P.K. RobertsonGregg Drilling & Testing Inc., Signal Hill, California, USA
ABSTRACT: Application of the Cone Penetration Test (CPT) for the evaluation of seismic performance isreviewed and updates presented. The role of the CPT in geotechnical earthquake engineering is presented. Theuse of the CPT to identify soil behavior type and the normalization of CPT parameters is also reviewed andupdates presented. The case-history based method to evaluate the resistance of sand-like soils to cyclic loading isreviewed and compared with the expanded and re-evaluated case history database. The laboratory based methodto evaluate the resistance of clay-like soils to cyclic loading is reviewed and modified for application using theCPT. A new combined CPT-based method to evaluate the resistance to cyclic loading is proposed that covers allsoils and is evaluated using an expanded case history database. The CPT-based method is extended to estimateboth volumetric and shear strains for all soils and evaluated using the expanded case history database.
1 INTRODUCTION
The seismic performance of geotechnical structuresoften requires an estimate of potential post-earthquakedisplacements. Historically, geotechnical earthquakedesign has focused extensively on evaluation of liq-uefaction in sandy soils since deformations tend to belarge when soils experience liquefaction. Liquefactionanalyses have traditionally focused on the evaluation offactor of safety and using this as an indicator of poten-tial post-earthquake deformations. Recently there hasbeen growing awareness that soft clays can also deformduring earthquake loading.
In North American building codes (e.g. NBC 2005,FEMA 356 and SEAOC 1995), the design philoso-phy for earthquake loading is to accept some level ofdamage to structures, i.e. to accept some level of defor-mation. The acceptable level of damage and deforma-tion is a function of the importance of the structureand the earthquake return period. The importance ofthe structure is a function of the risk. The evalua-tion of post-earthquake deformations is therefore a keyelement in any performance based earthquake design.
Due to size limitations, this paper will only discussthe application of the Cone Penetration Test (CPT) forthe evaluation of post-earthquake deformations. Theintent of this paper is not to imply that all earthquakegeotechnical design can be accomplished using onlythe CPT; other in-situ tests along with sampling andlaboratory testing also play a role, depending on therisk of the project.
2 ROLE OF CPT IN GEOTECHNICALEARTHQUAKE ENGINEERING
Since this paper is focused on the application of CPTresults for the evaluation of post-earthquake deforma-tions, it is appropriate to briefly discuss the role of theCPT in geotechnical earthquake engineering practice.Hight and Leroueil (2003) suggested that the appro-priate level of sophistication for a site characterizationand analyses program should be based on the followingcriteria:
• Precedent and local experience• Design objectives• Level of geotechnical risk• Potential cost savings
The evaluation of geotechnical risk was describedby Robertson (1998) and is dependent on hazards(what can go wrong), probability of occurrence (howlikely is it to go wrong) and the consequences (what arethe outcomes). Earthquake loading can be a significanthazard, but the resulting risk is primarily a function ofthe probability of occurrence and the consequences.General recommendations for the appropriate levelof sophistication for site investigation and subsequentdesign can be summarized in Table 1. Although Table 1indicates only two broad outcomes, Robertson (1998)and Lacasse and Nadim (1998) showed that the levelof risk cover a range from low to high and that theresulting site characterization program should varyaccordingly.
3
Table 1. Appropriate level of sophistication for site characterization and analyses.
Rating Criteria Rating
Good Precedent & local experience PoorSimple Design objectives ComplexLow Level of geotechnical risk HighLow Potential for cost savings HighLow risk project High risk projectTraditional Advanced
(simplified) methods (complex) methods
For low risk projects, traditional methods, such asin-situ logging tests (e.g. CPT, SPT) and index test-ing on disturbed samples combined with conservativedesign criteria, are often appropriate. For the evalua-tion of liquefaction and post-earthquake deformationsthe Simplified Procedure, first proposed by Seed andIdriss (1971) and recently updated by Youd et al.(2001), is appropriate for low risk projects. For mod-erate risk projects, the Simplified Procedure shouldbe supplemented with additional specific in-situ test-ing where appropriate, such as seismic CPT with porepressure measurements (SCPTu) and field vane tests(FVT) combined with selective sampling and basiclaboratory testing to develop site specific correla-tions. Sampling and laboratory testing is often limitedto fine-grained soils where conventional sampling iseasier and appropriate. For high risk projects, the Sim-plified Procedure can be used for screening to identifypotentially critical regions/zones appropriate to thedesign objectives. This should be followed by selec-tive high quality sampling and advanced laboratorytesting. The results of laboratory testing should becorrelated to in-situ test results to extend the resultsto other regions of the project. The Simplified Proce-dure for liquefaction evaluation should be used onlyas a screening technique to identify potentially crit-ical regions/zones for high risk projects. Advancedtechniques, such as numerical modeling, are oftenappropriate for more detailed evaluation of potentialpost-earthquake deformations for high risk projects.
One reason for the continued application of theStandard Penetration Test (SPT) as a basic loggingtest is that the test provides a soil sample suitable forindex testing, even though the test can be unreliable.A common complaint about the CPT is that it does notprovide a soil sample. Although it is correct that a soilsample is not obtained during the CPT, most commer-cial CPT operators have a simple push-in soil samplerthat can be pushed using the CPT installation equip-ment to obtain a small (typically 25 mm diameter)disturbed sample of similar size to that obtained fromthe SPT. Often the most cost effective solution is toobtain a detailed continuous stratigraphic profile usingthe CPT, then to move over a short distance (<1 m)
and push a small diameter sampler to obtain discreteselective soil samples in critical layers/zones that wereidentified by the CPT. Continuous push samplers are alsoavailable to collect plastic-lined near continuous smalldiameter, disturbed soil samples. The push rate toobtain soil samples can be significantly faster than the2 cm/s required for the CPT therefore making sam-pling rapid and cost effective for a small number ofdiscrete samples. For low risk projects the efficiencyand cost effectiveness of CPT, combined with adja-cent discrete push-in soil samples, is usually superiorto that of CPT plus adjacent boreholes with SPT.
Many of the comments and recommendations con-tained in this paper are focused on low to moderate riskprojects where traditional (simplified) procedures areappropriate and where empirical interpretations tendto dominate. For projects where more advanced proce-dures are appropriate, the recommendations providedin this paper can be used as a screening to evaluatecritical regions/zones where selective additional in-situ testing and sampling may be appropriate. Riskbased site investigation and analysis is consistent withperformance based design principles where the designcriteria are in terms of deformation based on the riskof the structure.
3 BASIC SOIL BEHAVIOR UNDEREARTHQUAKE LOADING
Boulanger and Idriss (2004b, 2007) showed that, forpractical purposes, soils can be divided into either‘sand-like’ or ‘clay-like’ soils, where sand-like soilscan experience ‘liquefaction’ and clay-like soils canexperience ‘cyclic failure’. In a general sense, sand-like soils are gravels, sands, and very-low plasticitysilts, whereas clay-like soils are clays and plastic silts.
