
load tests, where appropriate) or by use of various
descriptive assessments such as have been published
by the International Society of Rock Mechanics
(Brown 1981). The values given in many case are
often general and usually it is better to avoid them
for design purposes. This same problem of strength
estimation is discussed again later, as it is one of the
most thorny issues complicating reliable design
Figure 5. Most common GSI ranges for typical gneisses, granites, ophiolites, limestones, schists, siltstones/mudstones/shales,
molassic and flysch formations in conjunction with a range of mi and ci. (Refer to text and reference list for original papers for more
details on charts)
strength in areas of the rockmass with limited to no

confinement (i.e. with the minor principal effective
stress (3) close to zero), the modified generalized
equation that was formulated in 1992, was structured to allow greater curvature of the envelope
through the tensile range, while still covering the
full range from 0 < s < 1. This however has given
rise to other problems for rocks of particularly low
strength and competence, requiring inclusion of a
tensile cut-off, such as shown in Figure 6 for the siltstone/mudstone group. This approach is one method
to tackle this problem. Other approaches include using the use of transition function equations as explained later in this paper. The issue can also be
tackled during the analysis stage as changing envelope curvature may not be necessary if the need for a
tension cut off is understood and is used correctly
(Hoek and Martin, 2014).
As basis for decisions, where data is missing,
some typical values for mi and ci for various rock
material types commonly encountered worldwide in
engineering projects are presented in Table 3.
Typical
ci
(MPa)
Igneous
Metamorphic
125-250
Figure 6. Hoek-Brown typical Intact Strength envelopes for a

range of igneous, metamorphic and sedimentary rocks (Hoek
and Martin, 2014)
Experience has shown that there is a common

tendency to underestimate the value of the intact
strength in many cases where actual laboratory data
is not available. This is particularly true for weak
and/or tectonically disturbed rockmasses where the
characteristics of the intact rock components tend to
be masked either by structural tectonic elements or
by surrounding sheared or weathered material. These underestimates can have serious implications for
any engineering design and care has to be taken to
ensure that realistic estimates of intact strength are
made as early as possible in the project. In some situations, early estimates can be refined through detailed back-analysis, for example of tunnel deformation, and, while this may require considerable
effort and even the involvement of numerical analysis, the attempt will generally be repaid many times
over in time and cost savings achieved by more realistic designs.
Care must also be paid to ensuring that tensile
tests are carried out so that the strength envelope can
be properly defined. This is of major importance at
the intact sample scale. Because the original HoekBrown criterion had problems with predicting too
high an axial strength and also a finite tensile
8
Intrusive
Felsic
Coarse
(Granite)
Medium
(GranodioriteDiorite)
Mafic
Extrusive
(Volcanic)
Sedimentary
mi
31-33
100-300
Granular
Texture
(Granulites,
Quartz Gneiss)
85-350
Medium,
amorphous
(Amphibolite)
Coarse
(Gabbromember of
ophiolites)
75-350
Fine,
amorphous
(Homfels,
Quartzite)
50-200
Bended/
Gneissose
(Biotitic Gneiss)
30-100
Folliated
(Phyllite, Slate)
20-60
Schistose
(Schist)
10-50
Mylonitized
(Sericitic Schist,
Mylonite)
28-30
Mafic
(Basalt)
Intermediate
(Andesite)
Felsic
(Rhyolite)
Coarse
(Conglomerate-not
clayey)
25-27
Medium
(Dolerite /
Diabase
member of
ophiolites)
Medium quartz
cemented
(Sandstone/
Sandstone members
of flysch or
molasse/greywacke)
17-20
Fine
(Serpentinitemember of
ophiolites)
Medium
carbonates
(Limestone)
13-16
Fine, (clastics)
(Siltstone/ Siltstone
members of flysch or
molasse/tuff)
Fine, Calc-rock
(Chalk/marl)
Ultrafine
(Claystone,
Mudstone / sheared
Siltstone, Shale
members of flysch)
10-12
7-9
4-6
Table 3. Typical values for ci and mi for range of igneous, metamorphic and sedimentary rocks (to be read in conjunction with
Figure 6 regarding parent rock type characteristics)
This table can be used for preliminary estimating,

but it is far better that the value of the constant mi, as
for the intact strength ci, be determined by triaxial
tests, yielding envelopes similar to those in Figure 6.
When this is not possible, estimates can be made
based on published values (e.g. from the program
RocLab or from published tables from Brown 1981
and/or from Stacey & Page 1986 or other sources).
Note though that great care must be taken when
applying such strengths depending on the required
engineering application, as the data in Table 3 and
much of the data in Rocklab is for dry rock. Values
of mi can also differ somewhat depending on the

state of saturation. In this regard it is always better
to carry out site specific laboratory triaxial testing in
an attempt to obtain realistic test data, not just for
intact strength ci, but also for determining intact
tensile strength ti, so that the full Hoek-Brown envelope can be established and a representative mi
value properly established. This is particularly important for both ends of the rock competence scale
i.e. for both the weak and for the very competent
rocks, and also for weathered or altered intact rock
materials.
For brittle rocks towards the high end of the rock
competence scale, where designs might be needed
for high stress situations, it is critical that the onset
of brittle failure behaviour is established by proper
laboratory testing. For such rock, and for the weak
rocks also there is a need for more attention being
directed to undertaking representative laboratory
tests. In the opinion of Professor Brown (Hoek, pers.
comm.), the fact that plug-in strength tables are
available in codes such as Rocklab has resulted in a
significant decrease in the amount and the quality of
triaxial testing carried out on projects because many
users tend to take the easy way out by simply relying
on tabulated data. In this light, the relative expense
of acquiring a little testing data should be seen as
trivial compared with the potential cost implications
of a lack of good reliable data that might otherwise
compromise subsequent design decisions.
such as frequent bedding or schistosity planes, often

with Fair or Poor conditions on these discontinuities.
In tectonic areas, particularly if compressional,
GSI values may be considerably reduced since the
structure is more fractured (more joints) or even
sheared; such rock behaves in a more ductile manner
(e.g. mudstones, shales, siltstones). In the last case
joint condition is probably Poor to Very Poor (with
slickensided or soft clay coatings due to shearing).
The intact rock strength ci and the mi value may also be reduced in the case of shearing.
2.2 GSI and intact rock strength

