Rock Mechanics and Rock Engineering

https://doi.org/10.1007/s00603-020-02177-1

ORIGINAL PAPER

Putting Geological Focus Back into Rock Engineering Design

Trevor G. Carter1 · Vassilis Marinos2 

Received: 26 May 2018 / Accepted: 12 June 2020

© Springer-Verlag GmbH Austria, part of Springer Nature 2020

Abstract
The trend today to ever increasing modelling sophistication demands that much more attention be paid by practitioners to
achieving better appreciation and characterization of geology and rockmass variability, so that rock–structure interaction
effects can be analysed more realistically in better calibrated models. This paper is thus directed towards focussing attention
on risk-based geological characterization as basis for helping modellers and designers improve calibration of their models.
A sequential approach of appropriate input parameter refinement is outlined as a path forward methodology for consistently
achieving maximum reliability in modelling. Processes that are needed for identifying key controlling geological structural
features and rockmass domain characteristics that may be critical influences on true rockmass behaviour are explored so
that rationalization steps can be followed in model building to ensure that actual behaviour drivers are not only properly
represented, but are reliably characterized through rigorous calibration. Suggestions for the use of the observational and
quantitative GSI charts at various scales appropriate to specific geological domains are presented as a means for achieving
such calibration. Illustration is then given of how quantification can be achieved of rock quality variability throughout the
complete range of rock competence, from intact pseudo-homogeneous high strength rockmasses subject to brittle spalling,
through blocky, folded or foliated rockmasses, where kinematic controls are typically of paramount importance, through to
completely degraded, fault process core zones and saprolites, where material matrix strength almost entirely dominates behav-
iour. Guidelines are given for suggested ranges of classification applicability for use with the Hoek–Brown failure criterion.

Keywords  GSI · Rock mass characterization · Hoek–Brown failure criterion · Geotechnical parameter definition · Rock
engineering design · Numerical modelling calibration · Geological structural domaining · Q · RMR · RMi

1 Introduction: Risk Appreciation interest and attention directed towards reliably improving the
definition of key input parameters needed to ensure adequate
In recent years, in rock engineering, significant advances, calibration of these increasingly sophisticated models. In
mainly in numerical modelling capability have occurred fact, worrying trends have been observed across the industry
worldwide. Modelling codes now exist that can not only (as highlighted by several Journal Editors, one of whom, to
afford better and more advanced insight into rock-support which this paper is posthumously dedicated, clearly fore-
interaction and rockmass progressive failure processes, but saw the many pitfalls created on projects through his review
they are now capable enough to allow synthetic rock masses of submitted journal articles). Prof. Dr. Giovanni Barla on
to be efficiently built so that design layouts can be more real- many occasions expressed concern over the fact that authors
istically evaluated. Unfortunately, there has not been equal of many such submitted articles showed what he considered
a completely delusionary sense of understanding and accom-
plishment in their submissions. He found the fact that most
* Vassilis Marinos often the worst cases were papers submitted for publication
[email protected]
that were written about completing some complex modelling
Trevor G. Carter that was then used as the basis for some project’s design, but
[email protected]; [email protected]
without actually ever benchmarking the models with real
1
Golder Associates, Toronto, ON, Canada data (Barla et al. 2013). On several occasions in recent years,
2
School of Civil Engineering ‑ Geotechnical Dept.,
right up to the 2019 ARMA and ISRM conferences he con-
National Technical University of Athens (NTUA), 9, Iroon tinued to express concern that, despite this worrisome trend
Polytechniou Street, 15780 Athens, Greece

13
Vol.:(0123456789)
 T. G. Carter, V. Marinos

having been discussed and highlighted many times in Con- have been coalesced alongside one another to better high-
ference Keynotes by competent practitioners, and despite the light the fundamental delusion that is at the heart of this
fact that a number of similar minded journals to the RMRE problem, that increased sophistication in modelling with
had completely stopped publishing articles, where model- no improvement in base data markedly increases the reli-
ling submissions were made without proper reliable bench- ability of a design. In reality, as shown in the left diagram
marking, this troublesome trend continued. He was in fact only a marginal decrease in risk can actually be achieved
becoming even more concerned in recent years that the trend by increased analytical sophistication, even when properly
was not reversing, largely because modelling codes were calibrated. With uncalibrated modelling, more sophistica-
becoming ever easier to use, and much more graphically tion plausibly might even increase risk levels due to serious
impressive. Rather, he felt the opposite was occurring, that misrepresentation in the models of realistic actual rockmass
this worrying trend (of totally uncalibrated modelling being behaviour as a result of incorrect modelling of the true con-
thought reliable enough alone for use for real project design) trolling failure mechanisms existing at the site.
was actually becoming more the norm than modelling within In fact, delusionary misrepresentation of stable conditions
a well calibrated framework of real field data. impressively presented in modelling output, may further
The fact that recent practitioner feedback on these top- increase risks of unexpected failures, a spate of which have
ics, and up-to-date guidance documents to help modellers occurred in recent years right across all industry sectors—
focus on how they should benchmark their models are also ranging from tunnel and crown pillar collapses, through to
largely lacking in the literature (particularly in the pages of large open pit failures, many of which have caught public
the numerical journals), he felt somewhat contributed to the and press attention, but have never been reported in detail
current widespread lack of proper appreciation on how mod- in the technical literature. However, as discussed in Carter
ellers should move ahead to overcome this problem. Hence, (2014, 2018), a number of the most spectacular of these fail-
this paper, and the inclusion below in Fig. 1 of an updated ures have occurred due to lack of an adequate appreciation of
version of two 25-year-old diagrams. structural geological controls. Modelling misrepresentation
in a number of these cases is also known to have contributed
1.1 Risk Implications of Over‑Sophistication to unexpectedness of the failures, due to the lack of appre-
in Uncalibrated Modelling ciation through the modelling results of correct rockmass
behaviour, largely because of a worrying insufficiency of
In Fig. 1, the two original diagrams taken from a MIR con- attention to structural features and fabric of the rockmass
ference paper, presented again at the behest of Dr. Barla, in a number of these cases, but also because of inclusion

Fig. 1  Distinction in “Real” Project Risk Reduction achieved by (i) limited risk reduction achieved through increased design sophistica-
increased understanding of fatal flaws, geological anomalies and tion. Base diagrams from Carter (1992), Figs.  1 and 2. Note: blue
parameter uncertainty (through acquisition of more pre-construction shaded zone, which is equivalent on both diagrams, shows approxi-
geological field information—right diagram) vs (ii) left diagram— mate reduction in risk at various stages of Project Implementation

13
Putting Geological Focus Back into Rock Engineering Design

of inappropriate strength and deformation parameters and, projects, financiers may overrule the technical folk, and
principally for underground designs, erroneous boundary make decisions based purely on comparison of cost esti-
stress assumptions. mates, not realizing or not being adequately informed that
It is fundamental that the industry (and particularly mod- undertaking a design for an engineered structure or a project
elling practitioners) understand that real risk can only be in rock or on rock without adequate investigation data is
decreased by increasing real understanding through addi- foolish at best, and downright dangerous at worst.
tional data, so that models and analysis can be properly The remainder of this paper, therefore, attempts to pro-
representative of actual conditions and most probable rock- vide explanatory background based on insight from case
mass behaviour. The relational risk reduction, achieved by studies that hopefully will help increase awareness of the
increasing design sophistication, is illustrated by the blue need for better and more in-depth investigations as a stand-
zone included in the right-hand diagram of Fig. 1. This blue ard across the industry. The hope is that as a result, not only
zone plots the degree of risk reduction achievable through will uncalibrated modelling become seen as technically
good calibrated modelling per the left diagram in a relative unacceptable, industry-wide, but more importantly, that
sense compared to total project risk reduction achievable decision makers see project design submissions at Feasibil-
with additional information acquisition, as benchmarked to ity and Detailed Design stages made without an adequate
real data, per the right diagram. degree of real data acquisition as also unacceptable as bank-
With more user-friendly interfaces being incorporated able documents for project financing.
into modelling software, it is becoming ever easier to be The following sections of this paper summarize case
able to carry out quite sophisticated modelling, often with- study, experience-based, rationale as to what in rock engi-
out realizing deficiencies in the generated model. This ease, neering is needed on a project basis for achieving cost-effec-
together with the proliferation of readily available “pull- tive, geotechnically de-risked designs, so that construction
down” menus of instantly accessible parameters, is increas- implementation does not run into unforeseen delays and
ingly leading to modelling being undertaken without suffi- difficulties. Step-by-step guidance is also provided to help
cient calibration to real data. Proper and reliable calibration practitioners not only to take all the appropriate steps needed
to real data on material parameters and on geological and to proceed efficiently through their design, but also so that
geometrical variability is critical for verifying any design, perhaps they can utilize this paper to help them convince
particularly when that design is heavily based on numerical financiers and decision makers not familiar with geotechni-
modelling. cal risk and uncertainty why thorough in-depth understand-
This brings us to question why in the design stages of ing of rockmass fabric (through often expensive site investi-
many projects today, even major ones, these trends away gation effort) is essential for de-risking any rock engineering
from perceiving the need to calibrate models with real physi- project. Non-engineering financiers and decision makers
cal data, have developed. Unfortunately, one of the key rea- need to understand that expenditure on site investigation is
sons is purely financial, based on the lack of understanding not an impediment to project design but an essential baseline
of project authority decision makers, who may well not be component.
familiar with the real uncertainties of geotechnical engineer-
ing, whereby they perceive design through modelling (even
very sophisticated modelling) as being much cheaper and 1.2 Cost‑Effective Advances in Site Investigation
certainly much faster than design through undertaking addi- Approaches
tional site investigations. This is an unrealistic equivalence
that is commonplace and needs to be dispelled by wider Most of what has been said in the previous section relates to
education through the technical press of the consequences of advances that have been made in modern numerical model-
designing with inappropriate attention to real data acquisi- ling—particularly with respect to ease of applications and
tion through adequate site investigation. increased sophistication in display output. Hopefully this
Experienced project managers and decision makers who has not given the impression that progress has only been
have been exposed to real geotechnical risk implications made in computational aspects of design evaluation. This is
through having to deal with the legal and financial ramifica- not the case. There have been equally major advances made
tions of collapses and ground condition claims and major in the sophistication that is now available for data acquisi-
construction delays, truly understand that these approaches tion of key information about rockmass conditions. Consid-
are not equivalent. These practitioners appreciate that design erable improvements have been made in site investigation
must be based on real data, and will sanction the neces- methods, and particularly in equipment, with the advent of
sary expenditures, as they inherently understand, Fig. 1, the televiewer, arguably being of equally stellar importance
even though they may not have a geotechnical or geological to investigation data acquisition, as development of parallel
or rock engineering background. Unfortunately, on many processors and advanced graphics have been to computing.

