COPYRIGHT NOTICE

The following document is subject to copyright agreements. The attached copy is provided for your personal use on the understanding that you will not distribute it and that you will not include it in other published documents.

Dr Evert Hoek Evert Hoek Consulting Engineer Inc. 3034 Edgemont Boulevard P.O. Box 75516 North Vancouver, B.C. Canada V7R 4X1 Email: ehoek@mailas.com

V. Marinos

P. Marinos

E. Hoek

The geological strength index: applicationsand limitations

Received: 7 September 2004Accepted: 9 October 2004Published online: 2 February 2005� Springer-Verlag 2005

Abstract The geological strength in-dex (GSI) is a system of rock-masscharacterization that has beendeveloped in engineering rockmechanics to meet the need for reli-able input data, particularly thoserelated to rock-mass properties re-quired as inputs into numericalanalysis or closed form solutions fordesigning tunnels, slopes or founda-tions in rocks. The geological char-acter of rock material, together withthe visual assessment of the mass itforms, is used as a direct input to theselection of parameters relevant forthe prediction of rock-mass strengthand deformability. This approachenables a rock mass to be consideredas a mechanical continuum withoutlosing the influence geology has on itsmechanical properties. It also pro-vides a field method for characteriz-ing difficult-to-describe rock masses.After a decade of application of theGSI and its variations in quantitativecharacterization of rock mass, thispaper attempts to answer questionsthat have been raised by the usersabout the appropriate selection of theindex for a range of rock masses un-der various conditions. Recommen-dations on the use of GSI are givenand, in addition, cases where the GSIis not applicable are discussed. Moreparticularly, a discussion and sug-gestions are presented on issues suchas the size of the rock mass to beconsidered, its anisotropy, the influ-ence of great depth, the presence of

ground water, the aperture and theinfilling of discontinuities and theproperties of weathered rock massesand soft rocks.

Resume Le Geological Strength In-dex (GSI) est un systeme de classifi-cation des massifs rocheuxdeveloppe en mecanique des roches.Il permet d’obtenir les donnees rel-atives aux proprietes de massesrocheuses, donnees necessaires pourdes simulations numeriques ou per-mettant le dimensionnement d’ouv-rages:tunnels, pentes ou fondationsrocheuses. Les caracteristiquesgeologiques de la matrice rocheuseainsi que celles relatives a la struc-ture du massif correspondant sontdirectement utilisees pour obtenir lesparametres appropries relatifs a ladeformabilite et la resistance de lamasse rocheuse. Cette approchepermet de considerer une masserocheuse comme un milieu continu,le role des caracteristiques geologi-ques sur les proprietes mecaniquesn’etant pas oblitere. Elle apporteaussi une methode de terrain pourcaracteriser des masses rocheusesdifficiles a decrire. Apres une decen-nie d’application du GeologicalStrength Index et de ses variantespour caracteriser des masses roche-uses, cet article tente de repondreaux questions formulees par lesutilisateurs concernant le choix leplus approprie de cet index pour unelarge gamme de massifs rocheux.

Bull Eng Geol Environ (2005) 64: 55–65DOI 10.1007/s10064-004-0270-5 ORIGINAL PAPER

V. Marinos (&) Æ P. MarinosSchool of Civil Engineering,Geotechnical Department,National Technical University of Athens,9 Iroon Polytechniou str.,157 80 Athens, GreeceE-mail: vmarinos@central.ntua.grE-mail: marinos@central.ntua.gr

E. HoekConsulting Engineer,Vancouver, CanadaE-mail: ehoek@attglobal.net

Introduction

Design in rock masses

A few decades ago, the tools for designing tunnelsstarted to change. Although still crude, numericalmethods were being developed that offered the promisefor much more detailed analysis of difficult undergroundexcavation problems which, in a number of cases, falloutside the ideal range of application of the tunnelreinforcement classifications such as the RMR systemintroduced by Bieniawski (1973) and the Q systempublished by Barton et al. (1974) both furthermore ex-panded in the following years. There is absolutely noproblem with the concept of these classifications andthere are hundreds of kilometres of tunnels that havebeen successfully constructed on the basis of theirapplication. However, this approach is ideally suited tosituations in which the rock mass behaviour is relativelysimple, for example for RMR values between about 30–70 and moderate stress levels. In other words, slidingand rotation of intact rock pieces essentially control thefailure process. These approaches are less reliable forsqueezing, swelling, clearly defined structural failures orspalling, slabbing and rock-bursting under very highstress conditions. More importantly, these classificationsystems are of little help in providing information for thedesign of sequentially installed temporary reinforcementand the support required to control progressive failure indifficult tunnelling conditions.