Moving on from the discussion above, it must be
remembered that the strength of the intact rock does
not affect the GSI value. GSI is only dependent on
the fracturing degree and on discontinuity surface
condition. GSI values can be high (e.g. >70) in a
rock of low ci. For example marl or siltstone with
low ci can form a very good structure that may be
described as blocky or even intact.
On the contrary, a rock with high strength can be
initially characterised as Good due to its high ci but
may become Poor when it is heavily fractured (e.g.
along a fault zone). In this case, GSI can be low (e.g.
25-30), if the discontinuity surface condition is also
Poor. An indicative example of how tectonism (low
to severe) affects the GSI is shown in Figure 7.
However, if the low strength of the intact rock
has derived from weathering, alteration or shearing
through tectonism then GSI is also reduced with the
disturbance of the structure.
2.3 GSI and tectonism
When tectonism is low, GSI values
very high (Intact to Blocky structure).
values can however be acquired when
are crossed by discontinuities formed
9
are high to
Lower GSI
rockmasses
by genesis
Figure 7. Indicative example of how tectonism (from low to

severe) affects the GSI.
2.4 GSI, weathering and alteration

Before leaving the topic of correct input parameter
definition it is important to realize that rarely will a
project be executed entirely in fresh rock. Only in
deep tunnels and mining situations is it likely that
one would be entirely in dry, fresh and unweathered
rock. On the other hand, alteration can be from none
to severe in different depths. Accordingly, some
consideration needs to be given to the impact that
these natural processes of weathering and alteration
can have in degrading intact rock material quality,
strength and deformability from initial intact state.
There is much confusion in the engineering rock
mechanics (non-geological) literature regarding these two processes, as the terms are so often used interchangeably, largely because there is overlap in
characteristics, but the origin is different.
Alteration is an effect brought about by generally

deep geological processes hydrothermal alteration,
metamorphic alteration etc. Many rocks that are
heavily altered are common in mining situations as
they tend to be located in immediate proximity to
the main ore body, so oftentimes form the hangingwall or footwall of the stopes for which rock engineering design is needed.
Weathering by contrast, is almost universally a
shallow depth decomposition process brought about
by two different, but linked processes mechanical
and chemical degradation. Again in many engineering situations one has to design engineering projects
in weathered rocks. The various stages of weathering of intact rock, rockmass and discontinuities have
been described by ISRM (1981) in certain grades
(from fresh rock W-I to clayey-sandy soil W-VI).
Other descriptions for weathering have been made
by the engineering group of the British Geological
Society (Anon, 1995).
In the context of GSI, the influence that either
process exerts on the ascribed value of GSI is that
both degrade not just the parent intact rock material
but also they change the character and competence
of the rockmass fabric. According to the weathering
degree the discontinuity surface condition becomes
poorer and the interlocking of rock blocks becomes
loosened. The structure on the other hand may not
be in principle affected, at least if weathering is not
very advanced. An indicative example of how this
increase in weathering degree affects the GSI value
is illustrated in Figure 8. In weathering degrees W-II
and W-III, discontinuity condition is shifted to the
middle or right columns of the GSI chart (Figure 8).
In W-IV to W-V, joint condition is Poor to Very
Poor due to the weathering products along the joints.
By the same token, almost all of the processes of
mechanical weathering - which change the rockmass
fabric and break down the blocks into smaller pieces
are describable through routine application of the
GSI definition scales on the standard chart.
The degradation of the parent material within the
intact pieces within the fabric though needs special
consideration. Figure 9 provides some guidance for
assessing the influence of chemical/mineralogical
changes that occur due to pervasive weathering.
The main effect of weathering though is mainly
concentrated in the intact rock strength properties
like ci and mi, since interlocking and bonding of
grains is loosened. Reduction factors from intact
rock strength (Stacey and Page, 1986) according to
the weathering grade are presented in Figure 9. Note
that at W-III state the rock material is not yet friable
like in W-IV. The change from W-III to W-IV is
however a critical boundary since there is generally
considerable reduction to intact properties at this
change, often way in excess of 50% of fresh intact
material, while GSI may also by affected to some
degree by reduction of the structure also.
10
However, the actual changes that will occur in a

given situation are very much rock-type dependent.
Some rocks are resistant to intact material fabric
change. These are the rocktypes that the aggregate
industry chooses for high durability, some diabase,
some hornsfels, some diorites are typical of very
high durability rock materials. Other rocks tend to
decompose completely with penetrative weathering
or alteration. Even rocks, which are quite competent,
can be quite prone to degradation and decomposition. Many feldspathic granites, for instance can
weather appreciably, with the feldspars turning to
clay minerals. Some rock materials absorb water and
disintegrate as a consequence, some by swell processes, others by drying processes, depending on the
internal mineralogy. Control of decomposition is often cementitious material controlled, commonly
quartz cements are stronger and more durable than
calcite/carbonate cements and even better than iron
oxide/heamatitic cements. Processes of laterization which are typical in deep, tropically weathered areas
also reduce rock material strength, particularly at the
saprolite boundary. In such rocks the intact structure
may become pervasively degraded right within the
rock material itself.
Figure 8. Indicative example of how the weathering degree

(W-I W-V) affects the GSI.
Most weathering and alteration processes create

major changes in rock material character, which in
turn alters the parent material mi, ci and Ei, let
alone the changes that occur to the rockmass fabric -
all of which affect GSI. This is why it is not recommended just taking published mi and strength values
for dry unweathered intact material based on tables,
such as Table 3 or from RockLab listings as these
could be seriously in error for actual site conditions.
Alteration in principle also affects both the intact
rock properties of the material and the joint surface
condition. In fresh conditions, unaltered rockmasses
can be generally massive, strong rocks with sparsely
spaced discontinuities. Slightly to moderately altered
rockmasses often exhibit smoother or slickensided
joint surfaces (e.g. though serpentinization). The
structure, ci and mi in this case are not or only
slightly affected.
With severe alteration, GSI values can be reduced
considerably. The structure becomes disturbed (e.g.
from Blocky to Very Blocky or perhaps to Sheared)
according to the alteration degree (e.g. formation of
schistose or laminated planes). Joint condition in
this case is Poor to Very poor. The intact properties

ci and mi are considerably reduced. An indication of
how alteration affects GSI is illustrated in Figure 10.
However there are cases, eg., contact metamorphism, where alteration may result in stronger rocks
(such as Quartzites, Keratites etc).
Measuring intact rock strength, ci, from altered
rockmasses is always problematic. When testing for
example schisto-serpentinites, the influence of
schistosity results in a significant reduction in the
strength of a large proportion of the specimens. As a
consequence, it is very difficult to obtain reliable ci
values from laboratory tests. For schisto-serpentinite
rocks it is suggested that the uniaxial compressive
strength should be estimated from that of the normal
serpentinite and reduced by about 30% to account
for the schistosity (Marinos et al. 2005).
Figure 9. GSI and intact strength change for weathering grade W-I to grade W-VI (grades according to ISRM 1981)
11
2.1 GSI and other classification systems