13
 T. G. Carter, V. Marinos

While downhole acoustic and optical televiewer log- Unfortunately, use of straightforward mapping and
ging tools have seen some measure of widespread industry descriptive procedures have rather fallen into neglect, largely
acceptance as a viable means to more easily acquire rock- because more and more folk are moving towards numeric
mass fracture fabric data from site investigation boreholes codification of acquired field information. Fortuitously, like
than core alone, these tools are typically still only being modelling, numerical data acquisition is becoming every day
utilized in high end investigations. Rather, the industry, easier to accomplish with constant improvements in elec-
worldwide, has trended away from the perceived drudgery tronics. As a result, routinely, field compilation of factual
of field investigation, core logging and intensive laboratory information is nowadays made directly into custom-writ-
testing, toward increased use of tablets for numerical, rather ten applications, running on tablets or smart phones, with
than graphical logging with more and more use of database- a trend towards recording raw mapping and core logging
related pull-down dialogs of readily available parameter list- information simply as codified data, solely utilizing numeric
ings useful for subsequent direct modelling input. values. Gone generally in tunnelling and drift mapping are
Conceptually, this should be leading to increased effi- the extremely insightful hand-drawn sketches and plans that
ciency in conducting geotechnical evaluations, but in reality, used to be commonplace, such as shown in Fig. 2, plausibly
it is resulting in a decreased level of geological and geotech- resulting in less understanding being gained of site condi-
nical understanding of what is actually being investigated. It tions, by even the actual personnel undertaking the logging.
is not just the raw site investigation aspects that are suffering Together with all the complications of inaccurate geome-
from lack of adequate appreciation of basic geology; the chanical modelling, as previously discussed, reviews of fail-
merits of descriptive and graphical/mapping approaches as ure cases points to a diminished understanding nowadays of
a means to improve “big picture” geological understanding “big picture” site structural geology and reduced apprecia-
seems also to have been largely forgotten. tion of actually what any specific borehole had intersected.

Fig. 2  Typical geological detail included in drift mapping pre-numeric rockmass classifications (Extract from 1963 Geology Level Plan Cour-
tesy Goldcorp Inc., as presented in Carter (2015)

13
Putting Geological Focus Back into Rock Engineering Design

Rather, today, we see field programs collecting and generat- all subsequent mining. This is in turn leading to these mines
ing reams of raw data, with little immediate overview (in creating good, comprehensive, insightful, well calibrated 3D
the field) of what the data actually means. Although core numerical models as basis for mine design, with geo-engi-
photographs are still mostly being properly taken as core neers getting focussed full-circle back to having to under-
is recovered, increasingly many of the purely descriptive stand the structural geology of their sites and to collecting
aspects of core logging have been eliminated in favour of real structural and rockmass data pertinent to understanding
more direct numerical data acquisition and statistical sum- their actual rockmass conditions. Heavy use of televiewer
marization approaches. technology and other latest technology research tools for site
Arguably, these approaches are achieving marked effi- investigation data acquisition are accordingly being pushed
ciency improvements in data acquisition, but unfortunately, to new limits in attempts to speed up data collection and
yet again, like the increased sophistication in modelling, minimize the costs of the required investigations so that ade-
these major computational improvements are potentially quate data can be collected more efficiently. This rejuvenated
diminishing real understanding of what the collected site focus on collecting understandable structural geology data
investigation results physically mean. from core and from underground drift excavation mapping
Taken together with the model calibration issues dis- has also gone full-circle back to producing “old-fashioned”
cussed at the beginning of this paper, it is clear that trends conventional geological core logs and level plan maps and
over the last several decades while improving our ability sections (like that shown in Fig. 2), rather than just col-
to analyze rock engineering problems have unfortunately lecting reams of numerical classification data based on face
also diminished total understanding of real rock–structure inspections per blast or TBM push and logging solely Q
interaction issues. Symptomatic of this problem is the situ- values, not geology.
ation that has existed in the mining industry for the last few
decades, where it has been commonplace for mines to have 1.3 Detrimental Influence of Rockmass
been designed purely on the basis of geostatistics assess- Classifications
ments of block models of ore grade variability without an
overall structural geology understanding of the country rock While classification data collection and characterization of
within which the mine needs to be built. As a result, unex- rockmasses in terms of the well-known classification sys-
pected dilution and other operational problems have not tems (Q, RMR, GSI, RMi) yields significant useful informa-
infrequently been encountered, with, in some cases, viable tion that helps in moving a rock engineering project forwards
minable resources not being able to be extracted due to geo- in terms of collecting appropriate data needed for design,
technical reasons. For the more advanced mining companies, such systems should not be the sole basis for data collec-
particularly those transitioning to large scale block and panel tion. Heavy use of these geotechnical rockmass classifica-
cave-mining approaches, this previous industry trend is for- tion systems, without gaining a proper appreciation of the
tunately changing away from heavy sole reliance on geo- controlling structural geology, may in fact be detrimental to
statistical block modelling for mine layout design, back to developing a thorough understanding of intrinsic rockmass
concentrating on getting good structural models completed behaviour, based on the true fabric variability.
at an early stage to guide both exploration and operational It is common practice to utilize these systems in almost
mining aspects. all aspects of site investigation assessment of rockmass
Experience with cave-mining development has in fact conditions on any project nowadays. Sometimes these clas-
driven this trend and focussed attention on the need to sification systems are applied singly, sometimes multiple
understand geological structure much better than in recent approaches are applied. This latter course of action is pre-
decades, as structure has been found to critically control ferred, as, through duplication, specific problems of the
mining efficiency. It has become evident to the financiers and individual systems may be effectively compensated by use
decision makers in these situations that lack of understand- of one or other of the alternative systems (Carter 2010).
ing of geological structure has the potential to completely However, application of such classifications, alone, without
compromise the mine’s ability to achieve cost effectiveness parallel structural mapping of the same rockmass, unfortu-
in the whole cave-mining production operation. In conse- nately, rather than improving understanding, have, instead,
quence, serious effort and funding is again being directed tended to increasingly move practitioners away from true
into field geotechnical data collection. The two problems geological appreciation of the rockmass conditions that they
alluded to above—inappropriate uncalibrated modelling and are having to deal with, and more move them towards just
inadequate understanding of site conditions through inade- having made a numerical codification of conditions.
quate underfunding investigation effort, are thus both having This trend has unfortunately rather gone against the hope
to be tackled by the leading mining companies, because this of many of the original proponents of the systems, includ-
expenditure now is cost-efficient in the long run to de-risk ing various major organizations, such as the USACE, when

13
 T. G. Carter, V. Marinos

originally recommending the RMR (Bieniawski 1973/1976, that significant knowledge gaps exist in design understand-
1989) and the Q systems (Barton et al. 1974; Barton 1976) ing, even for major projects. Perhaps as a direct conse-
for rock mass quality ranking for design guidance (Murphy quence, failures continue to be evident across the industry,
1985). despite significant computer data processing and numerical
This trend has also detrimentally further reduced descrip- modelling effort frequently having been expended during
tive and diagrammatic representation of “big picture” geo- project designs.
logical characteristics. Preparation, for example, of longitu- Unexpected collapses of urban tunnels and/or mine
dinal tunnel geological sections and underground geological crown pillars, and failures of high slope cuts at large open
level plans, like the example shown in Fig. 2, is rare today, pit mines, are cases that catch media attention, Carter (2014,
both in the mining and civil industries. All too often on tun- 2018), Brox (2020). Lack of appreciation of “big picture”
nelling projects nowadays, longitudinal summary geology geological structure, oftentimes can be identified as causa-
drawings have been replaced by an Excel spreadsheet listing tive. However, conditions leading to such failures, can only
of rockmass classification values, based solely on individual be considered truly unexpected if extreme geological vari-
face maps acquired for each blast round advance or TBM ability makes prediction from collected facts inconclusive,
push. Sometimes, even the face maps are missing and only Carter (2018). The message is clear—detailed design should
a value for Q or RMR or GSI (Hoek 1994; Marinos and never be undertaken based solely on numerical rockmass
Hoek 2000; Hoek and Brown 2018) is recorded, represent- classifications without good structural geological domain-
ing the rock conditions. This really is not a satisfactory state ing, even when Q’s and/or RMR’s have been generated from
of affairs. quite reasonable field assessments.
This lack of detailed tunnel maps and sections or mining
underground drift and level plans that used to show intri-
cate details of geological structure and characteristics, is a 2 Application of Domain‑Based
worrying problem, as such plans, which were commonplace Classifications for Improved Design
prior to the advent of CAD drafting, were effective for com-
pilation and for allowing proper assimilation of all geologi- For the majority of rock engineering projects, it is rare that
cal information of significance. Such maps, which could be geological conditions are ever uniform. Accordingly, key
viewed and generally understood by all involved personnel, design information on rockmass characteristics is typically
showed all key information (as interpreted from the drift needed to be collected for more than one structural domain
mapping and the myriad of underground drill holes executed and often also for various different rock types even within
from the drift). Inspection of the interpreted geology shown one domain. Unfortunately, it is commonplace, particularly
on these types of routinely updated compilation drawings in early design stages, that insufficient emphasis is placed
was characteristically used every day to update everyone on adequately characterizing geological variability across a
on conditions encountered. As such, these maps provided a project site area in sufficient detail to help identify important
higher level of understanding to the underground geologists risk possibilities. Consequently, key perception of variability
and ground control engineers back then, than is the case in rock types and/or in rockmass fabric (that might exert
nowadays in many mining or tunnelling operations. Today, significant influence on constructability or might even lead
all too often, all that can be referenced are spreadsheet to failure), can be missed. As early as possible in project
numeric summaries which give no clue of potential prob- development, plotting of areal or better still volumetric vari-
lem situations unless plotted somehow (Muir-Wood 2000). ability in rock quality (even defined just as RQD, but better
Along with this diminished appreciation of what should be as GSI, Q or RMR) can help identify the presence of adverse
“known geology”, identification then of what is anomalous structural features that may need additional investigation, or
becomes increasingly more difficult, thereby further increas- at least additional characterization to ensure implementation
ing risk of encountering unknown, unexpected conditions. of an adequate design.