Numerical tools available today allow the tunneldesigner to analyse these progressive failure processesand the sequentially installed reinforcement and supportnecessary to maintain the stability of the advancingtunnel until the final reinforcing or supporting structurecan be installed. However, these numerical tools requirereliable input information on the strength and defor-mation characteristics of the rock mass surrounding thetunnel. As it is practically impossible to determine thisinformation by direct in situ testing (except for back-analysis of already constructed tunnels) there was a needfor some method for estimating the rock-mass propertiesfrom the intact rock properties and the characteristics of

the discontinuities in the rock mass. This resulted in thedevelopment of the rock-mass failure criterion by Hoekand Brown (1980).

The Geological Strength Index (GSI): developmenthistory

Hoek and Brown recognized that a rock-mass failurecriterion would have no practical value unless it could berelated to geological observations that could be madequickly and easily by an engineering geologist or geol-ogist in the field. They considered developing a newclassification system during the evolution of the criterionin the late 1970s but they soon gave up the idea andsettled for the already published RMR system. It wasappreciated that the RMR system (and the Q system)were developed for the estimation of undergroundexcavation and support, and that they included param-eters that are not required for the estimation of rock-mass properties. The groundwater and structuralorientation parameters in RMR and the groundwaterand stress parameters in Q are dealt with explicitly ineffective stress numerical analyses and the incorporationof these parameters into the rock-mass property estimateresults is inappropriate. Hence, it was recommendedthat only the first four parameters of the RMR system(intact rock strength, RQD rating, joint spacing andjoint conditions) should be used for the estimation ofrock-mass properties, if this system had to be used.

In the early days the use of the RMR classification(modified as described above) worked well because mostof the problems were in reasonable quality rock masses(30<RMR<70) under moderate stress conditions.However, it soon became obvious that the RMR systemwas difficult to apply to rock masses that are of verypoor quality. The relationship between RMR and theconstants m and s of the Hoek–Brown failure criterionbegins to break down for severely fractured and weakrock masses.

Both the RMR and the Q classifications include andare heavily dependent upon the RQD classificationintroduced by Deere (1964). Since RQD in most of the

Des recommandations quant al’usage du GSI sont donnees et, deplus, des cas ou le GSI n’est pasapplicable sont discutes. Plus par-ticulierement, des suggestions sontapportees sur des questions relativesa la taille de masse rocheuse a con-siderer, son anisotropie, l»influencedes grandes profondeurs, la presence

d’eau, l’ouverture et le remplissagedes discontinuites ainsi que les pro-prietes des masses rocheuses altereeset des roches tendres.

Keywords Geological StrengthIndex Æ Rock mass Æ Geologicalstructure Æ Mechanical properties ÆSelection of the GSI

Mots cles Geological StrengthIndex Æ Massif rocheux Æ Structuregeologique Æ Proprietes mecaniques ÆConditions d»utilisation du GSI

56

weak rock masses is essentially zero or meaningless, itbecame necessary to consider an alternative classifica-tion system. The required system would not includeRQD, would place greater emphasis on basic geologicalobservations of rock-mass characteristics, reflect thematerial, its structure and its geological history andwould be developed specifically for the estimation ofrock mass properties rather than for tunnel reinforce-ment and support. This new classification, now calledGSI, started life in Toronto with engineering geologyinput from David Wood (Hoek et al. 1992). The index

and its use for the Hoek and Brown failure criterion wasfurther developed by Hoek (1994), Hoek et al. (1995)and Hoek and Brown (1997) but it was still a hard rocksystem roughly equivalent to RMR. Since 1998, EvertHoek and Paul Marinos, dealing with incredibly difficultmaterials encountered in tunnelling in Greece, developedthe GSI system to the present form to include poorquality rock masses (Fig. 1) (Hoek et al. 1998; Marinosand Hoek 2000, 2001). They also extended its applica-tion for heterogeneous rock masses as shown in Fig. 2(Marinos and Hoek 2001).