In general, all soils deform under earthquake load-ing. Earthquakes impose cyclic loading rapidly andsoils respond undrained during the earthquake. Ingeneral, all soils develop some pore pressure duringearthquake loading and at small strains these pore pres-sures are almost always positive. Sand-like soils candevelop high positive pore pressures during undrained
4
cyclic loading and can reach a condition of zeroeffective confining stress. At the condition of zeroeffective stress, the initial structure of the soil is lostand the stiffness of the soil in shear is essentially zeroor very small and large deformations can occur duringearthquake loading. The condition of zero effectivestress is often defined as ‘liquefaction’ or ‘cyclic liq-uefaction’. Loose, young, uncemented sand-like soilsare more susceptible to ‘liquefaction’ than dense sand-like soils. The ability of sand-like soils to liquefy is afunction of in-situ state (relative density and effectiveconfining stress), structure (age, fabric and cementa-tion) and the size and duration of the cyclic loading.Most liquefaction cases occur in young uncementedsand-like soils. During earthquake loading, loosesand-like soils can experience very large shear strainswhich can result in large lateral and vertical defor-mations, depending on ground geometry and externalstatic loads (e.g. buildings, embankments, slopes,etc.). Very loose sand-like soils can also experiencestrength loss after earthquake loading that can resultin flow slides with very large deformations dependingon ground geometry and drainage. Following earth-quake loading, sand-like soils can also experiencevolumetric strains and post-earthquake reconsolida-tion settlements. The resulting volumetric strains canbe large due to the loss of initial soil structure at zeroeffective stress and resulting small volumetric stiffness(constrained modulus) during initial reconsolidation.These settlements generally occur rapidly after theearthquake (i.e. in less than a few hours), dependingon soil stratigraphy and drainage conditions.
Clay-like (cohesive) soils can also develop porepressures during undrained cyclic loading, but gener-ally do not reach zero effective stress and hence retainsome level of stiffness during cyclic loading and gen-erally deform less than sand-like soils. Traditionally,clay-like soils are considered not susceptible to lique-faction, since they generally do not reach a conditionof zero effective stress. However, clay-like soils candeform during cyclic earthquake loading. The amountof pore pressure buildup is a function of in-situ state(overconsolidation ratio), sensitivity, structure (age,fabric and cementation) and size and duration of cyclicloading. Soft normally to lightly overconsolidated andsensitive clay-like soils can develop large positivepore pressures with significant shear strains duringearthquake loading that can result in lateral and ver-tical deformations, depending on ground geometryand external static loads (e.g. buildings, embank-ments, slopes, etc.). Very sensitive clay-like soils canalso experience strength loss after earthquake loadingthat can result in flow slides with very large defor-mations depending on ground geometry. Followingearthquake loading, clay-like soils can also experiencevolumetric strains and post-earthquake reconsolida-tion settlements. However, these settlements generallyoccur slowly after the earthquake due to the lower
permeability of clay-like soils and are also a functionof soil stratigraphy and drainage conditions. The volu-metric strains during post-earthquake reconsolidationare generally small since clay-like soils often retainsome original soil structure and hence, maintain a highvalue of volumetric stiffness during reconsolidation.
Following earthquake loading, pore-water redistri-bution can result in some sand-like soils changing voidratio and becoming looser. This can result in strengthloss and the potential for instability.
Recent research has indicated that the transitionfrom sand-like to clay-like soils can be approximatelydefined by Atterberg Limits (e.g. plasticity index) ofthe soil (Seed et al, 2003; Bray and Sancio, 2006;Boulanger and Idriss, 2007). Sangrey et al. (1978)suggested that the transition was controlled by thecompressibility of the soil, where, in general, clay-likesoils have a higher compressibility than sand-like soils.In a general sense, soft normally consolidated clay-like fine grained soils respond in a similar manner toloose sand-like soils in that they are both contrac-tive under shear and develop positive pore pressuresin undrained shear. Highly sensitive clay-like soilsare similar to very loose sand-like soils in that bothcan experience a large increase in pore pressure underundrained shear and can experience significant strengthloss (i.e. strain soften). Stiff overconsolidated clay-likefine grained soils respond in a similar manner to densesand-like soils in that they both dilate under shear athigh strains. Soil response in fine grained soils is con-trolled partly by the amount and type of clay minerals.The plasticity index is an approximate measure of themineralogy of the soil, where the amount and type ofclay mineral influences soil behavior.
Traditionally, the response of sand-like and clay-likesoils to earthquake loading is evaluated using differentprocedures. It is common to first evaluate which soilsare sand-like, and therefore susceptible to liquefactionbased on grain size distribution and Atterberg Lim-its, and then to determine the factor of safety (FSliq)against liquefaction. A key element in performancebased geotechnical earthquake design is the evalua-tion of post-earthquake deformations. The challengeis to develop procedures that capture the correct soilresponse as soil transitions from primarily sand-liketo clay-like in nature. The objective of this paper is tooutline a possible unified approach for all soils usingCPT results with the ultimate goal to evaluate possiblepost-earthquake deformations.
4 CPT SOIL BEHAVIOUR TYPE
One of the major applications of the CPT has been thedetermination of soil stratigraphy and the identifica-tion of soil type. This has been accomplished usingcharts that link cone parameters to soil type. Earlycharts using qc and friction ratio (Rf ) were proposed
5
by Douglas and Olsen (1981), but the charts proposedby Robertson et al. (1986) have become popular.Initially these charts were based on empirical cor-relations, but theoretical studies have supported thegeneral concepts. Robertson et al. (1986) and Robert-son (1990) stressed that the charts were predictive ofSoil Behaviour Type (SBT) since the cone responds tothe mechanical behaviour of the soil and not directly tosoil classification criteria based on grain-size distribu-tion and soil plasticity. Fortunately, soil classificationcriteria based on grain-size distribution and plastic-ity often relate reasonably well to soil behaviour andhence, there is often good agreement between soilclassification based on samples and SBT based onthe CPT. Several examples can be given when dif-ferences arise between soil classification and SBTbased on CPT. For example, a soil with 60% sand and40% fines may be classified as ‘silty sand’ using theunified classification system. However, if the finesare composed of a highly active clay mineral withhigh plasticity, the soil behaviour may be controlledmore by the clay and the SBT from the CPT willreflect this behaviour and predict a more clay-likebehaviour, such as ‘clayey silt’. If the fines were non-plastic the soil behaviour may be controlled more bythe sand, the CPT SBT would predict a sand likesoil type, such as ‘silty sand’. Saturated loose siltsoften behave like soft clay in that their undrainedstrength is low and undrained response often gov-erns geotechnical design. Hence, SBT based on CPTin soft saturated silts is often defined as clay. Verystiff heavily overconsolidated fine-grained soils tendto behave similar to coarse-grained soil in that theydilate at large strains under shear and can have highundrained shear strength compared to their drainedstrength. These few examples illustrate that the SBTbased on the CPT may not always agree with tradi-tional classification based on samples. Geotechnicalengineers are usually interested in the behaviour of thesoil rather than a classification based only on grain-size distribution and plasticity, although knowledge ofboth is useful.
The corrected cone (tip) resistance (qt) responds tothe average shear strength (depending on soil sensitiv-ity, heterogeneity and macro fabric) of the soil aheadand behind the advancing cone, whereas the sleevefriction (fs) and measured pore pressure (u2) respondsto the larger strain behaviour of the soil in contact withthe cone. There is also a small scale effect and physicaloffset between the qt and fs measurements. Typicallymost commercially available CPT data acquisitionsystems adjust the two readings to present them at thesame depth in the soil profile (i.e. the fs reading isrecorded when the center of the sleeve has reached thesame depth/elevation as the cone tip). Soils with gravelparticles can produce rapid unrepresentative variationsin sleeve friction due to large particles touching thefriction sleeve.
where:σvo = pre-insertion in-situ total vertical stressσ ′
vo = pre-insertion in-situ effective vertical stressu0 = in-situ equilibrium water pressure�u = excess penetration pore pressure.