1.4 Missing “Big Picture” Geology Appreciation 2.1 Principles of Structural Domaining

Curiously, this trend to more data acquisition but with less Characterization of project area rockmasses should always
“big picture” appreciation of overall geological structure is be two pronged—domain scale and engineering structure
growing. This is problematic, since for effective rock engi- scale; as rockmass characteristics are typically never the
neering design, reliable understanding of broad scale geol- same over the entire width of a typical project site (Barnett
ogy is essential, such that strength and deformation charac- and Carter 2020). Characteristically, multiple geological
teristics can be well defined, where engineering structures domains exist, each with different engineering geologi-
are to be created. It is an unfortunate reality nowadays also, cal features and each exhibiting a unique set of definable

13
Putting Geological Focus Back into Rock Engineering Design

consistent conditions. These domains are typically well must be taken in this domaining process to not overly com-
bounded within definable contact margins, which may be plicate design by creating too many domain divisions. It is
geological weaknesses (faults/shears etc.) or specific litho- important to synthesize and simplify complex geology, but
logical contacts (ref. yellow coloured boxes in upper part without losing the essence of variability in diagnostic condi-
of Fig. 3). tions. For instance, depending on the scale of sedimentary
The domain itself, within which conditions show con- layering present in a rockmass, distinction might be made,
sistency, could in addition be a specific lithological unit or between one rockmass area and another based on the distri-
maybe a multifaceted rockmass. It would likely be jointed, bution of specific lithological units within each area. One
but also could be layered, foliated or schistose (ref. in par- area perhaps might be lumped together as a composite bed-
ticular the blue coloured boxes in the lower part of Fig. 3, as ded sequence, compared to another area, where each major
shown in enlarged detail in Fig. 4). lithology change might be separately characterized.
Major engineering structures (e.g., dams, tunnels, open Key diagnostic discontinuity assessment parameters that
pits, mine shafts etc.) frequently are sited across multiple aid definition of tectonic/structural domain consistency need
geological domains. In consequence, no one GSI, RMR or to be carefully examined (viz. discontinuity pole concentra-
Q value will likely be sufficiently diagnostic of site con- tions need to be defined on the stereonet, and surface condi-
ditions to be representative enough for design purposes. tion (JCond, Jr/Ja and JRC) parameters need proper quantifi-
Defining appropriate ranges in classification assessments cation. The question that needs also asking is whether these
should, therefore, be the focus for engineering structure parameters are consistent with geological appreciation of the
site assessment. Characterization should start with careful defined domain? ref. green shaded questions on the right-
domaining of the region around (and including) the project hand side of Fig. 3). If the proposed domain is structurally
site, so that unique, unambiguous zoning is achieved of the complex—the fabric should be checked to see if the domain
entire site area of engineering importance. Care, however, is within a fold zone or criss-crossed by patterns of fractures

Fig. 3  Components requiring definition for unambiguous GSI classification of a structural domain

13
 T. G. Carter, V. Marinos

Fig. 4  Detail of left part of domaining classification diagram illustrated in Fig. 3

that are sympathetic or conjugate to bounding faults (i.e., micro-faults shown in this same diagram in Fig. 5, typically
are Riedel structures present? (Riedel 1929, ref. Fig. 5). Is named X-features, per Bartlett et al. (1981), align at around
consistency of fabric maintained across the domain—ref. 105° (i.e., at 90 + ϕ/2°). Particular care should always be
red shaded questions (from the right-hand side of Fig. 3). taken to identify such structures, and also whether they are
The questions included on the right side of Fig. 5 are cross-cut by other, geologically later structures. Character
typically those that need to be asked at early project stages, and orientations of veins often help in this task, as they fre-
and then re-examined periodically as design progresses. For quently give clues to age dependence of structures, and as
instance, if a fold fabric exists across a potential domain, such they should also be clearly documented during domain
this will most likely manifest itself by discontinuity pole characterization. Veins, in fact, need special attention as they
concentrations showing up as a “Girdle” of poles distributed can severely alter presumed rockmass competence if not
across the stereonet, and should be interpreted as such, rather adequately addressed (Day et al. 2019).
than picking three, four or more joint sets from seemingly The last aspect of concern to check (in defining a GSI
minor pole concentrations on the net (a common problem domain) is whether or not changes in rockmass condi-
seen in many engineering reports lacking adequate geo- tion also occur across the proposed domain as a result of
logical input). When Riedel styles of conjugate structures weathering and/or alteration (ref. orange shaded questions
exist, again pole concentration patterns on a stereonet will included in Fig. 5, as extracted from Fig. 3). Either pro-
be diagnostic. Typically, as shown in the diagram on the cess can significantly affect rockmass quality—as clearly
lower right of Fig. 4, Riedel fabric R-style micro-normal shown in the GSI plots in Figs. 6 and 7 of Marinos and
faults and P-style micro-thrust faults plot at about ϕ/2 = 15° Carter (2018).
off axis to the plane of a major fault, with Riedel fabric R′ A check on consistency of fracture intensity across a
style micro-faults plotting at around 75° (i.e., at 90 − ϕ/2° domain needs also to be made. This can be achieved either
to the major deformation plane). The remaining suite of by taking three orthogonal standard uni-directional (scanline

13
Putting Geological Focus Back into Rock Engineering Design

Fig. 5  Details abstracted from right-hand side of domaining diagram, ing (modified from Rovida and Tibaldi 2005, after Skempton 1966,
with inset stereoplots and nomenclature for riedel and secondary fault Tchalenko 1970 and Bartlett et al. 1981)
structures with respect to example right lateral (dextral) wrench fault-

or core) index measurement estimates to assess relative frac- domain should, however, always be sought. Ideally, repre-
ture density, or better still by utilizing a direct volumetric sentativeness should be checked by preparing histograms
measure of rockmass fracture intensity. The linear meas- and statistical plots for each of the defining index parameters
ures that allow some appreciation of rockmass brokenness listed in the centre of Fig. 3, as enlarged on the left side of
(blockiness) include RQD (Deere 1964) and Fracture Fre- Fig. 5. Defining domain conditions using multiple (alterna-
quency (Priest and Hudson 1976), also definable as P10—per tive) characterization approaches, including checking areal
Dershowitz and Herda 1992 and Dershowitz et al., 2003, or volumetric variography of parameters and also validating
while the volumetric measures that have applicability are stationarity statistics per approaches explained in Barnett
VFC, also definable as P30 (Schlotfeldt and Carter 2019), or and Carter (2020) will help with completing the assessment
Jv (Palmstrom 1982). Ideally, one would wish to measure by clarifying uncertainties and assisting in ensuring that
P32 but this is extremely problematic in real rockmasses due consistency is maintained across the defined domain.
to lack of sufficient three-dimensional exposure of actual
fracture geometries and terminations. Arguably VFC thus 2.2 Applying Multiple Classifications in Parallel
provides the best field measure for defining fracture intensity for Improving Insight
for determination of a domain GSI estimate.
Notwithstanding the need for ensuring that the two key Each of the major classifications has different strengths and
geological principles of structural domaining set out in Fig. 3 limitations in terms of their appropriateness for defining
are met, viz—(i) consistent internal fabric and (ii) clearly the rock quality of a domain or a specific rockmass zone or
definable bounding structure or lithological change, statis- sample segment (drill core, tunnel face, bench slope, foun-
tical representativeness of collected data across a proposed dation slab, rock outcrop, adit portal zone etc.). At early

13
 T. G. Carter, V. Marinos

Fig. 6  Illustrative Use of the Quantitative (left side) and (right side) of ­JCond89 for the discontinuities (ranging from 8 to about 22, but
Observational GSI Charts for geological characterization of an most commonly 15), with typical RQD’s estimated as around 50%
observed very blocky to disturbed rockmass with planar to slightly (but ranging from 25% to 60%)
rough discontinuities, as also reflected by quantitative determinations

design stages when maybe only “non-oriented” drill core is (2013), helps in highlighting data gaps, thus aiding focus for
available, it is recommended that duplicate characterization improving investigation knowledge.
be attempted using only RMR (either the 1976 or the 1989 At this earliest (back of the envelope), concept design
versions, but clearly stating which has been used), plus the phase, when very limited “hard” data is typically all that
“quantitative” GSI chart (introduced by Hoek et al. 2013, is available for defining rockmass quality and variability, it
as illustrated in the left-hand diagram within Fig. 6). This is recommended that sequential application of at least two
is recommended solely, because characterization from core of the routinely applied, numerically based rockmass clas-
alone is easiest to get correct when using these approaches sification approaches (Q, RMR, RMi or quantitative GSI, per
alone. Estimates for RMi and particularly for Q, really should left-hand chart in Fig. 6) is attempted, with each estimate
not be made on “non-oriented” core, without availability of checked one against each other and also against a “blob”
reliable stereonets defining the correct number of joint sets plotted on the observational GSI chart (ref. right-hand dia-
and their inter-relationships (Carter 1992). gram in Fig. 6). It is recommended that careful comparative
Ironically, there is more merit in applying more than one graphical plotting per the style of Fig. 6 be undertaken for
classification approach in the earliest stages of project devel- each of the various rock types and structural domains across
opment, per the x-axis of the right-hand diagram in Fig. 1, the project site. Graphical “blobs” from each of the multiple
rather than when one has reached the detailed design stage. rockmass quality determinations should then be overlaid one
At early concept design stages, when there is the least known over the other so as to gain some visual appreciation of true
about the characteristics of the rockmasses of concern, the variability in rockmass conditions across the entire site.
rigour of collecting the quantitative data needed for applying Sequentially preparing these type of comparative plots—
Q, RMR, RMi or defining a quantitative GSI, per Hoek et al. first by rocktype and domain and then as a composite