Fig. 1 General chart for GSIestimates from the geologicalobservations

57

Functions of the Geological Strength Index

The heart of the GSI classification is a careful engi-neering geology description of the rock mass which isessentially qualitative, because it was felt that the num-bers associated with RMR and Q-systems were largelymeaningless for the weak and heterogeneous rock mas-ses. Note that the GSI system was never intended as areplacement for RMR or Q as it has no rock-massreinforcement or support design capability—its onlyfunction is the estimation of rock-mass properties.

This index is based upon an assessment of thelithology, structure and condition of discontinuity sur-

faces in the rock mass and it is estimated from visualexamination of the rock mass exposed in outcrops, insurface excavations such as road cuts and in tunnel facesand borehole cores. The GSI, by combining the twofundamental parameters of the geological process, theblockiness of the mass and the conditions of disconti-nuities, respects the main geological constraints thatgovern a formation and is thus a geologically soundindex that is simple to apply in the field.

Once a GSI ‘‘number’’ has been decided upon, thisnumber is entered into a set of empirically developedequations to estimate the rock-mass properties whichcan then be used as input into some form of numerical

Fig. 2 Geological strengthindex estimates for heteroge-neous rock masses such asFlysch

58

analysis or closed-form solution. The index is used inconjunction with appropriate values for the unconfinedcompressive strength of the intact rock rci and the pet-rographic constant mi, to calculate the mechanicalproperties of a rock mass, in particular the compressivestrength of the rock mass (rcm) and its deformationmodulus (E). Updated values of mi, can be found inMarinos and Hoek (2000) or in the RocLab program.Basic procedures are explained in Hoek and Brown(1997) but a more recent refinement of the empiricalequations and the relation between the Hoek–Brownand the Mohr–Coulomb criteria have been addressed byHoek et al. (2002) for appropriate ranges of stressencountered in tunnels and slopes. This paper and theassociated program RocLab can be downloaded fromhttp://www.rocscience.com.

Note that attempts to ‘‘quantify’’ the GSI classifica-tion to satisfy the perception that ‘‘engineers are happierwith numbers’’ (Cai et al. 2004; Sonmez and Ulusay1999) are interesting but have to be applied with caution.The quantification processes used are related to thefrequency and orientation of discontinuities and arelimited to rock masses in which these numbers can easilybe measured. The quantifications do not work well intectonically disturbed rock masses in which the struc-tural fabric has been destroyed. In such rock masses theauthors recommend the use of the original qualitativeapproach based on careful visual observations.

Suggestions for using GSI

After a decade of application of the GSI and its varia-tions for the characterization of the rock mass, this pa-per attempts to answer questions that have been raisedby users about the appropriate selection of the index forvarious rock masses under various conditions.

When not to use GSI

The GSI classification system is based upon theassumption that the rock mass contains a sufficientnumber of ‘‘randomly’’ oriented discontinuities suchthat it behaves as an isotropic mass. In other words, thebehaviour of the rock mass is independent of thedirection of the applied loads. Therefore, it is clear thatthe GSI system should not be applied to those rockmasses in which there is a clearly defined dominantstructural orientation. Undisturbed slate is an exampleof a rock mass in which the mechanical behaviour ishighly anisotropic and which should not be assigned aGSI value based upon the charts presented in Figs. 1, 2.However, the Hoek–Brown criterion and the GSI chartcan be applied with caution if the failure of such rockmasses is not controlled by their anisotropy (e.g. in the

case of a slope when the dominant structural disconti-nuity set dips into the slope and failure may occurthrough the rock mass). For rock masses with a struc-ture such as that shown in the sixth (last) row of the GSIchart (Fig. 1), anisotropy is not a major issue as thedifference in the strength of the rock and that of thediscontinuities within it is small.

It is also inappropriate to assign GSI values toexcavated faces in strong hard rock with a few discon-tinuities spaced at distances of similar magnitude to thedimensions of the tunnel or slope under consideration.In such cases the stability of the tunnel or slope will becontrolled by the three-dimensional geometry of theintersecting discontinuities and the free faces created bythe excavation. Obviously, the GSI classification doesnot apply to such cases.