Global correlation to other classifications has been
an ongoing discussion for many years. with the pros
and cons of different approaches debated intensely.
Several decades ago, with arrival of the personal
computer, the tools for rock engineering design
started to change. Although still crude, a number of
useful numerical methods were being developed at
that time that offered the promise for being able to
undertake much more detailed analysis of difficult
surface and underground excavation problems. In
tunnelling the advent of numerical techniques to
tackle rock-support interaction and the development
of ideas associated with the ground reaction curve
concept allowed problems to be tackled sometimes
well outside the ideal range for application of the
various then available tunnel support classification
systems such as RMR, Bieniawski (1973, 1976) and
the Q system Barton et al. (1974), Barton (1976).
Experience in the early application of more sophisticated modelling suggested that there was good correspondence between guidelines from these classifications and modelling results and reality when
rockmass behaviour was relatively simple, for example for RMR values between about 3070 and
with moderate stress levels; in other words where
sliding and rotation of intact rock pieces essentially
controlled the overall failure process. The database
of experience on which the classifications had been
built and the fact that there were literally hundreds
and hundreds of kilometres of tunnels that had been
successfully constructed on the sole basis of their
application attest to this. However, consistently it
was found that the classifications or for that matter
simplistic modelling was less reliable for predicting
excavation behaviour in squeezing or swelling
ground or for cases with clearly defined structural
failures or where spalling, slabbing and/or rockbursting was feasible under very high insitu stress
conditions. It was also found difficult to apply the
then available classification systems for the design
of sequentially installed temporary reinforcement or
for establishing the support required to control progressive failure in difficult tunnelling conditions,
where for example, the excavation and installation
of support is needed to be defined for a tunnel being
advanced by a drill and blast operation involving top
heading and benching. In such a system an intricate
sequence of excavation, face support, installation of
rockbolts, steel sets and/or shotcrete, support of the
top heading while the bench is excavated, then the
completion of the lower sidewalls and the invert and
then installation of the final concrete lining (if one is
required) all require design. In such a case, defining
the proper sequence of excavation and establishing
12
the timing of support installation is very critical elements of achieving a successful tunnel design. Describing such as case, Hoek et al. (2007) state
it is unacceptable to assume that an adequate design
can be based upon a simple estimate of the final bolt
pattern and lining thickness
Needless to say though, there is in fact merit in
undertaking more than one classification of a project
rockmass, if not for any other reason than simply
from the point of view of providing a redundancy
check to ensure that ones classification estimates
are not out of line. It has long been recognized
(Carter 1992), that for example RMR is more readily
applied when only core is available than Q, as there
are elements of the joint system that are difficult to
describe from core alone, estimating J N in particular
is particularly problematic. Similarly it has long
been established that RMR76 and GSI are interchangeable in the mid-range blocky rockmass-zone,
defined in the quantitative paper, (Hoek et al. 2013)
but that RMR is difficult to apply for discrimination
between rockmasses of low general competence.
This is an area where the GSI system excels in that it
provides the descriptive controls allowing ready observational characterization that are perhaps missing
from the other systems.
Figure 10. Indicative example of how alteration (none to severe) affects the GSI.
At the upper end of the scale of rock competence

it can be problematic discriminating characteristics
of importance, for example, differentiating between
different rock types from a brittle strain bursting
perspective. Again this is an area within which the
observational approach of GSI characterization can
be useful. This is where the Q system and the RMR
system sometimes have been found by practitioners
to be difficult to apply to get the correct numbers
that excavation performance, for example, might
show. Doing more than one classification has value
in that gaps in what might be termed classification
accuracy, can be filled and a proper spectrum of
characteristics assigned for the rockmass into or upon which engineering structures are required.
3 METHODOLOGY FOR GSI APPLICATION
3.1 Understanding geological reality
Irrespective of whether one wishes to use the now
available quantitative parameter scales as an aid to
characterization or not, the very heart of the GSI
classification and what makes it different from the
other systems is that is requires a careful engineering
geology description of the rockmass. In dealing with
specific rockmasses it is suggested that the selection
of an appropriate zone location on a GSI chart not be
limited merely to checking the visual similarity with
the sketches of the structure of the rockmass as they
appear in the charts, but rather should be based on
gaining real understanding by examining actual face
exposures or outcrops. Just using the charts and even
reading the associated descriptions carefully, is no
substitute for examining conditions in situ so that
the most suitable rockmass structure is chosen. The
most appropriate case may well lie at some intermediate point between the limited number of rockmass
sketches or descriptions included around the charts.
Outcrops, excavated slopes, tunnel faces and drill
cores are the most common sources of information
for the estimation of a GSI value for a rockmass, but
these are not of equal calibre.
Outcrops are an extremely valuable source of
good observational data, particularly in the initial
stages of a project, but like any surface exposure,
they suffer from the disadvantage that relaxation and
weathering may have significantly influenced the
appearance of the exposed rockmass as compared
with what might be met deep underground; but in
the context of a surface problem, such insight may
be very valuable for comparison to core to assess
degree of weathering decomposition. Improved data
on weathering depth and influence can sometimes
be acquired (where permissible) by trial trenches or
specific drilling. Judgment is however required in
order to allow for weathering effects in assessing the
most probable GSI value applicable at the depth of
the proposed excavation.
13
Excavated slopes and exposed tunnel faces are