13
Putting Geological Focus Back into Rock Engineering Design

Fig. 7  Typical Matrix Diagram from Marinos and Carter 2018 illus- diagram, (from Carter et al. 2008), illustrating mi and σC controls on
trating distribution of published GSI charts for many common rock- range of applicability for GSI and conventional H–B equations com-
types in mi-σC space (left diagram), ranging from strong, high com- pared with brittle spalling (top right) or Mohr–Coulomb (lower left)
petence rocktypes, with high mi and σC values (top left) to lower (squeezing) transitions
right, typical low competence low mi-σC rocktypes, set against right

overlay, will not only aid design understanding, but has the knowledge base for these approaches were the hundreds of
added benefit of constituting a quick reliability check on kilometres of tunnels that had already to that date been suc-
the correctness of all the classification estimates. This in cessfully constructed under moderate stresses and in fair to
turn should further improve insight—another benefit for the good quality rock conditions.
design team. At this early project evaluation stage, applica- Although the original purpose for all these classifications
tion of multiple classifications can be thought of as a critical was tunnel support estimation, because of their origin (US,
design check—for not only establishing areas of good and Germany, Austria, Scandinavia and South Africa, respec-
poor rock mass quality, but also of adequacy of understand- tively), their databases of experience were from quite a wide
ing of these areas. While inevitably data gaps will get high- range of different rockmasses and stress conditions. Further-
lighted in this process, so too will any potential fatal flaw more, some of the classification codings and rankings were
geotechnical conditions. Any, and all worrisome data gaps tunnel-size specific and thus they could not easily be directly
should trigger rigorous follow-up investigation. applied for characterizing a non-tunnelling situation.
In consequence of acceptance of the general principle
2.3 Applicability Limitations of Available that classifications were a useful step forward for char-
Classifications for Rockmass Strength Definition acterizing a rockmass for a given engineering purpose, a
plethora of modified classifications appeared soon after
In the preceding discussion, it has been taken as read that the original Q and RMR system publications were intro-
any or all of the well-known available classifications can be duced in the mid 1970s. Most of these modified classi-
applied simply to assess rockmass quality and hence make fications were built on the original classification tables,
design strength estimates. This, however, is not always the but with explicitly generated codings and rankings for
case, and reliability of such estimates needs careful con- a variety of non-tunnelling situations—for slopes (e.g.,
sideration. Firstly, it should be appreciated that when the Romana 1985, Sonmez and Ulusay 1999),—for open stop-
earliest of the more well-known and more commonly applied ing (Mathews et al. 1981; Potvin et al. 1989) and—for
rock mass classification systems were initially developed cave mining (Laubscher 1977). All of these approaches
(Terzaghi 1946; Lauffer 1958; Barton et  al. 1974; Bar- have their merits and detractions, but correctly used they
ton 1976; Bieniawski 1973, 1976), they were principally can be of great benefit for design of the specific problem
intended for tunnelling purposes, and mainly as an empiri- type that they were developed for. However, with the focus
cal design method for estimating underground support. The that greater understanding is needed of rockmasses for

13
 T. G. Carter, V. Marinos

numerical modelling calibration per Sect. 1.1, it should 3.1 Inception of the Geological Strength Index (GSI)
be appreciated that many alterations have been made to
the individual sub-coding parameters of the original Q In concept, the original extension of the Hoek–Brown intact
and RMR systems by some of all these modified classi- rock failure criterion for use for rockmasses was predicated
fications (particularly MRMR) yet retaining very similar on the assumption that rockmass strength would be domi-
nomenclature. Accordingly, great care must be exercised nated by interaction and interlock of rock blocks (of various
in avoiding mixing and matching datasets from differ- scales), controlled by the condition of the surfaces between
ent classifications. Checks need to be rigorously made to these blocks. The relationships between s and the mb/mi ratio
ensure uniformity of descriptive characterization, so that for different rockmass qualities (Hoek and Brown 1980a, b,
proper comparisons can be made back to the original Q 1988; Hoek et al. 1992), involving equations relating RMR′,
and RMR tabulations if these are then to be used as basis were seen to work well. However, with improvement in ana-
for deriving modelling input parameters. lytical approaches and greater availability of boundary and
finite element modelling tools in the late 1980s and early
1990s, cases were starting to be observed that suggested
that these equations through RMR′ did not always gener-
3 Classifications and the Hoek–Brown ate appropriate mb and s values that matched what more
Criterion rigorous numerical back-analyses suggested, particularly
for underground rockmass behaviour (Carter and Marinos
It should immediately be apparent from the above discus- 2014).
sion and from reading the original references for the Q and Introduction of the Geological Strength Index (GSI) into
RMR classifications that neither was created specifically for the 1995 equations was seen by Hoek and others through
rockmass strength definition, despite the fact that nowadays 1994/1995 as the most effective way of improving and clar-
they are commonly used often out of context for this exact ifying procedures for H–B parameter definition for vary-
purpose. Hoek and Brown in the 1980s, however, foresaw ing quality rockmasses, aiming—(i) at reducing confusion
the applicability of classifications in this specific role and between the different RMR codings (1989 vs 1976) and (ii)
made direct use of both systems in parallel as the basis for emphasizing a return to more of an observational geological
initial rockmass indexing for extending the Hoek–Brown emphasis as basis for defining variations in rockmass quality.
(H–B) failure criterion from intact rock to application for The 1995 tables thus list mb/mi ratios, as well as s, a, Em and
rockmasses (Hoek and Brown 1980a). By this time, clear v value estimates alongside GSI, rather than alongside RMR
understanding of how to define and classify rockmass char- or Q (ref. Table 8.4 in Hoek et al. 1995).
acteristics was emerging industry-wide (Anon 1977, 1978). The initial, truly graphical observational chart concept
A table with a widely circulated generic tabulation of intact came a few years later, following development through the
strength de-rating relationships tied to different rockmass late 1990s, resulting in publication of a suite of rock-type-
qualities, along with minor modifications also to the H–B specific GSI charts by Marinos and Hoek (2000) using the
criteria, was presented in Hoek’s (1983) Rankine Lecture same basic graphical framework as the chart included in
(updated from Hoek and Brown 1980b). the right of Fig. 6, termed since Hoek et al. (2013), as the
At this stage, the sole purpose of estimating a rock qual- “Observational Chart”. Marinos and Hoek in 2000 (rein-
ity value (via Q or RMR) for application of these tables was forced subsequently by Marinos et al. 2005), strongly rec-
so that an appropriate scaling correction could be applied ommended that GSI should be estimated directly from this
for downgrading intact rock strength for application of the observational chart, as per the examples shown in the papers,
H–B criterion to fractured rock masses. Accordingly, the suggesting that practitioners ignore the previously published
modified RMR’ and Q′ (prime) indices were introduced, equations, due to their inaccuracy over certain rock quality
recognizing even at this stage that use of the H–B criterion ranges. To further emphasize this difference of purpose, and
for subsequent design analyses likely would involve double also assist users with the process of parameter definition
counting some facets of parameters within both the Q and for difficult and varied weak and complex rockmass condi-
RMR systems, namely the strength term, the structural ori- tions, various extensions to the original set of half dozen
entation term and the groundwater term in RMR, and the Jw rock-type-specific charts included in the 2000 publication
and SRF parameter terms for groundwater and stress in the were added by Hoek, Marinos and Marinos from 2005 to
Q system. Hoek et al. (1995) addressed these modifications date (see Marinos and Carter 2018; Hoek et al. 2005; Hoek
for the inclusion of RMR89 in the H–B strength estimates. and Marinos 2007; Marinos and Hoek 2001; Marinos et al.
Q′ by this stage had been omitted from inclusion in the scal- 2007; Marinos 2007, 2014, 2017 more details). Mainly these
ing relationship equations for simplicity reasons, primarily additional charts were developed from experience gained
because of Q’s logarithmic behaviour. during excavation of more than 80 tunnels driven as part

13
Putting Geological Focus Back into Rock Engineering Design

of the Egnatia project (Egnatia S.A.) in Northern Greece, right-hand diagram in Fig. 7), with the 2002 H–B equa-
in addition to the Railway tunnels of Ergose S.A. and the tions only considered valid in the centre region (viz.
tunnels of Athens Metro S.A. With these additions, it was 25 < GSI < 65).
hoped that practitioners could use GSI for quite a wide range It was further noted that the modified observational GSI
of rock masses. charts that had been introduced in the intervening years
Despite this hope that GSI could be observationally (many of which are shown as small inset diagrams set
assessed and then utilized directly in the 1995 equations as within the left-hand diagram in Fig. 7, derived from Mari-
a straightforward approach for design, this was not found nos and Carter 2018) were in reality in part attempting to
totally the solution, as inconsistencies were still being noted compensate for these differences in rockmass behaviour
between modelling results utilizing this methodology and ascribed with each of the three divisions delineated in the
field experiences. This was particularly found to be the situ- right-hand diagram in Fig. 7. Evaluation of detailed case
ation for anisotropic rockmass behaviour, suggesting that for history data in fact suggests that there has thus been an
these types of rockmass the H–B criteria governing equation element of compensation for this lack of equation accu-
may not be exactly a square root relationship, but may vary racy built into such charts through overlap between diver-
with rocktype. Accordingly, a change was made to the 1995 gent data groupings across the individual GSI charts, par-
H–B expressions in 2002 to replace the square root term ticularly towards the top left and lower right-hand side
with a variable power term, designated as “a”, and defined of the matrix of small-scale charts. The kaleidoscope of
to be dependent on GSI as a means to more appropriately published GSI charts, included as Fig. 7 thus provides a
address in particular foliated and other anisotropic types of ready guide to aid practitioners select an appropriate mi
rockmass (per Hoek et al. 1992). The equations also were and UCSi range of applicability for their rockmasses and
updated with the introduction of the parameter D repre- rocktypes guided by the characterization portrayed by
senting a variable Disturbance Factor (per guidelines also these various example charts.
included in Hoek et al. 2002), aimed at alleviating the need Within the central division in the right-hand diagram
for two sets of equations (“disturbed” and “undisturbed”) per in Fig. 7 (which encompasses most “normal” blocky rock-
Hoek and Brown 1988. Removing this “switch” at GSI = 25, masses with GSI’s above 25 and below 65), the original H–B
which had been required in the earlier equations published equations have full validity. It is at the upper and lower ends
by Hoek et al. 1995, was one of the most important goals of of the scale that care must be taken of using the original
this 2002 revision. H–B equations.