Geological description in the GSI chart

In dealing with specific rock masses it is suggested thatthe selection of the appropriate case in the GSI chartshould not be limited to the visual similarity with thesketches of the structure of the rock mass as they appearin the charts. The associated descriptions must also beread carefully, so that the most suitable structure ischosen. The most appropriate case may well lie at someintermediate point between the limited number of sket-ches or descriptions included in the charts.

Projection of GSI values into the ground

Outcrops, excavated slopes tunnel faces and boreholecores are the most common sources of information forthe estimation of the GSI value of a rock mass. Howshould the numbers estimated from these sources beprojected or extrapolated into the rock mass behind aslope or ahead of a tunnel?

Outcrops are an extremely valuable source of data inthe initial stages of a project but they suffer from thedisadvantage that surface relaxation, weathering and/oralteration may have significantly influenced the appear-ance of the rock-mass components. This disadvantagecan be overcome (where permissible) by trial trenchesbut, unless these are machine excavated to considerabledepth, there is no guarantee that the effects of deepweathering will have been eliminated. Judgement istherefore required in order to allow for these weatheringand alteration effects in assessing the most probable GSIvalue at the depth of the proposed excavation.

Excavated slope and tunnel faces are probably themost reliable source of information for GSI estimatesprovided that these faces are reasonably close to and inthe same rock mass as the structure under investigation.In hard strong rock masses it is important that an

59

appropriate allowance be made for damage due tomechanical excavation or blasting. As the purpose ofestimating GSI is to assign properties to the undisturbedrock mass in which a tunnel or slope is to be excavated,failure to allow for the effects of blast damage whenassessing GSI will result in the assignment of values thatare too conservative. Therefore, if borehole data areabsent, it is important that the engineering geologist orgeologist attempts to ‘‘look behind’’ the surface damageand try to assign the GSI value on the basis of theinherent structures in the rock mass. This problem be-comes less significant in weak and tectonically disturbedrock masses as excavation is generally carried out by‘‘gentle’’ mechanical means and the amount of surfacedamage is negligible compared to that which alreadyexists in the rock mass.

Borehole cores are the best source of data at depth,but it has to be recognized that it is necessary toextrapolate the one-dimensional information providedby the core to the three-dimensional in situ rock mass.However, this is a problem common to all boreholeinvestigations, and most experienced engineering geolo-gists are comfortable with this extrapolation process.Multiple boreholes and inclined boreholes can be ofgreat help in the interpretation of rock-mass character-istics at depth.

For stability analysis of a slope, the evaluation isbased on the rock mass through which it is anticipatedthat a potential failure plane could pass. The estimationof GSI values in these cases requires considerable judg-ment, particularly when the failure plane can passthrough several zones of different quality. Mean valuesmay not be appropriate in this case.

For tunnels, the index should be assessed for thevolume of rock involved in carrying loads, e.g. for aboutone diameter around the tunnel in the case of tunnelbehaviour or more locally in the case of a structure suchas an elephant foot.

For particularly sensitive or critical structures, suchas underground powerhouse caverns, the informationobtained from the sources discussed above may not beconsidered adequate, particularly as the design advancesbeyond the preliminary stages. In these cases, the use ofsmall exploration tunnels can be considered and thismethod of data gathering will often be found to behighly cost effective.

Figure 3 provides a visual summary of some of theadjustments discussed in the previous paragraphs. Whendirect assessment of depth conditions is not available,upward adjustment of the GSI value to allow for theeffects of surface disturbance, weathering and alterationare indicated in the upper (white) part of the GSI chart.Obviously, the magnitude of the shift will vary from caseto case and will depend upon the judgement and expe-rience of the observer. In the lower (shaded) part of thechart, adjustments are not normally required as the rock

mass is already disintegrated or sheared and this damagepersists with depth.

Anisotropy

As discussed above, the Hoek–Brown criterion (andother similar criteria) requires that the rock mass behaveisotropically and that failure does not follow a prefer-ential direction imposed by the orientation of a specificdiscontinuity or a combination of two or three discon-tinuities. In these cases, the use of GSI is meaningless asthe failure is governed by the shear strength of thesediscontinuities and not of the rock mass. Cases, how-ever, where the criterion and the GSI chart can rea-sonably be used were discussed above.