In the original paper by Robertson (1990) the nor-malized cone resistance was defined using the termQt. The term Qt1 is used here to show that the coneresistance is the corrected cone resistance, qt withthe stress exponent for stress normalization n = 1.0.Note that in clean sands, qc = qt, but the more correctqt is used in this paper.
In general, the normalized charts provide morereliable identification of SBT than the nonnormal-ized charts, although when the in-situ vertical effec-tive stress is between 50 kpa to 150 kpa there isoften little difference between normalized and non-normalized SBT. The term SBTn will be used todistinguish between normalized and non-normalizedSBT. Robertson (1990) suggested two charts basedon either Qt1 – Fr or Qt1 − Bq but recommended thatthe Qt1 – Fr chart was generally more reliable, espe-cially for onshore geotechnical investigations wherethe CPT pore pressure results are more problematicand less reliable.
Jefferies and Davies (1993) identified that a SoilBehaviour Type Index, Ic, could represent the SBTnzones in the Qt1 − Fr chart where Ic is essentially theradius of concentric circles that define the boundariesof soil type. Robertson and Wride (1998) modifiedthe definition of Ic to apply to the Robertson (1990)Qt1 – Fr chart, as defined by:
Contours of Ic are shown in Figure 1 on the Robert-son (1990) Qt1 – Fr SBTn chart. The contours of Ic canbe used to approximate the SBT boundaries.
Jefferies and Davies (1993) suggested that the SBTindex Ic could also be used to modify empirical cor-relations that vary with soil type. This is a powerfulconcept and has been used where appropriate in thispaper.
Robertson and Wride (1998) and updated by Zhanget al. (2002) suggested a normalized cone parameter,using normalization with a variable stress exponent, n,where:
Qtn = [(qt − σv)/pa](pa/σ′vo)
n (5)
where:(qt – σv)/pa = dimensionless net cone resistance,
6
Figure 1. Contours of soil behaviour type index, Ic onnormalized SBT Qtn − Fr chart.
(pa/σ′vo)
n = stress normalization factorn = stress exponent that varies with SBTnpa = atmospheric pressure in same units as qt and σv.
Robertson and Wride (1998) and Zhang et al.(2002) use the term, qc1N instead of Qtn. This paperwill use the more general term, Qtn. Where the term‘Qt’ denotes normalized corrected cone resistanceand the subscript ‘n’ denotes normalization with avariable stress exponent. Note that, when n = 1,Qtn = Qt1. Zhang et al. (2002) suggested that thestress exponent, n, could be estimated using the SBTnIndex, Ic, and that Ic should be defined using Qtn.
Robertson (2008) recently updated the stress nor-malization by Zhang et al. (2002) to allow for a vari-ation of the stress exponent with both SBTn Ic andeffective overburden stress using:
n = 0.381(Ic)+ 0.05(σ ′vo/pa)− 0.15 (6)
where n ≤ 1.0.Robertson (2008) suggested that the above modifi-
cation to the stress exponent would capture the correctstate response for soils at high stress level and wouldavoid the need for a further stress level correction (Kσ )in liquefaction analyses.
There have been several publications regarding theappropriate stress normalization (Olsen and Malone,1988; Robertson, 1990; Jefferies and Davies, 1991;Robertson and Wride, 1998; Zhang et al., 2002;Boulanger and Idriss, 2004a; Moss et al., 2006; Cetinand Isik, 2007; Robertson, 2008). The contours ofstress exponent suggested by Cetin and Isik (2007)are very similar to those first suggested by Robertsonand Wride (1998), updated by Zhang et al. (2002) andfurther modified slightly by Robertson (2008). Thecontours by Moss et al. (2006) are similar to thosefirst suggested by Olsen and Malone (1988). The nor-malization suggested by Boulanger and Idriss (2004a)
only applies to sands where the stress exponent varieswith relative density with a value of around 0.8 inloose sands and 0.3 in dense sands. Figure 2 shows acomparison of the stress exponent contours suggestedby Robertson (2008) for σ ′
vo/pa = 1.0, Moss et al.(2006), and Boulanger and Idriss (2004a) on the nor-malized SBTn chart of Qtn – Fr. The regions where thethree methods provide similar values are highlightedand show that the methods agree on or close to thenormally consolidated zone suggested by Robertson(1990). Wroth (1984) showed that the stress exponentis 1.0 for clays based on Critical State Soil Mechanics(CSSM) theory, which is reflected in the Robertson(1990 & 2008) contours. The contours suggested byOlsen and Malone (1988) and Moss et al. (2006) arenot supported by CSSM.
Robertson (1990) stated that the soil behaviourtype charts are global in nature and should be usedas a guide for defining Soil Behaviour Type (SBT).Caution should be used when comparing CPT-basedSBT to samples with traditional classification systemsbased only on grain size distribution and plastic-ity. Factors such as changes in stress history, in-situstresses, macro fabric, cementation, sensitivity andvoid ratio/water content will also influence the CPTresponse and resulting SBT. The rate and manner inwhich the excess pore pressures dissipate during apause in the cone penetration can significantly aid inidentifying soil type.
Robertson (1990) and others have suggested thatsoils that have a SBTn index Ic < 2.5 are generallycohesionless where the cone penetration is generallydrained and soils that have Ic > 2.7 are generally cohe-sive where the cone penetration is generally undrained.Cone penetration in soils with 2.5 < Ic < 2.7 is oftenpartially drained.
Figure 2. Comparison of contours of stress exponent ‘n’ onnormalized SBTn chart Qtn – Fr.
7
5 SOIL STRATIGRAPHY—TRANSITIONZONES
Robertson and Campanella (1983) discussed how thecone tip resistance is influenced by the soil ahead andbehind the cone tip. Ahmadi and Robertson (2005)illustrated this using numerical analyses and con-firmed that the cone can sense a soil interface up to 15cone diameters ahead and behind, depending on thestrength/stiffness of the soil and the in-situ effectivestresses. In strong/stiff soils, the zone of influence islarge (up to 15 cone diameters), whereas, in soft soils,the zone of influence is rather small (as small as 1 conediameter). Ahmadi and Robertson (2005) showed thatthe zone of influence decreased with increasing stress(e.g. dense sands behave more like loose sands at highvalues of effective stress).
For interbedded soil deposits, the thinnest stiff soillayer for which the measured cone resistance repre-sents a full response is about 10 to 30 cone diameters.Hence, as described by Robertson and Campanella(1983), soil parameters may be under-estimated in thinstiff layers embedded within a softer soil (e.g. thin sandlayers in a softer clay). Fortunately, the cone can sensea thin soft soil layer more precisely than a thin stiffsoil layer. The fact that the cone can underestimate thesoil resistance in thin stiff layers has led to the thinlayer correction for liquefaction analyses (Robertsonand Wride, 1998, Youd et al., 2001).