3.2 Applicability Limitations of GSI 3.2.1 Near Intact Competent Rock‑Masses

within the Hoek–Brown Equations
At the high end of the rock competence scale (where
The 2002 equations and the visual observational chart GSI > 65, and mi > 15, per the right-hand diagram in Fig. 7),
method of definition of GSI, as introduced in 2000 has because discontinuities are very widely spaced, block sizes
stood the test of time quite well for most rockmass condi- have become so large that for most engineering scale pro-
tions met in typical engineering project applications. Most jects intact material strength, rather than pre-existing frac-
experienced practitioners found defining GSI from the obser- turing, dominates and controls rockmass behaviour. Here,
vational chart with reference to the margin axis descriptive the controlling behaviour is tensile splitting, thus spalling
listings quite straightforward, over almost the complete spec- H–B parameters need definition and great care must be taken
trum of rockmass competence. Not so the application of the in assigning GSI values and ensuring that these are not indis-
equations. The 2002 equations seemingly still proving to criminately used within the conventional H–B equations.
yield problematic results for rockmass strengths at the two Typically, the conventional H–B equations do not deal with
ends of the rock competence scale—for weak soft rocks and initiation of spalling failure without modification, but they
for high strength, high competence rocks. can be used for modelling compressive failure of already
Based on careful examination of rockmass behaviour tensionally split and damaged rock (Diederichs et al. 2007).
at the two ends of the rock competency scale (as reported Under these conditions, they will give an acceptable predic-
in Diederichs et al. (2007), for behaviour in high strength tion of damaged strength for hard brittle rocks. The upper
(spalling condition) rockmasses and in Carvalho et  al. end HB transition equations, as listed on the right-hand side
(2007), for behaviour at the low end of the competency of Fig. 8, and graphed in the left-hand diagram of the same
scale, in weak, low strength (near Mohr–Coulomb) rock- figure, however, allow one to also deal with the tensile ini-
mass conditions), Carter et al. (2008) suggested that three tiation of brittle fracture in these more intact, essentially
different regimes of rockmass behaviour might exist (per unfractured types of rockmass.

13
 T. G. Carter, V. Marinos

Fig. 8  Transition and Conventional Hoek–Brown Equations and Plot 2008, where UCSi = σci  = 
Intact Uniaxial Compressive Strength;
of Normalized Rockmass Strength (σcm/σci) as a function of Rock T = σi 
= Intact (Direct) Tensile Strength; UCS* = crack initiation
Quality (GSI), illustrating marked differences between conventional threshold strength; and where fSP and fT(σci) are the Spalling and
Hoek–Brown behaviour and Transition Functions. Spall transition Weak Rock Transition Functions)
threshold set at typical value, UCS* = 0.45σci (from Carter et  al.

3.2.2 Weak, Low Competence Rock‑Masses values across the complete rock competence spectrum. Not
so the quantitative GSI chart (Hoek et al. 2013) as this also
At the other, low end of the rock competence scale, again only covers the blocky rockmass range for which the origi-
rockmass structure has diverged away from a blocky fabric nal H–B equations remain valid and for which RQD can
to something much closer to an intact material—in this case, adequately describe block size. For larger block sizes, where
closer to a soil. At this low end of the rock competence scale RQD fails to capture variability, the vertical chart axis scale
(typically, where UCSi < 10 MPa and GSI < 25), disconti- can be replaced with VFC, following the methodology out-
nuities in the rockmass fabric again start to play less of a lined in Schlotfeldt and Carter 2018. While more difficult
role, as the rockmass becomes increasingly more isotropic to describe than the central, more blocky, rockmasses, for
and rockmass strength again tends more towards that of the which the Q and RMR classification systems and the quan-
“rock” matrix. In this regime, cohesive strength becomes titative GSI chart all work reasonably well, rockmasses at
much more important and thus the transition equations listed both ends of the competence spectrum are best described in
on the right-hand side of Fig. 8, and graphed in the left-hand terms of their parent rock material fabric.
diagram, allow transfer from Hoek–Brown blocky non-linear
behaviour to where the traditional linear Mohr–Coulomb
criterion gains more applicability (ref. Carvalho et al. 2007, 4 Intact Rock Parameters
and subsequent update calibrations presented in Castro et al.
(2013). From the preceding discussion, it will be clear that for the
majority of blocky rockmasses in competent rock, spacing,
3.2.3 GSI Chart Applicability character and interaction of discontinuities constitutes the
principal control on rockmass strength and stiffness, and
It is important to appreciate that while the original H–B hence on behaviour. The rock material forming the blocks,
criterion applicability range (including the 2002 equations), as long as it remains competent, plays less of a role, rela-
needs to be constrained to be used only within the blocky tively speaking, than the discontinuities, hence the lack
rockmass regime per Fig. 8, the qualitative, visual GSI chart of any geological differentiation in, for example, the Q
(per Marinos and Hoek 2000) can be used to define GSI or GSI classifications. Some geological differentiation is,

13
Putting Geological Focus Back into Rock Engineering Design

however, included in the basic RMR approach, as strength 4.1 Characterization Insights for Defining Intact
is included as a parameter descriptor, but this term specifi- Rocktype Parameters
cally needs to be excluded or nullified when RMR′76 is uti-
lized within the H–B criterion. Only when analysis moves The matrix diagram in the left-hand side of Fig. 7 hope-
to attempting to define these types of blocky rockmass as fully helps clarify this issue for users confused with
an H–B material does any serious consideration get given respect to the concept of different GSI’s all being possible
to intact rock block characteristics and any lithological at the same time in different rock conditions. The scales
variability. Not so for the materials at the ends of the rock and details on this 2018 matrix diagram, which have been
competence spectrum—here defining lithological charac- improved and clarified from the initial 2014 version (per
teristics and material matrix strength are of paramount Carter and Marinos 2014), identify key ranges for mi and
importance. Hence inclusion of UCSi in both the high UCSi for each of the already published rock-type-specific
(spalling) and low (squeezing) equations for the upper and GSI charts introduced over the 2000–2015 period, but dia-
lower ends of the competence spectrum, and also through- grammatically set within a framework defined in mi-σ C
out the transitions, per the curves plotted in Fig. 8. It is space.
here in the transition zones that it becomes clear how close While differing from the type of divisions commonly
is the inter-dependence of rockmass strength with GSI seen in geological texts, where classification usually starts
and with the other key intact rock parameters (ref. (i) plot by subdividing all rocktypes into three principal groups—
on left-hand side of Fig. 8, where variations in GSI and igneous, metamorphic and sedimentary, as characterized by
mi along with variations in UCSi can be seen to control their mode of origin, mineralogy, grain size and fabric, the
the shape of the curves, and (ii) the division lines drawn matrix diagram in Fig. 9 of Marinos and Carter (2018) (as
across the right-hand diagram in Fig. 7, which again are reproduced in a smaller version on the left side of Fig. 7),
defined by all three parameters). sets out each of the individual GSI charts with respect to
However, unfortunately there is a dearth of good reli- basic rock competence.
able guidance on appropriate combinations of param- The small charts are laid out, from strong, competent
eter values for incorporation into modelling analysis to rocks (top left), to weak (generally fine grained, almost soil
properly reflect the true variability evident in lithological strength) rocks (bottom right). Each chart thus represents a
characteristics and material matrix strength shown by the unique, discrete region within this overall rock competence
different litho-based GSI charts in Fig. 7. Despite the huge framework, that, in essence, should also group all common
advances that have occurred over the last several decades rocktypes by their fabric. In broad outline, this is indeed the
in computational power and sophistication in modelling case, but not in detail, so again, there is overlap, as rocks of
software, currently much of what has been the defacto differing fabric and/or of differing origin, can yield identi-
standard for parameter definition remains still based on cal Hoek–Brown mi values. So, useful as this matrix chart
only partial data, and mostly on data that was compiled is for gaining a conceptual appreciation of how all the GSI
prior to the 1980s when the original H–B equations were charts inter-relate one with another, it does not provide a
first published. This again is an industry-wide deficiency precise enough framework for it to be used to directly select
that needs addressing. an appropriate mi value.
Inspection of the various small GSI charts included in Conceptually, it would seem straightforward that one
the matrix diagram on the left-hand side of Fig. 7, particu- should be able to overlay geological origin—viz., igneous,
larly for the weaker rocks clearly shows an inter-depend- metamorphic or sedimentary, in combination with rock com-
ence of parameters—is this a variable that has not as yet petence, per Fig. 8, thus creating a workable engineering
been [properly captured? Depending on specific details classification chart to allow mi to be defined texturally and
within each of the individual charts, GSI values can be by rocktype. Although seemingly trivial, this task has proven
anywhere between 0 and 100 per chart, with associated a bit of a holy grail, as unfortunately, even though many
applicable changes in mi varying along the x-axis in step attempts have been made over the years to create just such
with major changes in UCSi occurring up and down the a comprehensive nomogram diagram, no simple engineer-
y-axis. This inter-dependence of these three parameters ing characterization chart has as yet been developed that
may seem at first confusing, particularly when consider- adequately ties engineering intact rock properties with vari-
ing that one may have the same specific GSI value for any ability in rocktype geology.
number of mi and UCSi combinations, and vice versa. This Deere (1968), following Deere and Miller 1966, made
has led to serious questions as to what really constitutes a one of the earliest attempts to develop a workable chart for
GSI of 100, and how can one have a GSI of 100 for com- characterizing intact rock material, introducing the concept
petent strong granite and also for a weak flysch mudrock? of the Modulus Ratio (MR = Ei/σC) as a guiding parameter
pairing (as summarized on the left side of Fig. 9).