probably the most reliable source of information for
estimating GSI provided that these faces are related
with the project. For design of tunnels for example,
in hard strong rockmasses it is important that some
appropriate allowance be made for damage due to
mechanical excavation or blasting. As the purpose of
estimating GSI values for such situations is to assign
properties to the insitu undisturbed rockmass, failure
to allow for effects of blast damage when assessing
GSI will result in assigning values that may be too
conservative. Therefore, attempts should be made in
such circumstances to properly estimate the degree
of disturbance, and thereby define D. This problem
becomes less significant in weak and tectonically
disturbed rockmasses as excavation then is generally
carried out by more gentle mechanical means but also because the rocks are more ductile and are less
influenced by blast damage. Hence, surface damage
effects are negligible compared to the damage which
already exists in the rockmass.
Drillhole cores of reasonable diameter provide
one of the best sources for acquiring data from
depth, but one has to be cognizant that it is necessary to extrapolate the very one-dimensional information provided by the core to the threedimensional insitu rockmass. However, this is a
problem common to all borehole investigations, and
almost all experienced engineering geologists are
comfortable with this extrapolation process. Multiple boreholes and also inclined boreholes are of
great help to interpretation of rockmass characteristics at depth.
For stability analysis of a slope, the evaluation
should be based on the rockmass through which it is
anticipated that a potential failure plane could pass.
The estimation of GSI values in these cases requires
considerable judgment, particularly when the failure
plane can pass through several zones of different
quality. Mean values may not be appropriate in such
situations.
For tunnels, the index should be assessed for the
volume of rock involved in carrying loads, e.g. for
about one diameter around the tunnel in the case of
overall tunnel behaviour or more locally in the case
of an engineered structure such as an elephant foot
at the end of lattice girders or steel rib elements.
For particularly sensitive or critical structures,
such as underground powerhouse caverns, the data
and information obtained from the sources discussed
above may not be considered adequate, particularly
as the design advances beyond preliminary stages. In
these cases, the use of small exploratory tunnels can
be considered, allowing better collection of reliable
data. Remarkably, often when compared to multiple
programs of deep drillholes this type of approach to
data gathering has often been found cost effective.
4 APPLICATION WITHIN THE HOEK-BROWN

CRITERIA
Use of GSI as a basic input parameter within the GSI
system is straightforward throughout the range of
possible description, but is more problematic at the
two ends of the rock competence scale weak soft
rocks and high strength, high competence rocks.
Figure 11 shows the range of applicability of GSI
when described qualitatively and quantitatively.
In the upper quartile of the standard GSI chart
rockmass conditions can be considered near intact,
and descriptive techniques work best, so the chart
with quantitative scales should not be used. For the
weak, low competence rocks, which typically are
trending ever closer to soils, the same basic precepts
apply. In the middle, both charts work equally well.
for explicitly. In this regard even at a crude modelling scale, any such fabrics can be readily examined
as a ubiquitous joint set within the overall HoekBrown material considered appropriate for characterizing the problem.
It is reasonable to extend this argument further
and to suggest that, when dealing with large scale
rockmasses, the strength will reach a constant value
when the size of individual rock pieces is sufficiently small in relation to the overall size of the structure
being considered. This suggestion is embodied in
Figure 12, which schematically shows the transition
from an isotropic intact rock specimen, through a
highly anisotropic rockmass in which behaviour is
controlled by one or two discontinuities, to again a
fully isotropic rockmass, in this case heavily jointed.
The Hoek-Brown failure criterion, which assumes
isotropic rock and rockmass behaviour, should only
be applied to those rockmasses in which there are a
sufficient number of closely spaced discontinuities
that isotropic behaviour involving failure on these
discontinuities can be assumed. Where the block
size of the rockmass is of the same order as that of
the engineering structure being analyzed, GSI and
the Hoek-Brown criterion should not be used.
Figure 11. Zones of applicability of GSI and conventional

Hoek-Brown equations with relation to transitions to brittle
spalling criterion (top right) and (lower left) to Mohr-Coulomb
criterion (with potential in tunnelling for squeezing, (photo,
courtesy E. Hoek, from Carter et al. 2008).
4.1 Conventional applicability range

For the conventional range where rockmasses can be
thought of as blocky, the qualitative and quantitative
charts (Figures 1 and 4 respectively) both can be
used for assessing GSI. Input of GSI values into the
Hoek-Brown criterion is then straightforward, and
depending on the scale of the engineering problem
with respect to the rockmass scale (Figure 12) the
appropriate m, s and a constants can be derived.
Getting the relative scale of the problem to the
scale of the rockmass correct is the key factor in any
application of GSI into the Hoek-Brown criterion.
While there has been some criticism that GSI is only
applicable to heavily jointed rockmasses, which at
the scale of an engineering problem can be treated
as homogeneous and isotropic, this limitation can be
readily overcome with a little judgment.
For example, if there are weak discontinuities or
a pervasive weak fabric throughout a rockmass, such
as bedding or foliation, that would render the overall
rockmass anisotropic, it would be recommended that
that weak fabric be ignored in developing the global
GSI and then it should be modelled for or analyzed
14
Figure 12. Limitations on the use of GSI depending on scale

(Hoek et al, 2013)
Rather, the overall stability of the engineering

structure itself should be analyzed by considering
the behaviour of blocks and/or wedges defined by
the intersection of geological structural features.
When the slope or the underground excavation is
large and the block size small in comparison, the
rockmass can be treated as a Hoek-Brown material.
An example of this transition is frequently seen in
large open pit mines in which individual structure or
intersecting structural features control bench scale
stability, while the overall slope can sometimes be

considered as a homogeneous rockmass to which the
Hoek-Brown criterion can be applied.
These basic differences in the various scales of
structures of importance are apparent in the photograph in Figure 13 of the overall slopes of the more
than 1000m deep Chuquicamata open pit mine in
Chile. In this case many scales of structure are involved and for much of the mechanics, the rockmass
can be considered as a Hoek-Brown material.
Figure 13. Examples of different scales of fabric - stability of

individual benches clearly controlled by intersecting structural
features while the overall slope can be characterized as a HoekBrown materials on which major faults and shear zones must be
superimposed (Hoek, pers comm)
4.2 Application at ends of Rock Competence Scale