However, in a numerical analysis involving a singlewell-defined discontinuity such as a shear zone or fault,it is sometimes appropriate to apply the Hoek–Browncriterion to the overall rock mass and to superimpose thediscontinuity as a significantly weaker element. In thiscase, the GSI value assigned to the rock mass shouldignore the single major discontinuity. The properties ofthis discontinuity may fit the lower portion of the GSIchart or they may require a different approach such aslaboratory shear testing of soft clay fillings.

Aperture of discontinuities

The strength and deformation characteristics of a rockmass are dependent upon the interlocking of the indi-vidual pieces of intact rock that make up the mass.Obviously, the aperture of the discontinuities that sep-arate these individual pieces has an important influenceupon the rock-mass properties.

There is no specific reference to the aperture of thediscontinuities in the GSI charts but a ‘‘disturbancefactor’’ D has been provided in the most recent versionof the Hoek–Brown failure criterion (Hoek et al. 2002).This factor ranges from D=0 for undisturbed rockmasses, such as those excavated by a tunnel boringmachine, to D=1 for extremely disturbed rock massessuch as open pit mine slopes that have been subjected tovery heavy production blasting. The factor allows forthe disruption of the interlocking of the individual rockpieces as a result of opening of the discontinuities.

The incorporation of the disturbance factor D intothe empirical equations used to estimate the rock-massstrength and deformation characteristics is based uponback-analysis of excavated tunnels and slopes. At thisstage (2004) there is relatively little experience in the useof this factor, and it may be necessary to adjust itsparticipation in the equations as more field evidenceis accumulated. However, the limited experience thatis available suggests that this factor does provide a

60

reasonable estimate of the influence of damage due tostress relaxation or blasting of excavated rock faces.

Note that this damage decreases with depth into therock mass and, in numerical modelling, it is generallyappropriate to simulate this decrease by dividing therock mass into a number of zones with decreasing valuesof D being applied to successive zones as the distancefrom the face increases. In one example, which involvedthe construction of a large underground powerhousecavern in interbedded sandstones and siltstones, it wasfound that the blast damaged zone was surrounding

each excavation perimeter to a depth of about 2 m(Cheng and Liu 1990). Carefully controlled blasting wasused in this cavern excavation and the limited extent ofthe blast damage can be considered typical of that forcivil engineering tunnels excavated by drill and blastmethods. On the other hand, in very large open pit mineslopes in which blasts can involve many tons of explo-sives, blast damage has been observed up to 100 m ormore behind the excavated slope face. Hoek and Karz-ulovic (2000) have given some guidance on the extent ofthis damage and its impact on rock mass properties.

Fig. 3 Suggested projection ofinformation from observationsin outcrops to depth. Whitearea: a shifting to the left or tothe left and upwards is recom-mended; the extent of the shiftshown in the chart is indicativeand should be based on geo-logical judgement. Shadowedarea: shifting is less or notapplicable as poor quality isretained in depth in brecciated,mylonitized or shear zones

61

Geological Strength Index at great depth

In hard rock, great depth (e.g. 1,000 m or more) therock-mass structure is so tight that the mass behaviourapproaches that of the intact rock. In this case, the GSIvalue approaches 100 and the application of the GSIsystem is no longer meaningful.

The failure process that controls the stability ofunderground excavations under these conditions isdominated by brittle fracture initiation and propagation,which leads to spalling, slabbing and, in extreme cases,rock-bursts. Considerable research effort has been de-voted to the study of these brittle fracture processes anda recent paper by Diederichs et al. (2004) provides auseful summary of this work. Cundall et al. (2003) haveintroduced a set of post-failure flow rules for numericalmodelling which cover the transition from tensile toshear fracture that occurs during the process of brittlefracture propagation around highly stressed excavationsin hard rock masses.

When tectonic disturbance is important and persistswith depth, these comments do not apply and theGSI charts may be applicable, but should be used withcaution.

Discontinuities with filling materials

The GSI charts can be used to estimate the character-istics of rock-masses with discontinuities with fillingmaterials using the descriptions in the columns of pooror very poor condition of discontinuities. If the fillingmaterial is systematic and thick (e.g. more than few cm)or shear zones are present with clayey material then theuse of the GSI chart for heterogeneous rock masses(Fig. 2) is recommended.