13
 T. G. Carter, V. Marinos

Fig. 9  Characteristic Ranges of Modulus Ratio (Ei/σC) for Various of Marinos and Carter (2018) along with listings from Table 4 from
Rocktypes (left diagram, from Deere and Miller 1966); and right Hoek (1999) and Modulus Ratio (Ei/σC) tabulations derived from
diagram, plotted with respect to mi (by combining data from Fig.  9 Table 3 from Hoek and Diederichs (2006))

Although numerous, slightly modified or updated ver- trendline MR value of 350–400, irrespective of rock suite,
sions of this type of Modulus Ratio chart still continue to suggesting that mi might better relate to variability away
appear in the literature (e.g., Fig. 2 of Feng and Jimenez from some mean trend line, rather than providing a direct
2014, as reproduced on the left of Fig. 10), this issue of index for defining MR or indeed differentiating between
finding a better means for characterizing different rocktypes rocktype origins. This same MR data when plotted with
remains unresolved, with Deere’s Modulus Ratio concept respect to known tensile strengths, however, shows similar
still remaining about the best consolidated inter-relationship disparity trends, suggesting that the wide divergence in MR
available, even today. values at a given mi is largely attributable to tensile strength
Previous attempts have, however, not looked specifically variability.
at mi pairing with Modulus Ratio or other laboratory param- To explore whether plotting the matrix diagram of small
eters as a possible better rocktype variability index. Rigorous GSI charts more accurately by rocktype might yield some
plotting of each of the main data combinations suggests that better trends than apparent in Fig. 7, an attempt has been
Modulus Ratio vs mi doesn’t appear to hold as much promise made in the diagram on the right side of Fig. 10 to more
as can be achieved by comparing the individual components properly scale the actual parameter ranges with respect to the
Ei and UCSi vs mi. This is illustrated quite well in the right- x and y, mi and UCSi axes. This Fig. 10 plot, which has been
hand diagram in Fig. 9, which has been compiled by com- prepared in much the same style as the original Deere charts
bining data from Fig. 9 of Marinos and Carter (2018) along (e.g., that shown on the left side of Fig. 9, and used as basis
with listings from Table 4 from Hoek (1999) and Modulus for the two plots on the left side of Fig. 10), unfortunately,
Ratio (Ei/σC) tabulations derived from Table 3 from Hoek yet again, suggests significant overlap of characteristics
and Diederichs (2006). As can be observed, comparing MR between rocks of different origins, lithologies and fabrics
directly with respect to mi with the rocktypes also divided by (particularly at either end of the mi scale).
rock suite (sedimentary, metamorphic, igneous) shows con- An attempt was accordingly then made to split the tabu-
siderable scatter, particularly at low mi values. The graph, lated Modulus Ratios per Table 3 of Hoek and Diederichs
however, seems to suggest what seems like a constant mean (2006) directly back into their component UCSi and Ei

13
Putting Geological Focus Back into Rock Engineering Design

Fig. 10  Possible inter-relationship between Hoek–Brown constant vs Ei/σC compilation (per Fig.  9, right diagram), plus (ii) mi vs σC
mi and Modulus Ratio (Ei/σC) (left diagram and inset mi definition compilation plot of Table 2 and Fig. 9 data from Marinos and Carter
plot)—based on (i) Feng and Jimenez 2014, Fig.  2, updating Deere (2018) (right diagram)
1968, and Deere and Miller 1966 (per Fig.  9, left diagram) and mi

ranges, to see if cross-plotting the individual components utilized by Hoek (1999) listed along the base of the chart and
rather than the Modulus Ratio might yield better correlations colour-coded with respect to geological origin. To achieve
that would allow a guidance chart to be prepared for better this a wide cross section of published sources of UCSi data
aiding parameter selection for modelling practitioners. Some for these different rocktypes were researched from available
concern was felt that the data might be compromised by the literature, with some of the most important references listed
fact that more than 50% of the Hoek (1999) mi table cor- below Fig. 11.
responding to the same rocktypes as Hoek and Diederich’s The results of this particular plot need further evaluation
2006 table of MR’s are in fact estimated values rather than and verification with real data replacing the assumed val-
measured data, and that the mi range spreads were likely ues from the Hoek (1999) tabulations, but initial results are
seriously under-estimated as shown by Richards and Read’s encouraging. As is evident, when only the individual com-
2011 et seq. However, as an approach was sought that would ponents of the Modulus Ratio for each of the 42 rocktypes
allow better estimates to be made of mid-range values of key are plotted with respect to a linearized monotonic sorting of
parameters per rocktype, rather than extreme end member the 42 rocktypes based on Ei, reasonable “best fit” regres-
values, it was decided to retain consistency with the two sion lines can be readily drawn through the data, allowing
key tables (Hoek 1999, Table 4; Hoek and Diederichs 2006, quite simple regression equations to be developed between
Table 3). the various parameters, as summarized in matrix form in
Figure 11 plots the best regression trend relationship Table 1.
results derived from cross-plotting the three component As discussed above, the plot in Fig. 11, which ranks
intact material parameters with respect to lithology. In this the 42 rocktypes from Hoek (1999) across the base of the
best case plot, the data has been ranked not by mi or Mod- chart has deliberately been created, with reference not to
ulus ratio, MR, but by Ei, with each of the 42 rocktypes mi, but rather to Ei based on interrogation of the tabulation

13
 T. G. Carter, V. Marinos

Fig. 11  Comparison of Table  4 mi dataset from Hoek (1999) with (1968), (ii) the MR plot in Fig.  2 of Feng and Jimenez (2014), plus
Table  3 Modulus Ratio (Ei/σC) dataset from Hoek and Diederichs (iii) data from Table 9 of Hoek and Brown (1980b); plus (iv) Table 2
(2006) for 42 different rocktypes, sorted by Ei based on ranges and from Chapter 1 of Coates (1970) and (v) the MR plot in Fig. 10 from
means estimated for UCSi for each mi/MR datapoint pair, extracted Peck (1976), plus additional data from Douglas (2002), Sheorey
from (i) the MR plots presented as Figs. 1.1 through 1.7 within Deere (1997) and from Rocscience (2018) (as extracted from RocData)

Table 1  Approximate inter- Ei (GPa) UCSi (MPa) mi only yielded plots with much worse divergences, but the
relationship matrix for deducing relationships were clearly less linear.
first order estimates of mi, UCSi 1 2.33 0.20 It was also noted that Ei also seemed to sort the rocktypes
and Ei values, based on linear
0.43 1 0.08 into their parent rocktype origin suites better than achieved
regressions shown in Fig. 11
5.11 12.00 1 by mi, MR, or by UCSi—perhaps pointing to degree of indu-
ration being the key differentiator, as suggested in Carter and
Marinos (2020). As shown by the colour coding along the
of Modulus Ratio (MR) values given in Table 3 of Hoek base of Fig. 11, the igneous rocks clearly group to the high
and Diederichs (2006) for the same suite of rocktypes as end of the Ei-axis scale, with the sedimentaries towards the
Hoek’s (1999) Table 4. Ei has also been specifically used low end and the metamorphics spread throughout, in quite
as the base for generating the other regression equations logical progression based on typical characteristics for each
to mi and to UCSi, as listed in Table 1 above, not just of the named rocktypes.
because choosing the modulus avoids the “estimate” vs Some of these controls are better evidenced by more
reality uncertainty issues associated with the mi values detailed examination of textural characteristics, with fur-
listed in Table 4 of Hoek (1999) (more than 50% of the ther assessment of possible grain size, interlock and textural
values being identified as estimates), but more so because rocktype character and of induration on definition of mi and
the range bar spreads for Ei within Fig. 11 are much nar- Ei discussed separately in Carter (2019), Carter and Car-
rower and the data value rankings much more linear than valho (2020) and in Carter and Marinos (2020).
when plotted with either mi or UCSi as base reference. In Here, as discussed throughout the first three sec-
fact, this monotonic ranking by Ei shown in Fig. 11, con- tions of this paper, focus is on first defining Regions
stituted the best of the three plots that were generated in and Domains, then within each Domain, best defining
attempts to group the other two, non-ranked, independent an appropriate GSI for the rockmass of concern in that
parameters over the full rocktype range of Hoek’s (1999) Domain at a scale appropriate to the engineering struc-
suggested values. Sorting by mi, by MR, or by UCSi not ture under consideration for design, so that appropriate