As it will be apparent from Figure 11 and previous
discussions about the charts within Figures 1 and 4,
because of their development, both GSI and the
Hoek-Brown criterion admirably characterize most
normal rockmasses from the viewpoint of their
behaviour for rock excavations. While descriptive
GSI spans the full range of rockmass characteristics,
the governing equations describing a conventional
Hoek-Brown material however run into difficulties
when applied at the two ends of the rock competence scale. This is largely because incipient
strength and block size is such that rockmass behaviour in these domains tends not to be controlled by
inter-block shear strength, but rather by material
strength. Three rockmass competence regimes can
thus be postulated as shown in Figure 11.
As also illustrated in Figure 11, the behaviour of
the rockmass is different in these three zones, and
thus it is suggested that use of the conventional
transfer equations from GSI values to the m, s and a
parameters for the Hoek-Brown criterion equations
be restricted to the central range of applicability
shown in Figure 11, i.e. above GSI=30 and below
GSI = 65. At the low end of the rock competence
scale (where ci 15MPa and GSI <30), it can be
15
taken that discontinuities play less of a role and

rockmass strength tends towards matrix strength.
Similarly, at the high end of the scale (where
GSI>65, mi>15), because discontinuities are now
widely spaced, block size becomes so significant
that once again, intact material behaviour rather than
the fracturing becomes dominant in controlling rock
mass strength.
At the high end of the scale, in competent hard
rock, particularly at great depth (e.g. 1,000m or
more), rockmass structure is often so tight that the
mass behaviour again approaches that of the intact
rock. These conditions can be modelled using the
original Hoek-Brown equation directly applied for
the intact rock (Hoek 1983) as rockmass behaviour
can now be considered that of intact rock (i.e.
equivalent GSI = 100). For example, in the case of
the Drakensberg and Ingula projects (Hoek, pers
comm), where rockmass behaviour was controlled
by intact rock strength, (since GSI for the sandstones
and mudstones was equal to 100) the throughgoing
major horizontal bedding planes could then each be
modelled explicitly.
Where the rockmass is more brittle and spalling is
an important issue, consideration needs to be given
to brittle fracture initiation which occurs at about
40% of ci. In this case the non-linear relationship
between brittle spalling initiation and confining
stress can still be represented (at least as well as any
other criterion) by the Hoek-Brown equation with
adjusted parameters. In this situation, one is towards
the top of the GSI charts and absolute GSI values
likely meaningless (Hoek, pers. comm.). Here the
failure process that controls stability of underground
excavations under these conditions is dominated by
brittle fracture initiation and propagation, which
leads often to spalling, slabbing and, rock-bursts in
extreme cases. Considerable research effort has been
devoted in recent years to the study of these brittle
fracture processes, and a specific spalling criterion
developed to characterize rockmass behaviour under
such conditions. The paper by Diederichs, et al.
(2004) provides a useful summary of much of this
work, while that by Diederichs et al. (2010) outlines
practical application methods for spalling criterion
application for tunnelling problems.
Figure 14. Normalized rockmass strength (cm/ci) as a function of rock quality (GSI), illustrating marked differences between conventional Hoek-Brown behaviour and transition functions. Spall transition threshold set at typical value,
UCS*=0.45ci) (Carter et al. 2008)
At the other (low) end of the rock competence

scale, again the rockmass structure has diverged
from a blocky fabric to something much closer to an
intact material in this case a soil. In this regime
cohesive strength becomes more important and thus
the traditional Mohr-Coulomb criterion has more
applicability. It is possible however to extend the
applicability of the Hoek-Brown criterion to cover
both these ends of the competence scale if use is
made of the two transition equations suggested by
Carter et al., 2008. These two transition relationships allow continued use of the basic Hoek-Brown
strength criterion concept out into these both end
domains. Figure 14 illustrates the behaviour of these
transition functions with respect to the conventional
Hoek-Brown and Mohr-Coulomb relationships, as a
function of GSI, while Figure 15 summarizes the
transition relationships. It should be noted that since
the initial publication of these low and high end
transition relationships by Carvalho et al. (2007) and
by Diederichs et al. (2007) respectively, several minor modifications and improvements have been
made by other authors, e.g. Castro et al. (2013).
Figure 15. Suggested Applicability Ranges and suggested
Transition and Conventional Hoek-Brown Relationships (from
Carter et al. 2008); (where UCSi = ci = Uniaxial Compressive
Strength; T = i = Tensile Strength; UCS* = crack initiation
threshold strength; and where fSP and fT(ci) are the originally
suggested Spalling and Weak rock transition functions)
5 GSI LIMITATIONS
GSI classification is based upon the assumption that
the rockmass contains a sufficiently large number of
intersecting discontinuities that it can be considered
to behave as an isotropic mass (Figures 11 and 12).
The GSI system therefore should not be applied for
rockmasses with clearly defined dominant structural
fabrics. Undisturbed slate is an example of a highly
anisotropic rockmass, where applying a GSI value
should only be done if the mode of potential failure
wont be governed by shear strength of the incipient
discontinuities within the slate. Of importance
though, in the confined conditions of a typical tunnel
situation it is possible that stress dependent effects
may provoke very different degrees of deformations
dependent on location around the periphery of the
tunnel as a result of such anisotropy. An approach to
relate such anisotropy in deformations with GSI is
described in Fortsakis et al. 2012. For rockmasses
with completely sheared structure, such as that
16
shown in the sixth (very last) row of the GSI chart in

Figure 1, anisotropy is not an issue as the difference
in strength between the rock and the discontinuities
within it will likely be small.
GSI values should thus not be indiscriminately
estimated and used in the conventional Hoek-Brown
equations for rockmasses that tend towards either
the upper or lower end of the rock competence scale.
When rocks such as marls, claystones, siltstones
and weak sandstones are deposited and remain
through diagenesis in stable environments, with little
tectonism, they can often be found as quite massive
looking rocks with few discontinuities. In such
rocks, even when bedding planes exist they often do
not appear as clearly defined discontinuity surfaces.
In such cases, the GSI chart can still be used, but
care needs to be taken in defining the discontinuity
spacing range, based on observed conditions. The
character of the discontinuities needs to be carefully,
defined, as if they exert control on blockiness and
fabric, then it is likely that rockmass quality cannot
be much better than fair (usually fair or poor) and
hence GSI values should be in the range of 4060. If
the discontinuities do not appear to exert any block
size or fabric control, such rocks form continuous
masses with very few to no discontinuities evident.
For such cases the GSI of the rockmass is essentially
100 and as such, the rockmass can be treated as an
intact rock material with engineering parameters
given directly by laboratory testing. This situation
can be handled by the low end transition function as
included in Figure 15, but with any specific discrete
discontinuities treated explicitly.
At the upper end of the rock competence range,
again care must be taken in assigning GSI values
and ensuring that these are not indiscriminately used
within the conventional Hoek-Brown equations. The
upper end transition equations allow one to deal
with the initiation of fracture in hard brittle rocks.
The conventional equations do not deal with initia-
tion of spalling failure, without modification, but