The zone of influence ahead and behind a cone dur-ing penetration will influence the cone resistance at
any interface (boundary) between two soil types ofsignificantly different strength and stiffness. Hence,it is often important to identify transitions betweendifferent soils types to avoid possible misinterpreta-tion. This issue has become increasingly importantwith software (or spreadsheets) that provide interpre-tation of every data point from the CPT. When CPTdata are collected at close intervals (typically every20 to 50 mm) several data points are ‘in transition’when the cone passes an interface between two differ-ent soil types (e.g. from sand to clay and vice-versa).For thin stiff layers the two interface regions can joinsuch that the cone resistance may not represent the truevalue of the thin layer.
It is possible to identify the transition from one soiltype to another using the rate of change of either Icor Qtn. When the CPT is in transition from sand to clay,the SBTn Ic will move from low values in the sand tohigher values in the clay. Robertson and Wride (1998)suggested that the approximate boundary betweensand-like and clay-like behaviour is around Ic = 2.60.Hence, when the rate of change of Ic is rapid andis crossing the boundary defined by Ic = 2.60, thecone is likely in transition from a sand-like to clay-like soil, or vice-versa. Profiles of Ic provide a simplemeans to identify these transition zones. Figure 3 illus-trates a CPT profile through a deposit of interbeddedsands and clays and shows how computer software(CLiq, 2008) can identify transition zones on the Icprofile based on the rate of change of Ic as Ic crossesthe value 2.60. There are clear transitions from clay to
Figure 3. Example of interbedded soil profile with transition zones identified (in red) on SBTn Ic plot (CLiq Software,Geologismiki).
8
sand (and vice-versa) at depths of 4.5, 8.5, 12.5, 14.1,14.5, 16.9, 17.5, and 20.5 m. The region between 5.0to 8.0 m, and again between 20.5 to 21.8 m, representsoils close to the boundary of Ic = 2.60. Althoughthese transitions could be identified from combina-tions of Qtn, Fr and Bq, the algorithm (software) thatidentifies the zones on the profile of Ic appears to bemore effective. Figure 3 also illustrates that the porepressure measurements are less effective at shallowdepths where saturation of the CPT sensor may be lesseffective. At depths of about 14 m, 17 m and 21 mthere are thin sand layers where the maximum valuesin the sand are likely too low due to the adjacent tran-sition zones. Hence, identification of transition zonesaids in the recognition of thin layers that may requirecorrection (Youd et al., 2001).
6 RESISTANCE TO EARTHQUAKE LOADING
Idriss and Boulanger (2008) present a summary of thehistory and background on the evaluation of liquefac-tion resistance to earthquake loading. They describe indetail how the Factor of Safety (FSliq) against trigger-ing of liquefaction in sand-like soils can be computedas the ratio of the soils CRR to the earthquake-inducedCSR, with both the CRR and CSR values pertainingto the design earthquake magnitude (M) and in-situeffective overburden stress (σ ′
vo):
FSliq = CRRM,σ ′vo/CSRM,σ ′
vo(7)
Alternately, it is common to convert the earthquake-induced CSR into the reference condition applicableto M = 7.5 and σ ′
vo = 1 atm. (i.e. σ ′vo/pa = 1).
FSliq = CRRM=7.5, σ ′vo=1/CSRM=7.5, σ ′
vo=1 (8)
where:CRRM=7.5, σ ′
vo=1 = Cyclic Resistance Ratio applica-ble to M = 7.5 and an effective overburden stress ofσ ′
vo = 1 atm., sometimes presented as simply CRR7.5.CRRM=7.5, σ ′
vo=1 = earthquake induced Cyclic StressRatio adjusted to the equivalent CSR for the refer-ence values of M = 7.5 and an effective overburdenstress of σ ′
vo = 1 atm., sometimes presented as simplyCSR7.5.
For low-risk projects, CSR is typically estimatedusing the Simplified Procedure first described by Seedand Idriss (1971), using:
Alternate methods have been suggested for estimat-ing the correction factors, rd, MSF and Kσ .
Boundary lines have been developed that separatecase histories in which ‘liquefaction’ was observed,from case histories in which liquefaction was notobserved. This boundary line is used to provide therelationship between in-situ CRR7.5 and an in-situ test
index. Due to space limitations, this paper will onlypresent CPT-based methods to estimate CRR7.5.
6.1 Sand-like (cohesionless) soils
CRR7.5 for sand-like soils is generally defined in termsof ‘triggering’ liquefaction (i.e. reaching zero effec-tive stress) although laboratory testing often usesa critical shear strain level (e.g. γ = 3%). Trig-gering of ‘liquefaction’ in loose sands is the onsetof large strains. Therefore, since CRR7.5 is tradi-tionally used to define ‘liquefaction’ it can also beused to define the onset of large deformations. If thefactor of safety against ‘liquefaction’ is less than 1(i.e. FSliq < 1) shear strains can be large and tend toincrease as the factor of safety decreases, especiallyfor loose sands.
The evaluation of CRR has evolved primarily fromcase histories of past earthquakes. The earliest effortsbegan with attempts to use SPT data (Kishida, 1966,Seed et al, 1984). In the early 1980’s efforts were madeto use CPT data (Zhou, 1980; Robertson and Cam-panella, 1985). In 1996–97, a workshop by NCEERand NSF provided a summary and recommendationson SPT-, CPT-, and Vs-based correlations and proce-dures (Youd et al., 2001). Following the NCEER work-shop several major earthquakes provided new casehistories. Moss et al. (2006) produced a compilationof the expanded database.
The NCEER/NSF workshop provided a set of rec-ommendations by over 20 leading experts and wassummarized by Youd et al. (2001). Youd et al. (2001)recommended the Robertson and Wride (1998) methodfor the CPT-based approach to evaluate CRR for cohe-sionless soils (Ic < 2.60). However, since 1997 therehave been several publications attempting to updatethese recommendations. These updates have led tosome confusion in practice, since changes were sug-gested to both CSR and CRR, which often resulted inminor changes to the calculated FSliq.
Traditionally, case history data have been compiledby identifying the combination of the earthquake-induced cyclic stress ratio, CSR, and in-situ test resultsthat best represents the ‘critical zone’ where liquefactionwas estimated to have occurred for each site. It hasbeen common to adopt a magnitude M = 7.5 earth-quake, an effective overburden stress of σ ′
vo = 1 atmand case histories with modest static shear stress(i.e. essentially level ground conditions). The resultingCSR7.5 values are plotted against the in-situ test resultsnormalized to σ ′
vo = 1 atm. The resulting plots arethen used to develop boundary lines separating casesof ‘liquefaction’ from cases of ‘non-liquefaction’ and,therefore, a method to estimate the CRR7.5. This paperwill focus only on the approaches that use CPT results,since the CPT is generally considered more repeatableand reliable than the SPT and provides continuous datain a cost effective manner.
9
Although this traditional approach of using casehistory data has resulted in significant developments,the approach has some limitations. The following is ashort description of the main limitations.
‘Liquefaction’ and ‘Non-liquefaction’: field evi-dence of ‘liquefaction’ generally consists of surfaceobservations of sand boils, ground fissures or lat-eral spreading. Sites that show no surface featuresmay have experienced either liquefaction or the devel-opment of significant pore pressures in some soillayers, but no sand boils resulted, either due to thedepth of the layer or the overlying deposits. Also, sitesthat show no surface deformation features may haveexperienced significant pore pressure development insome soil layers, but showed limited post-earthquakedeformations due to ground geometry and lack of anysignificant static loads. Few case histories have welldocumented deformation records where deformationswere recorded with depth.