The influence of water

The shear strength of the rock mass is reduced by thepresence of water in the discontinuities or the fillingmaterials when these are prone to deterioration as aresult of changes in moisture content. This is particularlyvalid in the fair to very poor categories of discontinuitieswhere a shift to the right may be made for wet condi-tions (Fig. 4).

Water pressure is dealt with by effective stress anal-ysis in design and it is independent of the GSI charac-terization of the rock mass.

Weathered rock masses

The GSI values for weathered rock masses are shifted tothe right of those of the same rock masses when these areunweathered. If the weathering has penetrated into theintact rock pieces that make up the mass (e.g. in weath-

ered granites) then the constant mi and the unconfinedstrength of the rci of the Hoek and Brown criterion mustalso be reduced. If the weathering has penetrated therock to the extent that the discontinuities and the struc-ture have been lost, then the rock mass must be assessedas a soil and the GSI system no longer applies.

Heterogeneous and lithologically varied sedimentaryrock masses

The GSI has recently been extended to accommodatesome of the most variable of rock masses, includingextremely poor quality sheared rock masses of weakschistose materials (such as siltstones, clay shales orphyllites) sometime inter-bedded with strong rock (suchas sandstones, limestones or quartzites). A GSI chart forflysch has been published in Marinos and Hoek (2001)and is reproduced in Fig. 2. For lithologically varied buttectonically undisturbed rock masses, such as themolasses, a new GSI chart is (Hoek et al. 2005).

Rocks of low strength

When rocks such as marls, claystones, siltstones andweak sandstones are developed in stable conditions or apost tectonic environment, they present a simple struc-ture with few discontinuities. Even when bedding planesexist they do not always appear as clearly defined dis-continuity surfaces.

In such cases, the use of theGSI chart for the ‘‘blocky’’or ‘‘massive’’ rock masses (Fig. 1) is applicable. The dis-continuities, although they are limited in number, cannotbe better than fair (usually fair or poor) and hence theGSIvalues tend to be in the range of 40–60. In these cases, thelow strength of the rock mass results from low values ofthe intact strength rci and the constant mi.

When these rocks form continuous masses with nodiscontinuities, the rock mass can be treated as intactwith engineering parameters given directly by laboratorytesting. In such cases the GSI classification is notapplicable.

Precision of the GSI classification system

The ‘‘qualitative’’ GSI system works well for engineeringgeologists since it is consistent with their experience indescribing rocks and rock masses during logging andmapping. In some cases, engineers tend to be uncom-fortable with the system because it does not containparameters that can be measured in order to improve theprecision of the estimated GSI value.

The authors, two of whom graduated as engineers, donot share this concern as they feel that it is not mean-ingful to attempt to assign a precise number to the GSI

62

value for a typical rock mass. In all but the very simplestof cases, GSI is best described by assigning it a range ofvalues. For analytical purposes this range may be definedby a normal distribution with the mean and standarddeviation values assigned on the basis of common sense.

In the earlier period of the GSI application it wasproposed that correlation of ‘‘adjusted’’ RMR and Qvalues with GSI be used for providing the necessaryinput for the solution of the Hoek and Brown criterion.Although this procedure may work with the betterquality rock masses, it is meaningless in the range of

weak (e.g. GSI<35), very weak and heterogeneous rockmasses where these correlations are not recommended.

Estimation of intact strength rci and the constant mi

While this paper is concerned primarily with the GSIclassification, it would not be appropriate to leave therelated topic of the Hoek–Brown failure criterion with-out briefly mentioning the estimation of intact strengthrci and the constant mi.

Fig. 4 In fair to very poorcategories of discontinuities, ashift to the right is necessary forwet conditions as the surfaces ofthe discontinuities or the fillingmaterials are usually prone todeterioration as a result ofchange in the moisture content.The shift to the right is moresubstantial in the low qualityrange of rock mass (last linesand columns)

63

The influence of the intact rock strength rci is at leastas important as the value of GSI in the overall estimateof rock mass properties by means of the Hoek–Browncriterion. Ideally, rci should be determined by directlaboratory testing under carefully controlled conditions.However, in many cases, this is not possible because oftime or budget constraints, or because it is not possibleto recover samples for laboratory testing (particularly inthe case of weak, thinly schistose or tectonically dis-turbed rock masses where discontinuities are included inthe laboratory samples). Under such circumstances,estimates of the value of rci have to be made on thebasis of published information, simple index tests or bydescriptive grades such as those published by theInternational Society for Rock Mechanics (Brown1981).