13
Putting Geological Focus Back into Rock Engineering Design

geomechanical characteristics can be selected. This 5 Conclusions

requires not only good definition of the structural geology
of the site to properly domain every key part of the region Two primary conclusions can be drawn from the vari-
of interest, but good definition of GSI’s per controlling ous uncertainty elements of geological characterization
rockmass, together with good definition of appropriate addressed in this paper:
mi, Ei and UCSi values for the individual rock types
within that Domain. Only with such properly sequenced 1. Geologically based Domaining of Project Sites needs to
steps can representative H–B strength envelopes and stiff- be more widely practiced as a fundamental procedure to
ness characteristics and associated parameters be prop- be seriously taken into account in Design, particularly
erly assigned for use within any analytical or numerical where design is based on heavy utilization of numerical
models needed for assessing engineering structures to be models. Such domaining should follow the prescriptive
constructed in that domain rockmass. sequence of steps outlined in Table 2.
Checks, however, are essential as geological engineer-   Domaining is seen as key to achieving more represent-
ing as a discipline is not a precise science, and significant ative understanding of rock conditions and of parameter
variability is to be expected in certain conditions, but variability, hopefully leading to achieving better repre-
some uniformity should be always verifiable. Accord- sentativeness. Table 2 outlines a suggested sequential
ingly, once best estimate geometries and extents have approach to help rapidly coalesce often divergent data
been defined for each Domain, and parameter values for needed for firstly properly domaining a project site based
each step downwards from region to litho-specific rock- on structural geological understanding, and secondly,
type zone have been selected, it is then suggested that coming up with appropriate parameters by detailed and
the regression matrix in Table 1 be utilized as a basis rigorous assessment of litho-specific data.
for checking the validity of the chosen intact material   Following the steps outlined in this table should help
parameters. So long as the inter-relationships amongst allow more effective de-risking of future rock engineer-
the chosen mi, Ei and UCSi values are within ± 5% of the ing designs:
values determined from the matrix regression equations,   … and …
it can be assumed that the chosen parameters are suffi- 2. Much greater interaction is more than ever needed today
ciently representative of the individual rocktypes under between Geologists undertaking field data acquisition
consideration for that Domain. If, however, the checks and office-based Design Engineers involved in using key
show major inconsistency from the matrix inter-relation- input parameters in numerical models built to facilitate
ship equation results, then further checks should be made Rock Engineering Design.
on the laboratory data utilizing the granularity and pet-
rographic approaches suggested in Carter and Marinos With the widespread increasing use of greater sophistica-
(2020) and in Ganye et al. (2020) as basis for verifying tion in application of numerical modelling to rock engineer-
the appropriateness of the chosen values. ing design, it is becoming apparent that all too often there
is inadequate matching of sophistication in either collection
or analysis of raw field and laboratory data. It is hoped with

Table 2  Suggested sequential approach to domaining and parameter assessment as an aid for more effective de-risking of future rock engineer-
ing designs

1 Domain the project area a. Assess uniformity—by lithological controls

b. Assess boundaries—by structural controls
2 Per Domain: define mean and range of rockmass characteristics GSI, Q, RMR, RMi
Build domain a. Define block sizes & characteristics VFC, Jv, FF, RQD
rockmass b. Define intact rocktype parameters {G}, mi, UCSi, Ei
model
c. Define discontinuity parameters Individual & Per Set JCond, Jr, Ja, plus
& Scaled for Rockmass JRCo and JRCL
d. Define controlling shear strengths & stiffnesses (peak & residual & post- Per Set, Per Domain Discontinuities cres, øres,
peak behaviour) JCS, i, kn, ks
Rock Bridges ci_peak, øi_peak
Plus Equiv. cerm, øerm, Eerm
for Rock Blocks

13
 T. G. Carter, V. Marinos

application of the principles outlined in this paper (as syn- impossible to acknowledge the many industry and university colleagues
thesized into the summary roadmap table above) that Project who have contributed to the ideas expressed in this paper, the authors
specifically wish to acknowledge the encouragement of Prof. Giovanni
Geologists and Designers will be able to improve on this Barla extending over several years, for the authors to try to put together
situation and perhaps reverse this detrimental trend. Only some necessary guidelines to help analysts and modellers grasp the
with improved understanding of how best to characterize significance of correctness in definition of key input data for numerical
geological materials, so that data collection errors and other modelling. Insightful feedback from Dr. Evert Hoek at several stages
during the early preparation of this work is also greatly acknowledged.
uncertainties are minimized will Designers appreciate how
real project risk due to design errors can be reduced. Not
only is better understanding needed across the industry on
how to characterize inherent natural geological variability so
References
that this can be properly addressed in analyses and numerical
modelling, but better understanding is needed by many Pro- Anon (1977) The description of rock masses for engineering purposes.
ject Authorities of the erroneous, currently common miscon- Geological Society Engineering Group Working Party Report. Q
ception that greater sophistication in modelling and analysis J Eng Geol 10:355–388
Anon (1978) Suggested methods for the quantitative description of
during design is an effective risk minimization approach.
discontinuities in rock masses. International Society for Rock
It must be more widely appreciated that no level of Mechanics Working Group Compilation Report. Int J Rock Mech
increased sophistication in the design process can ever Min Sci Geomech Abstr 15:319–368
replace an insufficiency of site investigation data and a lack Barla G, Einstein H, Kovari K (2013) Manuscripts using numerical
distinct element methods. J Rock Mech Rock Eng 46:655
of comprehensiveness of geological evaluation. To improve
Barnett WP, Carter TG (2020) Structural domaining for engineering
general standards of rock engineering design and improve projects. In: Proceedings of the 54th US rock mechanics/geome-
the effectiveness of today’s sophisticated design tools for chanics symposium, Golden, Colorado. ARMA Paper 20-2105
de-risking a project’s design and even more so, its construc- Bartlett WL, Friedman M, Logan JM (1981) Experimental folding of
rocks under confining pressure Part IX wrench faults in limestone
tion, it is crucial that the steps outlined in this paper be more
layers. Tectonophysics 79:255–277
widely adopted and modellers, in particular, achieve more Barton NR (1976) Recent experiences with the Q-system of tunnel sup-
awareness of a number of salient design parameter uncer- port design. In: Bieniawski ZT (ed) Proceedings of the symposium
tainties that may be affecting the credibility of their model- on exploration for rock engineering. A.A. Balkema, Johannesburg,
pp 107–117
ling assumptions, for instance,
Barton NR, Lien R, Lunde J (1974) Engineering classification of rock
masses for the design of tunnel support. Rock Mech 6(4):189–239
i. the limitations of applicability of the original Hoek– Bieniawski ZT (1973) Engineering classification of jointed rock
Brown equations and associated application of the masses. Trans S Afr Inst Civ Eng 15:335–344
Bieniawski ZT (1976) Rock mass classification in rock engineering. In:
Quantified GSI chart (which spans the same range);
Bieniawski ZT (ed) Proceedings of the symposium on exploration
  … and … for rock engineering. A.A. Balkema, Johannesburg, pp 97–106
ii. the degree of uncertainty in the definition of many Bieniawski ZT (1989) Engineering rock mass classifications: a com-
of the key parameters describing the rockmasses and plete manual for engineers and geologists in mining, civil, and
petroleum engineering. Wiley, New York, p 251
rocktypes being characterized as basis for numerical Brox D (2020) Hydroelectric waterway inspections—risks and causes.
modelling. As a couple of examples … the unreliabil- TunnelTech Article. http://www.tunne​ltalk​.com/Tunne​lTECH​
ity of using single value estimates of key discontinuity -Mar20​20-Hydro​elect​ric-tunne​l-inspe​ction​s.php
or rocktype specific parameters, such as JRC or mi or Carter TG (1992) Prediction and uncertainties in geological engineer-
ing and rock mass characterization assessment. In: Proceedings
ϕ need to be more widely borne in mind, and … more of the 4th Italian rock mechanics conference, Torino, pp 1.1–1.22
care and understanding is needed of the risks inherent Carter TG (2010) Applicability of classifications for tunnelling—valu-
in utilizing “pull-down” styles of tabulated parameter able for improving insight, but problematic for contractual support
values as commonplace nowadays for selecting input definition or final design. In: Proceedings of the world tunnelling
conference (WTC 2010), 36th ITA Congress. Vancouver, Paper
parameters for concept or prefeasibility levels of mod- 00401, Session 6c
elling (before reliable laboratory testing information Carter TG (2014) Guidelines for use of the scaled span method for
is available). These pitfalls can compromise an other- surface crown pillar stability assessment. In: Proceedings of the
wise workable design. At the earliest stages of Project 1st international conference on applied empirical design methods
in mining, Lima-Perú, 9–11th June
Development is where such types of uncertainty need Carter TG (2015) On increasing reliance on numerical modelling and
most attention synthetic data in rock engineering. In: Proceedings 13th ISRM
international congress of rock mechanics, 10–13 May, Montreal,
Canada (paper 821)
Acknowledgements  This paper has been prepared to address an indus- Carter TG (2018) Suggested standards for improving structural geo-
try-wide need, recognized by many astute observers, that adequate geo- logical definition for open pit slope design. In: Proceedings of
logical insight is often lacking in the modern design process. While it is the XIV Congreso Internacional de Energia y Recursos Minerals:
Slope Stability, 10–13 April, Sevilla