Selecting the ‘critical zone’: the depth where‘liquefaction’ was assumed to have occurred requiresconsiderable judgment. Occasionally, this is based onlinking sand boil material to a specific soil layer, butoften the selection is more subjective.
Average data points to represent each site: consid-erable judgment is required to select an appropriateaverage value for the in-situ test. For SPT results thiswas simpler because there were often only 1 or 2 SPTvalues in the critical zone. However, for CPT resultsthis is more difficult, since there can be many CPTvalues within a layer. CPT results often show that asoil layer is not uniform either in terms of consistency(i.e. density/state) or grain characteristics (e.g. finescontent/plasticity). In critical soil layers, where the soilis non-uniform and the cone resistance is variable, an‘average’ value can be misleading.
Although the SPT- and CPT-based design methodswere developed using average values, the methods aregenerally applied to all data points for design. CPTdata are generally recorded at 5 cm depth intervalsto provide a near continuous profile. Hence, applica-tion of case-history based design methods, using thenear continuous CPT profile, incorporate some levelof conservatism. Applying the CPT-based methods toaverage in-situ test values for design requires judgmentin selecting appropriate representative average values,and details in the near continuous profile can be lost.
Although the traditional approach has limitations,it has resulted in relatively simply approaches to eval-uate a complex problem. Moss et al. (2006) (basedon Moss, 2003) compiled a comprehensive databasebased on CPT records. For this paper, the Moss (2003)database has been re-evaluated using the continuousdigital CPT records, where available, to confirm ormodify the estimated average in-situ test values. There-evaluation focused primarily on case histories thatplot close to the boundary lines, since these play amore important role in defining the boundary line.
The near continuous CPT records were processedthrough software that incorporates the updatedRobertson and Wride (1998); Zhang et al. (2002)and Zhang et al. (2004) CPT-based method as well astransition zone detection and the updated Robertson(2008) stress normalization (equation 6) (CLiq www.geologismiki.gr). The re-evaluation showed that theRobertson and Wride (1998) method performedextremely well on the database of near continuous CPTrecords. Some sites that appeared to have ‘liquefaction’average data points on the ‘non-liquefaction’ side ofthe boundary line actually predicted ‘liquefaction’(i.e. had regions in the critical layer where the com-puted FSliq < 1)when using the near continuous CPTdata. Hence, at sites where the Robertson and Wride(1998) method would appear to have incorrectly pre-dicted performance based on the case history resultsusing Moss et al. (2006) average values, the methodpredicted the correct performance using the mea-sured near continuous values in terms of liquefaction(i.e. FSliq < 1.0) and post-earthquake deformations.Some key sites, where the average values selectedby Moss et al. (2006) were considered inappropriate,are the sites at Whiskey Springs (1983 Borah Peakearthquake). These sites were composed of gravellysands to sandy gravels and the CPT results showedsignificant rapid variation caused by the gravel con-tent. The CPT measurements at these sites were lessreliable due to the gravel content, and the average val-ues selected by Moss et al. (2006) were consideredtoo high and unrepresentative of the loose sand matrixthat likely dominated the buildup of pore pressuresduring the earthquake. Other key sites are BalboaBlvd. and Malden St. (1994 Northridge, USA) andKornbloom (1982 Westmorland, USA). Average val-ues can be misleading in interbedded soils and maynot adequately represent the various individual soillayers.
Moss et al. (2006) and Juang et al. (2003) haveused the expanded case history database based onaverage values to provide criteria based on probabil-ity. The re-evaluation, using near continuous CPTrecords, suggest some uncertainty on proposed levelsof probability, due to the highly subjective nature of theaverage values selected and the observation that some‘liquefaction’ and ‘non-liquefaction’ sites were incor-rectly classified when using only the Moss et al. (2006)average values. It is recommended that the near contin-uous CPT data be used to evaluate various CPT-basedliquefaction methods and not average values that weresubjectively selected. It is also interesting to note that,to the authors knowledge, none of the more recentCPT-based methods (i.e. post-Youd et al., 2001) usedthe recorded near continuous CPT records from thecase histories to confirm the accuracy of the proposednew methods.
The Moss et al. (2006) database included 182 casehistory results (146 ‘liq’ and 36 ‘non-liq’). However,
10
30 cases (23 ‘liq’ and 7 ‘non-liq’) were describedas ‘Class C’ data that were case histories where theCPT results were obtained using either ‘non-standardor mechanical cone’ or ‘no friction sleeve data avail-able’. The Class C data are clearly less reliable thanthe rest of the data, especially for methods that makeuse of the friction sleeve results in the form of eitherfriction ratio, Rf (Moss et al., 2006) or soil behav-ior type, Ic (Robertson and Wride, 1998; Juang et al.,2003). Robertson and Campanella (1983) showed thatmechanical cone friction sleeve values can be signif-icantly different from standard electric cone values inthe same soil.
The database, (with Class C data removed) whereliquefaction was observed, had earthquake magnitudesin the range 5.9 < Mw < 7.7 and vertical effectivestress in the range 15 kpa < σ ′
v < 135 kpa. Theaverage vertical effective stress in the liquefied layerswas 60 kpa. No liquefaction, based on surface obser-vations, was considered to have occurred at a depthgreater than 16 m. The average depth for the criticalliquefiable layers was around 5 to 6 m.
All the CPT-based methods (to determine CSR7.5)typically include corrections for depth (rd), magni-tude scaling factors (MSF) and overburden correctionfactor (Kσ ). The variations in these correction fac-tors when applied to the database are generally small.Hence, the database is insufficient to clarify whichcorrection methods are appropriate for design. Mostmethods specify that consistency is required whenapplying the methods to design problems (i.e. usethe same correction factors on which the methodwas based). This paper uses the correction factors(rd, MSF, Kσ ) suggested by the NCEER workshop(Youd et al., 2001), with Kσ = 1.0.
Figure 4 shows a summary plot of the reevaluatedexpanded database in terms of CPT results in the formof CSR7.5 versus normalized cone resistance (Qtn).The Class C data are not included in Figure 4. Figure 4includes some case history data where the soil wasnot considered to be ‘clean sand’, however, the result-ing boundary line is unaffected, because the ‘liq’ datain soils that are not ‘clean sands’ have lower coneresistance (i.e. located to the left of the boundary line).The resulting boundary line is often referred to as the‘clean sand’ boundary line.
Figure 4 also shows some of the most recent pub-lished correlations superimposed over the updateddatabase. The comparison in Figure 4 is not strictlycorrect, since the various published procedures includedifferent normalization procedures for the CPT results.Fortunately, the differences, when applied to the casehistory data, are generally small (less than 20%),since all of the case history data are from sites wherethe range in vertical effective stress was small (15 kpa<σ ′
v < 135 kpa). The various correlations are similarin the region of maximum data (20 < Qtn < 100).When Qtn is larger than 100 the correlations differ,
Figure 4. Updated case history database in terms ofCSRM=7.5, σ ′vo=1 vs Qtn (Class C data excluded).
mainly due to the form of the suggested correlations.Hence, for ‘clean sands’ the baseline correlation toestimate CRR7.5 from CPT results is reasonably wellestablished, especially in the region defined by 20 <Qtn < 100. It is likely that there will be little gainedfrom further evaluation of current case history datausing average values for clean sands in the form ofCSR7.5 – Qtn plots. It is also recommended that furtherfine-tuning of the CRR7.5 relationships using averagevalues will be ineffective, since the location of theboundary is sensitive to the judgment used to selectappropriate average in-situ test values. The form ofthe relationship controls CRR7.5 for Qtn > 100, sincevery little field data exists in this range. The form of therelationship becomes important when the method isextended to estimate post-earthquake displacements.