they can be used for modelling failure of already
damaged rock and hence will give an acceptable
prediction of damaged strength for hard brittle
rocks.
For numerical analysis of a rockmass containing a
single well-defined shear zone or fault or other type
of similar major geological structure, it is sometimes
appropriate to model the overall rockmass as a
HoekBrown material and superimpose the specific
discontinuity as a significantly weaker element. In
this case, the GSI value assigned to the rockmass
should ignore the single major discontinuity. The
properties of this particular discontinuity may fit the
lower portion of the GSI chart or it may require a
different approach such as laboratory shear testing
of soft clay fillings in order to model its influence.
Last, but not least, care must be taken to ensure
that adequate attention is given to defining a tensile
strength cut-off for rocks with low mi, values, or else
the Hoek-Brown envelopes will be unrealistic. This
is the range where the low-end transition equations
might be better applied. Another possibility is to use
the basic Hoek-Brown relationship but incorporating
a tension cut-off, which is easy to implement in most
modelling codes.
6 CONCLUSIONS
Use of the GSI rockmass classification system
and the associated m, s and a parameter relationships linking GSI with the Hoek-Brown failure criterion provides a proven, effective and reliable approach for strength prediction for surface and
underground excavation design and for rock support
selection for most normal rockmasses:
with intact rock material strength, ci > 510MPa,
and
with matrix yield behaviour dominated by shear
mechanisms
for (mi < 15) for the full GSI range and
for (mi > 15 ) for GSI < 65.
The observational GSI approach of making visual
assessments of rockmass conditions and assigning
appropriate quality on charts such as Figure 1 seems
to work very effectively, not just for the original
range of applicability of the Hoek-Brown criterion
for the normal range of fractured rockmasses but
also for describing rockmasses at either end of the
rock competence scale. The quantitative GSI chart
however should not be used outside the original
blocky rockmass range.
Rockmass characterization has an important role
in the future of engineering geology in extending its
usefulness, not only to define a conceptual model of
the site geology, but also for the quantification
needed for analyses to ensure that the idealization
17
(for modelling) does not misinterpret actuality (Knill