Experience has shown that there is a tendency tounderestimate the value of the intact rock strength inmany cases. This is particularly so in weak and tecton-ically disturbed rock masses where the characteristics ofthe intact rock components tend to be masked by thesurrounding sheared or weathered material. Theseunderestimations can have serious implications forengineering design and care has to be taken to ensurethat realistic estimates of intact strength are made asearly as possible in the project. In tunnelling, such esti-mates can be refined on the basis of a detailed back-analysis of the tunnel deformation and, while this mayrequire considerable effort and even the involvement ofspecialists in numerical analysis, the attempt will gen-erally be repaid many times over in the cost savingsachieved by more realistic designs.

The value of the constant mi, as for the case of theintact strength rci, is best determined by direct labora-tory testing. However, when this is not possible, anestimate based upon published values (e.g. in the pro-gram RocLab) is generally acceptable as the overallinfluence of the value of mi on the rock-mass strength issignificantly less than that of either GSI or rci.

GSI and contract documents

One of the most important contractual problems in rockconstruction and particularly in tunnelling is the issue of‘‘changed ground conditions’’. There are invariablyarguments between the owner and the contractor on thenature of the ground specified in the contract and thatactually encountered during construction. In order toovercome this problem there has been a tendency tospecify the anticipated conditions in terms of the RMRor Q tunnelling classifications. More recently somecontracts have used the GSI classification for this pur-pose, and the authors are strongly opposed to this trend.As discussed earlier in this paper, RMR and Q weredeveloped for the purposes of estimating tunnel rein-

forcement or support whereas GSI was developed solelyfor the purpose of estimating rock-mass strength.Therefore, GSI is only one element in a tunnel designprocess and cannot be used, on its own, to specify tun-nelling conditions.

The use of any classification system to specify antic-ipated tunnelling conditions is always a problem as thesesystems are open to a variety of interpretations,depending upon the experience and level of conservatismof the observer. This can result in significant differencesin RMR or Q values for a particular rock mass and, ifthese differences fall on either side of a major ‘‘change’’point in excavation or support type, this can haveimportant financial consequences.

The geotechnical baseline report (Essex 1997)was introduced in an attempt to overcome some ofthese difficulties and has attracted an increasingamount of international attention in tunnelling1. Thisreport, produced by the Owner and included inthe contract documents, attempts to describe therock mass and the anticipated tunnelling conditionsas accurately as possible and to provide a rationalbasis for contractual discussions and payment. The au-thors of this paper recommend that this concept shouldbe used in place of the traditional tunnel classificationsfor the purpose of specifying anticipated tunnelconditions.

Conclusions

Rock-mass characterization has an important role in thefuture of engineering geology in extending its usefulness,not only to define a conceptual model of the site geol-ogy, but also for the quantification needed for analyses‘‘to ensure that the idealization (for modelling) does notmisinterpret actuality’’ (Knill 2003). If it is carried out inconjunction with numerical modelling, rock-mass char-acterization presents the prospect of a far better under-standing of the reasons for rock-mass behaviour(Chandler et al. 2004). The GSI has considerable po-tential for use in rock engineering because it permits themanifold aspects of rock to be quantified therebyenhancing geological logic and reducing engineeringuncertainty. Its use allows the influence of variables,which make up a rock mass, to be assessed and hence thebehaviour of rock masses to be explained more clearly.One of the advantages of the index is that the geologicalreasoning it embodies allows adjustments of its ratingsto cover a wide range of rock masses and conditionsbut it also allows us to understand the limits of itsapplication.

1A simple search for ‘‘geotechnical baseline report’’ on the Internetwill reveal the extent of this interest.