For soils that are not ‘clean sands’, the traditionalapproach has been to adjust the in-situ penetrationresults to an ‘equivalent clean sand’ value. This evolvedfrom the SPT-based approach where samples couldbe obtained and the easiest parameter to quantifychanges in grain characteristics was the percent finescontent.
Research has clearly shown that fines content alonedoes not adequately capture the change in soil behav-ior. Also, the average fines content of an SPT samplemay not always reflect the variation in graincharacteristics in heterogeneous soils, since it is com-mon to place the full SPT sample into a container forsubsequent grain size analyses, with resulting mis-leading ‘average’ fines content. The recent Idriss andBoulanger (2008) CPT-based approach thatuses only fines content from samples to make adjust-ments to cone resistance is a retrograde step and is notrecommended.
Several recent CPT-based liquefaction methods usemodified CPT results to estimate clean sand equivalentvalues based on either SBT Ic (e.g. Robertson andWride, 1998; Juang et al., 2006) or friction ratio, Rf ,(Moss et al., 2006). Figure 5 shows a summary plot of
11
Figure 5. Updated case history database in terms ofCSRM=7.5, σ ′vo=1 vs Qtn,cs (Class C data excluded).
the reevaluated expanded database, in terms of CPTresults in the form of CSR7.5 versus normalized cleansand equivalent cone resistance (Qtn,cs), based on thecorrections suggested by Robertson and Wride (1998)using Ic.
Good agreement exists between the expandeddatabase and the original Robertson and Wride (1998)CPT-based method.
Figures 6 and 7 show the updated database plot-ted on the normalized SBTn chart (Qtn – Fr), whereQtn and Fr were calculated using the method sug-gested by Zhang et al. (2002) and recently modifiedslightly by Robertson (2008). Figure 6 shows thecase history data where 0.20 < CSR7.5 < 0.50.Figure 7 shows the data where CSR7.5 < 0.20. Thecase history database is insufficient to subdivide thedata into smaller divisions in the Qtn – Fr format,since both are on log scales. Presenting the case his-tory data, in terms of the full CPT data (Qtn and Fr)on the SBT chart, provides a different view of theinfluence of changing soil type on the correlations.Superimposed on the SBTn chart are the contours forCRR7.5 suggested by Robertson and Wride (1998)in the region where Ic < 2.60. The Class C dataare also included in Figures 6 and 7 but are identi-fied using a different symbol. The Moss et al. (2006)corrections using friction ratio (Rf ), appear to be influ-enced by the questionable Class C data. It is alsointeresting to note that, excluding the questionableClass C data, there are no case histories of observed‘liquefaction’ based on average CPT values whereIc > 2.60. It is useful to remember that each datapoint, in terms of Qtn and Fr, represents an averagevalue for the critical layer.
Figure 8 shows the data where CSR7.5 < 0.20with the correlations suggested by Olsen and Koester(1995); Suzuki et al. (1995); Robertson and Wride(1998) and Moss et al. (2006), for comparison. Thisformat provides a way to compare the different ‘cor-rection’ factors to adjust CPT results for soil type. Thecorrelations suggested by Moss et al. (2006) appear
Figure 6. Updated database on SBTn Qtn – Fr chart for0.20 < CRR7.5 < 0.50 and Robertson and Wride (1998)contour for CRR7.5 = 0.50 (Ic < 2.60).
Figure 7. Updated database on SBTn Qtn – Fr chart forCRR7.5 < 0.20 and Robertson and Wride (1998) contourfor CRR7.5 = 0.20 (Ic < 2.60).
to be too conservative at high values of either fric-tion ratio or Ic. This was partly a result of using theunreliable Class C data, as well as inappropriate aver-age values for some key sites, especially the sitesfrom Whiskey Springs. The correlations suggested
12
Figure 8. Comparison of published correlations on SBTnQtn – Fr chart for CRR7.5 < 0.20.
by Suzuki et al. (1995) and Olsen and Koester (1995)appear to be unconservative at high values of Ic, whichwas also pointed out by Robertson and Wride (1998).
6.2 Clay-like (cohesive) soils
Since cohesive clay-like soils are not susceptible to‘liquefaction’, the criteria used to define CRR is defor-mation, which is often assumed to be a shear strainof γ = 3%. Since detailed deformation records areuncommon in many case histories, much of our under-standing regarding the response of cohesive soils toearthquake loading derives from undrained cyclic lab-oratory testing. Fortunately, it is also possible to obtainhigh quality undisturbed samples in many clay-likesoils.
Sangrey et al. (1978) showed that fine-grained soilstend to reach a critical level of repeated loading thatis about 80% of the undrained shear strength (su).Boulanger and Idriss (2006, 2007) provided a sum-mary of the response of cohesive soils to cyclic load-ing. There is a strong link between the cyclic undrainedresponse of fine-grained soils and their monotonicundrained response. The monotonic response of fine-grained soils is generally defined in terms of their peakundrained shear strength, su. Although the undrainedshear strength is not a unique soil parameter, sinceit varies with the direction of loading, it does pro-vide a simple way to understand the behavior ofcohesive soils and captures many features (e.g. stress
history, age and cementation). During earthquakeloading, the predominant direction of loading is simpleshear; hence, the undrained strength in simple shearis often the most appropriate parameter to link withCRR. Since earthquake loading is best defined interms of CSR (τcy/σ
′v), it is appropriate to compare
this with the undrained strength ratio (su/σ′v). In sim-
ple terms, if the earthquake imposes a shear stressratio that is close to the undrained strength ratio of thesoil, the soil will deform. Since earthquake loadingis rapid and cyclic, the resulting deformations maynot constitute ‘failure’ (i.e. unlimited deformations).However, shear deformations can be large and tend toprogress during the earthquake. Boulanger and Idriss(2004) used the term ‘cyclic softening’ to describe theprogression of shear strains during cyclic undrainedloading in fine-grained soils.
Boulanger and Idriss (2004b) presented publisheddata that showed that, when the CSR ratio approachesabout 80% of su/σ
′v, deformations tend to become
large. Wijewickreme and Sanin (2007) showed thatthe CRR(γ = 3%) in low plastic silts is also controlledby their peak undrained shear strength ratio (su/σ
′v).
Although it is common to treat low plastic silts as‘sand-like’, their CRR is controlled by their undrainedstrength ratio. Hence, soft low plastic silts tend to‘behave’ similar to soft clays, where their responseis controlled by the undrained strength ratio.
Boulanger and Idriss (2007) suggested that theCRR7.5 (for a shear strain of 3%) could be estimatedusing either:
CRR7.5 = 0.8(su/σ′vo) (10)
or
CRR7.5 = 0.18(OCR)0.8 (11)
Both methods are equivalent, since Ladd (1991) showedthat:
su/σ′vo = 0.22(OCR)0.8 (12)
Boulanger and Idriss (2004b) suggested a furtherreduction factor (Kα) to CRR7.5, based on the staticshear stresses existing at the time of the earthquake.Therefore, the factor of safety against cyclic softening(3% shear strain), for cases in which the static shearstresses are small (i.e. Kα = 1.0), can be expressed as:
FSγ=3% = CRRM/CSRM = CRR7.5/CSR7.5 (13)
Boulanger and Idriss (2007) showed that the MSFfor clays is different than that for sands. They alsoshowed that the CRR7.5 of saturated clays and plasticsilts can be estimated by three approaches:
• Directly measuring CRR by cyclic laboratory test-ing on undisturbed samples.