2003). If GSI rockmass characterization is carried
out in conjunction with careful numerical modelling,
together they can provide more insight on ground
behaviour than either approach would on its own
(Chandler et al. 2004). Use of GSI has considerable
potential in rock engineering because it permits the
multiplicity of fabrics within a given rockmass to be
quantified, thereby enhancing geological logic and
reducing engineering uncertainty. The use of the GSI
system allows the influence of variables, which
make up a rockmass, to be assessed and hence the
behaviour of the rockmass to be explained more
clearly.
One of the key advantages of the index is that the
geological reasoning it embodies allows adjustments
to be made of its ratings to cover a wide range of
rockmasses and conditions, but also allowing some
understanding to be gained on applicability limits.
Having become familiar with the GSI system most
people have no difficulty in using it for describing
the complete range of rock characteristics including
better and worse rockmasses than the blocky midrange for which the original Hoek-Brown criterion
was initially formulated.
The qualitative GSI chart provides an intelligent
tool for the user to rapidly assess the numerous key
combinations of rockmass characteristics that truly
influence the strength and deformation properties
that might control design of any structure on or within that rockmass.
As a system, visual GSI characterization linked
with Hoek-Brown strength determination as a basis
for modelling has been widely adopted by engineers
and geologists involved in design and construction
of structures in or on rock. One reason is that there
are very few practical alternatives. There are quite a
number of rockmass classification systems and rules
of thumb that can be applied for estimating tunnel
support, with several of them developed specifically
to provide first order estimates of final tunnel rock
support, foundation bearing capacity or rockslide
risks. Some of these classifications have been
pressed into service to provide estimates of rockmass strength and deformation modulus but that was
never the original aim of their developers. The GSI
approach has been solely developed for parameter
estimation, and this is the key other reason that its
use for input into the Hoek-Brown failure criteria,
both empirical in concept, have been proved to work
so well. Analytic or practical back-analyses of
rockmasses where it has been applied are continuously certifying this.
In spite of criticisms of the qualitative nature of
the original GSI classification, it appears to work
well within the hands of experienced observers.
With the advent of a parallel, more quantified chart,
it is hoped that those who felt uncomfortable with
the solely qualitative visual approach will be able to
more readily ascribe appropriate GSIs for their projects. There is no question than GSI should only be
expressed as a range, as rockmasses by their very nature are variable. As such, the expectation must be
that GSI characterization will produce approximate
answers only, but when these answers are refined as
a project typically progresses through its various
stages the final solutions are generally credible and
robust. We are still a long way perhaps from being
able to completely define a rockmass, but we have
come a long way towards this goal, as expressed
several years ago by Hoek My long term
hope is that numerical tools such as the Synthetic
Rock Mass and its off-shoots will eventually enable
us to replace classification type approaches or at
least to calibrate these classifications. It may be a
while before these hopes can be realized.
ACKNOWLEDGEMENTS
Thanks are due to Evert Hoek and Paul Marinos for
their insight in initially developing the GSI approach
and to Mark Diederichs and others for their critiques
over recent years of its applicability over the full
range of the rock competence scale.
6.1 References
ANON. 1995. The description and classification of weathered
rocks for engineering purposes. Geological Society Engineering. Group Working Party Report. Quarterly Journal of
Engineering Geology, 28, 207-242.
Barton N.R, Lien R, Lunde J. 1974. Engineering classification
of rockmasses for the design of tunnel support. In: Rock
Mech. 6 (4), pp 189-239.
Barton, N., 1976 Recent Experiences with the Q-System of
Tunnel Support Design. Proc.Symposium on Exploration
for Rock Engineering, pp. 107-117
Bieniawski Z.T. 1973. Engineering classification of jointed
rockmasses. In: Trans S. Afr. Inst. Civ. Engrs 15, pp 335344.
Bieniawski, Z.T. 1976. Rock mass classification in rock engineering, exploration for rock engineering. Z.T. Bieniawski
(ed), A.A. Balkema, Johannesburg: 97106.
Cai M., Kaiser P.K.,Uno H.,Tasaka Y., Minami M. 2004. Estimation of rockmass strength and deformation modulus of
jointed hard rockmasses using the GSI system. In: Int. Jour.
Rock Mechanics and Mining Sciences. 41, 1, pp 3-19.
Carter, T.G., 1992 Prediction and Uncertainties in Geological
Engineering and Rock Mass Characterization Assessment.
Proc. 4th Italian Rock Mechanics Conference, Torino. pp.
1.1 - 1.22.
Carter, T.G., Diederichs, M.S., Carvalho, J.L. 2008. Application of modified Hoek-Brown transition relationships for
assessing strength and post yield behaviour at both ends of
the rock competence scale. In Proc. The 6th International
Symposium on Ground Support in Mining and Civil Engineering Construction, 30 March 3 April 2008. Cape
Town, South Africa, pp.3759. Journal of the Southern African Institute of Mining and Metallurgy, Vol. 108: pp 325338.
18
Carvalho, J.L., Carter, T.G., Diederichs, M.S. 2007. An approach for prediction of strength and post yield behavior for
rock masses of low intact strength. In Proceedings 1st CanUS Rock Mechanics Symposium, Meeting Societys Challenges & Demands, Vancouver, Canada: 249257.
Castro, L.A.M., Carvalho, J., G. S. 2013. Discussion on how
to classify and estimate strength of weak rock masses. In
P.M. Dight (ed), Slope Stability 2013. Perth: Australian
Centre for Geomechanics.
Chandler R. J., de Freitas M. H. and Marinos P. 2004. Geotechnical Characterisation of Soils and Rocks: a Geological
Perspective, Proceedings of the Skempton Conference:
Advances in Geotechnical Engineering, Thomas Telford,
London, 1, 67-102.
Dearman, W. R. 1976. Weathering classification in the characterization of rock: A revision: Bulletin International Association of Engineering Geologists, Vol. 13, pp. 373381.
Deere D.U. 1964. Technical description of rock cores for engineering purposes. In: Rock Mechanics and Engineering Geology.1, 1, pp 17-22.
Diederichs M.S, Kaiser P.K, Eberhardt E. 2004. Damage initiation and propagation in hard rock during tunnelling and the
influence of near-face stress rotation. In: Int. Jour. Rock
Mechanics, Mining Science.5, 41, pp 785-812.
Diederichs, M.S. (2007). Mechanistic Validation and Practical
Application of Damage and Spalling Prediction Criteria for
Deep Tunnelling. The 2003 Canadian Geotechnical Colloquium. Canadian Geotechnical Journal. Vol.44:9 pp.10821116.
Diederichs, M.S, Carvalho, J.L., & Carter, T.G. (2007). A
modified approach for prediction of strength and post yield
behaviour for high GSI rockmasses in strong, brittle ground.
Proc. 1st Can-US Rock Symposium. Meeting Society's
Challenges and Demands. June, Vancouver. pp.277-28
Fortsakis P., Nikas K., Marinos V., Marinos P.. 2012. Anisotropic behaviour of stratified rockmasses in tun-nelling. Engineering Geology, ,Volumes 141142, 19, pp. 7483.
Hoek, E. (1982). Geotechnical Considerations in Tunnel Design
and Contract Preparation. Trans. lnstn. Min. Metall.
(Sect.A:Min.industry) 91, A101-9.
Hoek E. 1994. Strength of rock and rockmasses. In: News
journal of the International Society of Rock Mechanic, 2, 2,
pp 4-16
Hoek, E (1999) Putting numbers to geology an engineer's
viewpoint. Q. Jnl Eng.Geol. vol. 32, 1, pp. 1-19(19)
Hoek, E. and Brown, E. T. (1980), Empirical strength criterion
for rock masses. Journal of the Geotechnical Engineering
Division, ASCE v. 106, n. GT9, p. 1013-1035.
Hoek E, Brown E. T. 1997. Practical estimates of rockmass
strength. In: Int. Jour. Rock Mechanics, Mining Sciences
and Geomechanics Abstracts 34, pp 1165-1186
Hoek E, Diederichs M.S. 2006. Empirical estimation of rockmass modulus. In: International Journal of Rock Mechanics
and Mining Sciences, 43, pp 203-215.
Hoek, E., Marinos, P. 2000. Predicting tunnel squeezing problems in weak heterogeneous rock masses. In Tunnels and
Tunnelling International, Part 1 November 2000, Part 2,
December 2000: 21 p.
Hoek E, Marinos P. 2007. A brief History of the Development
of the Hoek-Brown Failure Criterion. Soils and Rocks, So
Paulo, 30(2): pp 85-92, May-August, 2007.
Hoek E., Martin C. D. 2014. Fracture initiation and propagation
in intact rock. Submitted to the Journal of Rock Mechanics
and Geotechnical Engineering.
Hoek E, Caranza-Torres C.T, Corcum B. 2002. Hoek-Brown
failure criterion - 2002 edition. In: Bawden, H.R.W., Curran, J., Telesnicki M., (Eds). Proc. North American Rock
Mechanics Society (NARMS-TAC 2002). Mining Innovation and Technology, Toronto, Canada, pp 267-273
Hoek, E., Kaiser, P.K. and Bawden. W.F. 1995. Support of underground excavations in hard rock.Rotterdam: Balkema.
Hoek E, Marinos P, Benissi M. 1998. Applicability of the Geological Strength Index (GSI) classification for weak and
sheared rockmasses The case of the Athens Schist formation. In: Bull. Eng. Geol. Env. 57, 2, pp 151-160.
Hoek, E., Carter, T.G., Diederichs, M.S. 2013. Quantification
of the Geological Strength Index chart. 47th US Rock Mechanics / Geomechanics Symposium, San Francisco: ARMA 13-672.
ISRM. 1981. Rock characterization, testing and monitoring
ISRM suggested methods. In Brown E.T. (ed), International Society of Rock Mechanics, Pergamon, Oxford.
ISRM. 1985. Suggested method for determining point load
strength. In: Int. J. Rock Mech. Min. Sci. and Geomech.
Abstr. 22, pp 51-62.
Lauffer H. 1958. Gebirgsklassifizierung fur den Stollenbau.
Geol. Bauwesen 74, pp. 46-51..
Marinos P, Hoek E 2000. GSI: A geologically friendly tool for
rockmass strength estimation. In: Proc. GeoEng2000 at the
Int. Conf. on Geotechnical and Geological Engineering,
Melbourne, Technomic publishers, Lancaster, Pennsylvania,
pp 1422-1446
Marinos P, Hoek E 2001. Estimating the geotechnical properties of heterogeneous rockmasses such as flysch. In: Bull.
Eng. Geol. Env., 60, pp 82-92.
Marinos P., Marinos V., Hoek E. 2007 Geological Strength
Index (GSI). A characterization tool for as-sessing engineering properties for rockmasses. Proceedings of the Intrnational Workshop on Rockmass Classification in Underground Mining held in 1st Canada-US Rock mechanics
symposium, Also permitted to be published in: Underground
works under special conditions, Romana, Perucho & Olalla
(eds) Taylor and Francis publ. 13-21, Lisbon.
Marinos V 2007..Geotechnical classification and engineering
geological behaviour of weak and complex rockmasses in
tunneling, Doctoral thesis, School of Civil Engineering, Geotechnical Engineering Department, National Technical
University of Athens (NTUA), Athens. (In greek)
19
Marinos V., Marinos P., Hoek E. 2005. The Geological