64

References

Barton NR, Lien R, Lunde J (1974) Engi-neering classification of rock masses forthe design of tunnel support. RockMech 6(4):189–239

Bieniawski ZT (1973) Engineering classifi-cation of jointed rock masses. Trans SAfr Inst Civ Eng 15:335–344

Brown ET (ed) (1981) Rock characteriza-tion, testing and monitoring—ISRMsuggested methods. Pergamon, Oxford,pp 171–183

Cai M, Kaiser PK, Uno H, Tasaka Y,Minami M (2004) Estimation of rockmass strength and deformation modu-lus of jointed hard rock masses usingthe GSI system. Int J Rock Mech MinSci 41(1):3–19

Chandler RJ, de Freitas MH, Marinos P(2004) Geotechnical characterization ofsoils and rocks: a geological perspective.Advances in geotechnical engineering:the Skempton conference, vol 1, Tho-mas Telford, London, pp 67–102

Cheng Y, Liu SC (1990) Power caverns ofthe Mingtan Pumped Storage Project,Taiwan. In: JA Hudson (ed) Compre-hensive Rock Engineering, vol 5, pp111–132

Cundall P, Carranza-Torres C, Hart R(2003) A new constitutive model basedon the Hoek–Brown criterion. In:Brummer et al. (eds) Proceedings of the3rd international symposium on FLACand FLAC3D numerical modelling inGeomechanics, Sudbury, October 21–24, pp 17–25

Deere DU (1964) Technical description ofrock cores for engineering purposes.Rock Mech Eng Geol 1(1):17–22

Diederichs MS, Kaiser PK, Eberhardt E(2004) Damage initiation and propaga-tion in hard rock during tunnelling andthe influence of near-face stress rota-tion. Int J Rock Mech Min Sci41(5):785–812

Essex RJ (1997) Geotechnical baseline re-ports for underground construction.American Society of Civil Engineers,Reston

Hoek E (1994) Strength of rock and rockmasses. News J ISRM 2(2):4–16

Hoek E, Brown ET (1980) Undergroundexcavations in rock. Institution ofMining and Metallurgy, London

Hoek E, Brown ET (1997) Practical esti-mates of rock mass strength. Int J RockMech Min Sci Geomech Abstr 34:1165–1186

Hoek E, Karzulovic A (2000) Rock massproperties for surface mines. In: Hu-stralid WA, McCarter MK, van ZylDJA (eds) Slope stability in surfacemining. Society for Mining, Metallur-gical and Exploration (SME), Littleton,pp 59–70

Hoek E, Wood D, Shah S (1992) A modi-fied Hoek–Brown criterion for jointedrock masses. In: Hudson JA (ed) Pro-ceedings of the rock mechanic sympo-sium. International Society of RockMechanics Eurock’’ 92, British Geo-technical Society, London, pp 209–214

Hoek E, Marinos P, Marinos V (2005)Characterization and engineering prop-erties of tectonically undisturbed butlithologically varied sedimentary rockmasses under publication. Int J RockMech Min Sci

Hoek E, Kaiser PK, Bawden WF (1995)Support of underground excavations inhard rock. AA Balkema, Rotterdam

Hoek E, Marinos P, Benissi M (1998)Applicability of the geological strengthindex (GSI) classification for weak andsheared rock masses—the case of theAthens schist formation. Bull Eng GeolEnv 57(2):151–160

Hoek E, Caranza-Torres CT, Corcum B(2002) Hoek–Brown failure criterion-2002 edition. In: Bawden HRW, CurranJ, Telsenicki M (eds) Proceedings of theNorth American Rock MechanicsSociety (NARMS-TAC 2002). MiningInnovation and Technology, Toronto,pp 267–273

Knill J (2003) Core values (1st Hans-Closslecture). Bull Eng Geol Env 62:1–34

Marinos P, Hoek E (2000) GSI: a geologi-cally friendly tool for rock massstrength estimation. In: Proceedings ofthe GeoEng2000 at the internationalconference on geotechnical and geolog-ical engineering, Melbourne, Technom-ic publishers, Lancaster, pp 1422–1446

Marinos P, Hoek E (2001) Estimating thegeotechnical properties of heteroge-neous rock masses such as flysch. BullEng Geol Env 60:82–92

Sonmez H, Ulusay R (1999) Modificationsto the geological strength index (GSI)and their applicability to the stability ofslopes. Int J Rock Mech Min Sci36:743–760

65