• Empirically estimating CRR based on su profile.• Empirically estimating CRR based on consolidation
stress history (i.e. OCR) profile.
13
Boulanger and Idriss (2007) described that the firstapproach provides the highest level of insight and con-fidence, whereas the second and third approaches useempirical approximations to gain economy. For lowrisk projects, the second and third approaches are oftenadequate. Based on the work of Wijewickreme andSanin (2007) it would appear that the CRR7.5 for softlow plastic silts can also be estimated using the sameapproach.
Robertson (2008) showed that CPT results in fine-grained soils are influenced primarily by both stresshistory (OCR) and soil sensitivity (St) and that the nor-malized cone resistance (Qtn) is stronglyinfluenced by OCR and almost unaffected by St,whereas, the normalized friction ratio (Fr) is stronglyinfluenced by St and almost unaffected by OCR.Hence, Robertson (2008) suggested that the peakundrained shear strength ratio in cohesive soils canbe estimated from:
(su/σ′vo) = qt − σvo
σ ′vo
(1/Nkt) = Qtn/Nkt (14)
when Ic > 2.60 and n ∼1.0)
where Nkt = empirical cone factor with an averagevalue of 15.
Hence, when Kα = 1.0:
CRR7.5 = 0.8 Qtn/15 = 0.053 Qtn (15)
Alternately, the OCR of clay can be estimated using(Kulhawy and Mayne, 1990):
OCR = 0.33 Qtn (16)
Hence, when Kα = 1.0:
CRR7.5 = 0.074 (Qtn)0.8 (17)
For values of Qtn < 10 (i.e. CRR7.5 < 0.5), bothapproaches produce similar values of CRR7.5.
Hence, estimates of CRR7.5 can be made from CPTresults using the normalized cone resistance Qtn, sinceCRR7.5 is controlled primarily by the peak undrainedshear strength ratio. Note that in clays and silts whereIc > 2.60, Qtn = Qt1.
6.3 All soils
By combining the Robertson and Wride (1998)approach for cohesionless sand-like soils with theBoulanger and Idriss (2007) recommendations forcohesive clay-like soils, it is possible to provide a sim-ple set of recommendations to estimate CRR7.5 fromCPT results for a wide range of soils.
The recommendations can be summarized, asfollows:
When Ic ≤ 2.60, assume soils are sand-like:
Use Robertson and Wride (1998) recommenda-tion based on Qtn,cs = Kc Qtn, where Kc is a
function of Ic. Robertson and Wride (1998) seta minimum level for CRR7.5 = 0.05.
When Ic > 2.60, assume soils are clay-like where:CRR7.5 = 0.053 Qtn Kα (18)
Boulanger and Idriss (2007) suggested that, in clay-like soils, the minimum level for CRR7.5 = 0.17 Kα
for soft normally consolidated soils.For a more continuous approach, it is possible to
define a transition zone between sand-and clay-likesoils:
When Ic ≤ 2.50, assume soils are sand-like:
Use Robertson and Wride (1998) recommenda-tion based on Qtn,cs = Kc Qtn, where Kc is afunction of Ic.
When Ic > 2.70, assume soils are clay-like, where:CRR7.5 = 0.053QtnKα (19)
When 2.50 < Ic < 2.70, transition region:
Use Robertson and Wride (1998) recommenda-tions based on Qtn,cs = KcQtn, where:
Kc = 6 × 10−7(Ic)16.76 (20)
The recommendations where 2.50 < Ic < 2.70 rep-resent a transition from drained cone penetration toundrained cone penetration where the soils transitionfrom predominately cohesionless to predominatelycohesive.
Figures 9 and 10 show the proposed combinedrelationships for CRR7.5 = 0.5 and 0.2, respectively,compared to the expanded database. Additional non-liquefaction data points (28 in total) have been addedfrom the published case history records. The ‘non-liquefaction’ points reflect soil layers (predominatelyclay-like soils) that did not ‘liquefy’ and did not showany observable/recorded deformations (i.e. no cyclicfailure). As noted above, the criteria to define CRR7.5in clay is a shear strain of 3%. Figure 9 includes twodata points (Yalova Harbour and Soccer Field sites,Kocaeli earthquake, Turkey, 1999) where cyclic soft-ening may have occurred in the soft clay layer duringearthquake shaking but no significant post-earthquakedeformations within the clay layers were observed ornoted. The lack of observed deformation in the claylayers at the two sites in Turkey may have been dueto small static shear stresses at the depth of the clay.Figure 10 includes one data point from the Moss Land-ing site (Sandholt Rd., Loma Prieta, 1995) where a softsilty clay (Qtn = 4 to 5, Fr = 3 to 4%) appears to havebeen close to cyclic failure and where a small amountof post-earthquake lateral deformation (approximatelyγ = 0.5%) was observed from slope indicator mea-surements (Boulanger et al., 1995) and where theCSR7.5 was about 0.25.
Data from three sites (Marina District, TreasureIsland Alameda) with deposits of soft, sensitive San
14
Figure 9. Proposed relationship to estimate CRR7.5 = 0.50for a wide range of soils compared to updated database.
Figure 10. Proposed relationship to estimate CRR7.5 =0.20 for a wide range of soils compared to updated database.
Fransico (SF) young Bay Mud are also identified inFigure 10. These sites likely experienced a CSR7.5of about 0.15 during the Loma Prieta earthquake butshowed no reported signs of deformations within theclay layer. This may have been, in part, due to therather small static shear stress at these sites withinthe soft clay. The less reliable Class C data have notbeen included in Figures 9 and 10.
Boulanger and Idriss (2004) showed that high staticshear stresses in soft clays can initiate cyclic failureduring earthquake loading. They presented resultsfrom sites that experienced ground failure duringthe Kocaeli 1999 earthquake in soft clays where thestatic shear stresses were high. The above CPT-basedapproach to estimate CRR also correctly predictsground failure at the sites presented by Boulangerand Idriss (2004) when Kα < 1.0.
Typically, when Ic > 2.60 the soils are generallyfine-grained and more easily sampled. Therefore, inthis region (Ic > 2.60), selective sampling and labora-tory testing can be appropriate, depending on the riskof the project.
7 POST-EARTHQUAKE DEFORMATIONS
Estimating deformations in soils is generally difficult,due to the non-linear, stress dependent stress-strainresponse of soils. Estimating deformation after earth-quake loading is more difficult, due in part to thecomplex nature of earthquake loading and the role ofsoil stratigraphy and variability.
Idriss and Boulanger (2008) present a summaryof alternate approaches to estimating post-earthquakedeformations depending on the risk and scope of theproject. For low to moderate risk projects it is commonto estimate post-earthquake deformations by estimat-ing strains and then integrate those strains over depthto estimate deformation. The estimated deformationsmay also be empirically adjusted on the basis of cal-ibration to case history observations. For high riskprojects it is appropriate to perform complex non-linear dynamic numerical analyses if initial screeningindicates a need.