Strength Index Applications and limitations. Bulletin of
Engineering Geology and the Environment, , 64/1, 55-65.
Marinos V., Fortsakis P., Prountzopoulos G. 2011. Estimation
of geotechnical properties and classification of geotechnical
behaviour in tunnelling for flysch rockmasses. Proceedings
of the 15th European Confer-ence on Soil Mechanics and
Geotechnical Engineering, A. Anagnostopoulos et al. (Eds.)
Part 1, pp. 435-440, Athens.
Palmstrm A. (2000): Recent developments in rock support estimates by the RMi. Journal of Rock mechanics and tunneling technology, Vol.6, pp.1-9
Palmstrm A., (2005) Measurements of and correlations between block size and rock quality designation. Tunnelling &
Undergr Space Tech; 20(4): pp362377
Russo G. (2007): Improving the reliability of GSI estimation:
the integrated GSI-RMi system. Proc. I.S.R.M. Workshop
"Underground Works under Special Conditions", Madrid,
pp 123-130
Russo, G. 2009. "A new rational method for calculating the GSI
". Tunnelling and Underground Space Technology n.24,
pp.103-111
Sonmez, H. and Ulusay, R., (1999), Modifications to the geological strength index (GSI) and their applicability to stability of slopes, International Journal of Rock Mechanics &
Mining Sciences, v. 36, p. 74360.
Stacey, T.R., Page, C.H. 1986. Practical Handbook for Underground Rock Mechanics. Trans Tech. Publications,
Clausthal-Zellerfeld publ.
Terzaghi, K. 1946. Rock defects and load on tunnel supports,
Introduction to rock tunnelling with steel supports, a book
by Proctor, R.V. and White, T.L. Commercial Shearing &
Stamping Co., Youngtown, Ohio, U.S.A.
Tzamos, S. and Sofianos, A. I., (2007), A correlation of four
rock mass classification systems through their fabric indices.
Intern. Journal of Rock Mechanics and Mining Sciences 44
pp.477495

Use of Gsi For Rock Engineering Design

Uploaded by

Copyright:

Available Formats

Use of Gsi For Rock Engineering Design

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Use of Gsi For Rock Engineering Design

Uploaded by

Copyright:

Available Formats

What are some methods used to define rockmass properties for design?

What are some methods used to define rockmass properties for design?

What are some limitations of using laboratory and in situ testing to define rockmass properties?

What are some limitations of using laboratory and in situ testing to define rockmass properties?

Use of GSI for Rock Engineering Design

complications in representativeness may also arise

well-known and more commonly applied rockmass

Figure 1. Basic GSI Chart for Visual Geologic

2 THE GEOLOGICAL STRENGTH INDEX

Fine grained polyminerallic igneous crystalline rocks,

Coarse grained polyminerallic igneous and

Intact rock samples

Arenaceous rocks with strong crystals and

Empirical failure criterion

Lithified argillaceous rocks,

It specifically was set up to avoid including other

Table 1. Original GSI tabulations for m and s for various

Kaiser and Bawden (1995). In place of disturbed

the rockmass as the basis for ascribing a GSI value.

Figure 2. Difficult Ground Conditions for undertaking

An extension of the original GSI application

Figure 3. Data entry stream for using the Hoek-Brown system

2.2 Observational characterization

2.3 Quantitative characterization

Figure 4. Quantification of GSI by Joint Condition, as graded

The chart in Figure 4 was also introduced because

Table 2. Guidelines for the selection of the Disturbance

In examining these transfer equations between

should be noted these equations have been set up to

Approximate values of the ratio between ci and ti

Under such circumstances estimates of the value

strength in areas of the rockmass with limited to no

Figure 6. Hoek-Brown typical Intact Strength envelopes for a

Experience has shown that there is a common

This table can be used for preliminary estimating,

of mi can also differ somewhat depending on the

such as frequent bedding or schistosity planes, often

2.2 GSI and intact rock strength

Figure 7. Indicative example of how tectonism (from low to

2.4 GSI, weathering and alteration

Alteration is an effect brought about by generally

However, the actual changes that will occur in a

Figure 8. Indicative example of how the weathering degree

Most weathering and alteration processes create

this case is Poor to Very poor. The intact properties

2.1 GSI and other classification systems

At the upper end of the scale of rock competence

Excavated slopes and exposed tunnel faces are

4 APPLICATION WITHIN THE HOEK-BROWN

Figure 11. Zones of applicability of GSI and conventional

4.1 Conventional applicability range

Figure 12. Limitations on the use of GSI depending on scale

Rather, the overall stability of the engineering

stability, while the overall slope can sometimes be

Figure 13. Examples of different scales of fabric - stability of

4.2 Application at ends of Rock Competence Scale

taken that discontinuities play less of a role and

At the other (low) end of the rock competence