13
Putting Geological Focus Back into Rock Engineering Design

Carter TG (2019) A suggested visual approach for estimating Hoek- Feng X, Jimenez R (2014) Bayesian prediction of elastic modulus of
Brown mi for different rock types. In: Proceedings of the 14th intact rocks using their uniaxial compressive strength. Eng Geol
ISRM congress on rock mechanics and rock engineering, Sept. 173:32–40
13–18, Foz do Iguassu, Brazil, Paper #14356 Ganye J, Vasileiou A, Perras MA, Carter TG (2020) Influence of grain
Carter TG, Carvalho JL (2020) Suggested tensile test data interpre- size and interlock on intact rock strength. In: Proceedings of the
tation for estimating Hoek-Brown mi. In: Proceedings of the 54th US rock mechanics/geomechanics symposium, Golden,
54th US rock mechanics conf., Paper ARMA 201860, Golden, Colorado: ARMA Paper 20-2097
Colorado Hoek E (1983) Strength of jointed rock masses. 1983 Rankine Lecture.
Carter TG, Marinos V (2014) Use of GSI for rock engineering Geotechnique 33(186):187–223
design. In: Proceedings of the. 1st international conference on Hoek E (1994) Strength of rock and rock masses. News J Int Soc Rock
applied empirical design methods in mining, Lima-Perú, 9–11th Mech 2(2):4–16
June, 2014 Hoek E (1999) Putting numbers to geology—an engineer’s view-point.
Carter TG, Marinos V (2020) Characterization of rockmass strength Q J Eng Geol 32(1):1–19
based on rock matrix competence (to be submitted for publica- Hoek E, Brown ET (1980a) Empirical strength criterion for rock
tion in Rock Mechanics and Rock Engineering) masses. J Geotech Eng Div ASCE 106(GT9):1013–1035
Carter TG, Diederichs MS, Carvalho JL (2008) Application of modi- Hoek E, Brown ET (1980b) Underground excavations in rock. Institu-
fied Hoek-Brown transition relationships for assessing strength tion of Mining and Metallurgy, London, p 527
and post yield behaviour at both ends of the rock competence Hoek E, Brown ET (1988) The Hoek-Brown failure criterion—a 1988
scale. In: Proceedings of the 6th international symposium on update. In: Proceedings of the 15th Canadian rock mechanics
ground support in mining and civil engineering construction, symposium, pp 31–38
30 March–3 April 2008. Cape Town, South Africa, pp 37–59. J Hoek E, Brown ET (2018) The Hoek-Brown failure criterion and
S Afr Inst Min Metall 108:325–338 GSI—2018 edition. J Rock Mech Geotech Eng. https​://doi.
Carvalho JL, Carter TG, Diederichs MS (2007) An approach for pre- org/10.1016/j.jrmge​.2018.08.001
diction of strength and post yield behaviour for rock masses of Hoek E, Diederichs MS (2006) Empirical estimation of rock mass
low intact strength. In: Proceedings 1st Can-US rock mechanics modulus. Int J Rock Mech Min Sci 43:203–215
symposium, meeting society’s challenges & demands, Vancou- Hoek E, Marinos P (2007) A brief history of the development of
ver, Canada, pp 249–257 the Hoek-Brown failure criterion. Soils Rocks (São Paulo)
Castro LM, Carvalho J, Sá G (2013) Discussion on how to classify 30(2):85–92
and estimate strength of weak rock masses. In: Dight PM (ed) Hoek E, Wood D, Shah S (1992) A modified Hoek-Brown criterion for
Proceedings of the 2013 International Symposium on Slope jointed rock masses. In: Hudson JA (ed) Proceedings of the rock
Stability in Open Pit Mining and Civil Engineering, Australian characterization, Symp. Int. Soc. Rock Mech.: Eurock‘92. Lond
Centre for Geomechanics, Perth, pp 205–217 Brit Geotech Soc, 209–214
Coates DF (1970) Rock mechanics principles. Mines Branch Mono- Hoek E, Kaiser PK, Bawden WF (1995) Support of underground
graph 874, Mining Research Centre, Department of Energy, excavations in hard rock. Rotterdam: Balkema, reprinted by CRC
Mines and Resources, Ottawa, Canada Press. ISBN 9789054101871
Day JJ, Diederichs MS, Hutchinson DJ (2019) Composite geological Hoek E, Carranza-Torres CT, Corkum B (2002) Hoek-Brown failure
strength index approach with application to hydrothermal vein criterion—2002 edition. In: Bawden HRW, Curran J, Telesnicki M
networks and other intrablock structures in complex rockmasses. (eds) Proceedings of the North American Rock Mechanics Soci-
Geotech Geol Eng 37:5285–5314. https​://doi.org/10.1007/s1070​ ety (NARMS-TAC 2002). Mining Innovation and Technology,
6-019-00980​-4 Toronto, Canada, pp 267273
Deere DU (1964) Technical description of rock cores for engineering Hoek E, Marinos P, Marinos V (2005) Characterisation and engineer-
purposes. Rock Mech Eng Geol 1(1):17–22 ing properties of tectonically undisturbed but lithologically varied
Deere DU (1968) Geological considerations. In: Stagg K, Zienkie- sedimentary rock masses. Int J Rock Mech Min Sci 42(2):277–285
wicz OC (eds) Rock mechanics in engineering practice. Wiley, Hoek E, Carter TG, Diederichs MS (2013) Quantification of the geo-
London, pp 1–20 logical strength index chart. In: 47th US rock mechanics/geome-
Deere DU, Miller R (1966) Engineering classification and index chanics symposium, San Francisco: ARMA Paper 13-672
properties for intact rock. Tech. Report No AFNL—AFWL- Laubscher DH (1977) Geomechanics classification of jointed rock
TR-65-116. Air Force Weapons Lab., Kirtland Air Force Base, masses—mining applications. Trans Inst Min Metall 86:A1–A8
New Mexico Lauffer H (1958) Gebirgsklassifizierung fur den Stollenbau. Geol Bau-
Dershowitz WS, Herda H (1992) Interpretation of fracture spacing wesen 74:46–51
and intensity. In: Proceedings, 32nd US rock mechanics sym- Marinos V (2007) Geotechnical classification and engineering geo-
posium, Santa Fe, New Mexico logical behaviour of weak and complex rock masses in tunneling.
Dershowitz W, Doe T, Uchida M, Hermanson J (2003) Correlations Doctoral thesis, School of Civil Engineering, Geotechnical Engi-
between fracture size, transmissivity, and aperture. In: Essen, P. neering Department, National Technical University of Athens
Culligan, H. Einstein, and A. Whittle (eds) Soil Rock America. (NTUA), Athens (In Greek)
Proceedings of the 39th US rock mechanics symposium, Cam- Marinos V (2014) Tunnel behaviour and support associated with the
bridge, MA. VGE, pp 887–891 weak rock masses of flysch. J Rock Mech Geotech Eng 6:227–239
Diederichs MS, Carvalho JL, Carter TG (2007) A modified approach Marinos V (2017) A revised, geotechnical classification GSI system
for prediction of strength and post yield behaviour for high GSI for tectonically disturbed heterogeneous rock masses, such as
rock masses in strong, brittle ground. In: Proceedings of 1st flysch. Bull Eng Geol Environ. https​://doi.org/10.1007/s1006​
Can-US rock symposium. Meeting Society’s Challenges and 4-017-1151-z
Demands, June, Vancouver, pp 277–285 Marinos V, Carter TG (2018) Maintaining geological reality in appli-
Douglas KJ (2002) The shear strength of rock masses. Unpublished cation of GSI for design of engineering structures in rock. J Eng
Doctoral thesis. University of New South Wales, Sydney, Geol 239:282–297. Corrigendum to Maintaining geological real-
Australia ity in application of GSI for design of engineering structures in
Rock

13
 T. G. Carter, V. Marinos

Marinos P, Hoek E (2000) GSI: a geologically friendly tool for rock- Priest SD, Hudson JA (1976) Discontinuity spacings in rock. Int J Rock
mass strength estimation. In: Proceedings of the GeoEng2000 Mech Min Sci 13:135–148
at the international conference on geotechnical and geological Riedel W (1929) Zur Mechanik Geologischer Brucherscheinungen.
engineering, Melbourne, Technomic publishers, Lancaster, Penn- Zentral-blatt fur Mineralogie, Geologie und Paleontologie B,
sylvania, pp 1422–1446 354–368
Marinos P, Hoek E (2001) Estimating the geotechnical properties of Rocscience (2018) https​://www.rocsc​ience​.com/help/rs3/Geolo​gy/
heterogeneous rock masses such as flysch. Bull Eng Geol Environ Mater​ials/Param​eter_Calcu​lator​.htm (in RocData Program)
60:82–92 Romana M (1985) New adjustment ratings for application of Bieniaw-
Marinos V, Marinos P, Hoek E (2005) The geological strength index— ski classification to slopes. In: International symposium on the
applications and limitations». Bull Eng Geol Environ 64(1):55–65 role of rock mechanics ISRM, Zacatecas, pp 49–53
Marinos P, Marinos V, Hoek E (2007) Geological Strength Index Rovida A, Tibaldi A (2005) Propagation of strike-slip faults across
(GSI). A characterization tool for assessing engineering properties holocene volcano-sedimentary deposits, Pasto, Colombia. J Struct
for rock masses » . In: Romana, Perucho & Olalla (eds) Proceed- Geol 27:1838–1855
ings of the international workshop on rock mass classification in Schlotfeldt P, Carter TG (2019) A new and unified approach to
underground mining held in 1st Canada-US rock mechanics sym- improved scalability and volumetric fracture intensity quantifica-
posium, Also permitted to be published in: Underground works tion for GSI and rockmass strength and deformability estimation.
under special conditions. Taylor and Francis Publishing, Lisbon, Int J Rock Mech Min Sci 110:48–67
pp 13–21 Sheorey PR (1997) Empirical rock failure criteria, 1st edn. A.A.
Mathews KE, Hoek E, Wyllie DC, Stewart SBV (1981) Golder Associ- Balkema, Rotterdam
ates Report 802-1571 to CanMet on prediction of stable excava- Skempton AW (1966) Some observations on tectonic shear zones. In:
tion spans for mining at depths below 1000 metres in Hard Rock. Congress of the international society for rock mechanics 1st pro-
CanMet Contract No.17SQ.23440-0-9020 ceedings, vol 1, pp 329–335
Muir-Wood AM (2000) Tunnelling: management by design. Pub- Sonmez H, Ulusay R (1999) Modifications to the Geological Strength
lished by E & FN Spon. ISBN 0-419-23200. https ​ : //doi. Index (GSI) and their applicability to Stability of Slopes. Int J
org/10.4324/97802​03477​663 Rock Mech Min Sci 36:743–760
Murphy WL (1985) Geotechnical descriptions of rock and rock masses. Tchalenko JS (1970) Similarities between shear-zones of different mag-
Technical Report GL85-3, Geotechnical Laboratory, Department nitudes. Geol Soc Am Bull 81:1625–1640
of the Army, Waterways Experiment Station, Corps of Engineers Terzaghi K (1946) Rock defects and load on tunnel supports, intro-
PO Box 631 Vicksburg, Mississippi 39180-0631 duction to rock tunnelling with steel supports, a book by Proc-
Palmstrom A (1982) The volumetric joint count - A useful and simple tor, R.V. and White, T.L. Commercial Shearing & Stamping Co,
measure of the degree of rock mass jointing. IAEG Congress, New Youngtown
Delhi, pp V.221–V.228
Peck RB (1976) Rock foundations for structures. In: Proceedings of Publisher’s Note Springer Nature remains neutral with regard to
the specialty conference on rock engineering for foundations and jurisdictional claims in published maps and institutional affiliations.
slopes, vol 2, Boulder, Colorado, pp 1–21
Potvin Y, Hudyma MR, Miller HDS (1989) Design guidelines for open
stope support. CIM Bull v.82, #926

13