7.1 Vertical settlements due to reconsolidation
Post-earthquake vertical displacements can developin two ways: (1) settlement caused by reconsolida-tion, and (2) vertical displacement caused by sheardeformation associated with lateral deformation. Thissection addresses only settlements caused by recon-solidation.
Post-earthquake reconsolidation volumetric strainsare generally estimated using relationships derivedprimarily from laboratory studies. Methods are thenevaluated using case history observations. One of theprimary laboratory studies used is that by Ishihara andYoshimine (1992) for cohesionless soils. Ishihara andYoshimine (1992) observed that volumetric strains ofsand samples were directly related to the maximumshear strain during undrained cyclic loading and to
15
the initial relative density of the sand. Ishihara andYoshimine (1992) showed that when FSliq > 1 someshear and volumetric strains still occur and that asthe FSliq decreases (FSliq < 1), shear and volumetricstrains increase but reach maximum values depend-ing on the relative density. When FSliq < 1.0, loosecohesionless soils have reached zero effective stresswith a loss of structure/fabric, the stiffness of the soil isthen very small during reconsolidation that can resultin large volumetric strains.
Zhang et al. (2002) coupled the Robertson andWride (1998) CPT-based method using clean sandequivalent values to determine FSliq with the Isihaharaand Yoshimine (1992) volumetric strain relationships,to provide a method to estimate the post-earthquakevertical reconsolidation settlements. Zhang et al. (2002)evaluated the approach using case history observationsand showed that the approach provided reasonablepredictions of settlements, although details on sitegeometry and soil stratigraphy play an important role.Since most cohesionless soils have relatively highpermeability, the post-earthquake reconsolidation set-tlements occur relatively soon after the earthquake,but depend on soil stratigraphy and drainage.
7.1.2 Volumetric strains—cohesive clay-like soilsFactors affecting vertical (1-D) settlement caused bypost-earthquake reconsolidation of clay layers are dis-cussed in Ohara and Matsuda (1988), Matsuda andOhara (1991) and Fiegal et al. (1998). The limitedlaboratory data indicate that reconsolidation volu-metric strains are controlled primarily by the max.shear strain which is function of the factor of safety(FSγ=3%) and stress history (OCR) of the soil. Dur-ing undrained cyclic loading, pore pressures developthat result in a decrease in effective confining stress.However, the effective stresses generally do not reachzero and the soil retains some structure and stiff-ness. Wijewickreme and Sanin (2007) showed that,on average, for a wide range of fine-grained soils,when FSliq = 1 the excess pore pressure representsabout 80% of the effective confining stress(i.e. �u/σ ′
vo = ru = 0.8). Volumetric strains occuras the soil reconsolidates back to the in-situ effectiveconfining stress. The volumetric strains in cohesivesoils during reconsolidation after earthquake loadingare generally much smaller than those observed incohesionless coarse-grained soils because cohesivesoils retain some level of stiffness during reconsol-idation. Case history field observations have alsoshown that post-earthquake settlements, due to recon-solidation, are generally small at sites with thickdeposits of cohesive soils. For example, the SanFransico Bay area in California has extensive thickdeposits of soft (young) Bay Mud (essentially nor-mally to lightly overconsolidated clay) but very fewobservations of measurable post-earthquake settle-ments within the clay deposits were made following
the Loma Prieta earthquake. The re-evaluation ofpost-earthquake reconsolidation settlements at theMarina District, Treasure Island and Moss Landingsites following the Loma Prieta earthquake and sitesin Taiwan following the Chi-Chi earthquake, sug-gest an average volumetric strain of less than 1% infine-grained soils.
Volumetric strains for cohesive soils can be estimatedusing the 1-D constrained modulus, M, and the changein effective stress due to the earthquake loading where,
εvol = (�σ ′v/M) (21)
�σ ′v = ruσ
′vo (22)
The buildup in pore pressure and hence, change ineffective stress, is a function of the factor of safety(FS) and the OCR of the soil. Laboratory test resultsindicate that ru is a function of FS. When FS = 1.0,ru = 0.8 and when FS = 2, ru = 0. Assuming alinear relationship between FS and ru and an inverserelationship with OCR gives:
ru = [0.8 − 2.66 log (FS)]/OCR (23)
where: ru </ = 1.0, when FS = 0.84Kulhawy and Mayne (1990) showed that OCR can
be estimated from the CPT using:
OCR = 0.33 Qtn (24)
Hence,
�σ ′v = [0.8 − 2.66 log (FS)] σ ′
vo/0.33 Qtn (25)
Assuming the 1-D constrained modulus duringreconsolidation is generally larger than the initialconstrained modulus estimated from the CPT:
M = A MCPT (26)
The 1-D constrained modulus estimated from theCPT is equivalent to the modulus from the in-situstress to a higher stress, whereas during reconsolida-tion the cohesive soil has become overconsolidateddue to the decrease in effective stress and the recon-solidation modulus is stiffer. For soft normally con-solidated cohesive soils the reconsolidation stiffnessis about 10 MCPT. Whereas, in stiff overconsolidatedcohesive soils, the reconsolidation stiffness is approxi-mately equal to MCPT. Therefore, assume that A varieswith OCR as follows:
A = 10 − 9 log (OCR) (27)
Since OCR = 0.33 Qtn
A + 10 − 9 log (0.33 Qtn) (28)
Robertson (2008) showed that in soft clays:
MCPT = (Qtn)2σ ′
vo (29)
Hence:
εvol = [0.8 − 2.66 log (FS)]/[0.33A(Qtn)3] (30)
When FS ≤ 0.84 set ru = 1.0 & limit εvol ≤ 1%.
16
The above procedure provides an approximateestimate of the post-earthquake reconsolidationvolumetric strains in clay-like soils based on CPTresults. The re-evaluation of the expanded case historydatabase shows good agreement between observedpost-earthquake settlements and those calculated usingthe Zhang et al. (2002) CPT-based method with thecontinuous CPT records incorporating the abovemethod to estimate volumetric strains in clay-like soils.
7.2 Lateral displacements due to shear deformation
7.2.1 Shear strains—cohesionless soilsZhang et al. (2004) coupled the Robertson and Wride(1998) CPT-based method to determine FSliq withthe Isihahara and Yoshimine (1992) maximum shearstrain relationships to provide a method to estimate thepost-earthquake lateral displacement index (LDI).Zhang et al. (2004) used case history observationsto modify the LDI based on ground geometry to esti-mate actual lateral displacements. Zhang et al. (2004)evaluated the approach using case history observationsand showed that the approach provided reasonable pre-dictions of settlements. Chu et al. (2007) showed thatthe Zhang et al. (2004) CPT-based method providedreasonable but generally conservative estimates of lat-eral displacements from the 1999 Chi-Chi (Taiwan)earthquake. Chu et al. (2007) also showed that shearstrains at a depth more than twice the height of the freeface should not be included in the method, since staticshear stresses are likely too small to contribute to thelateral deformation.
7.2.2 Shear strains—cohesive soilsThe potential for shear deformations or instability inclay-like cohesive soils depends heavily on the staticshear stresses (which can be captured via Kα) and thesensitivity of the soil.
Boulanger and Idriss (2004) have shown that highstatic shear stresses in soft clays can initiate highshear strains during earthquake loading. The CPT-based approach described here captures the decreasein FS in clay-like soils when an appropriate value ofKα is used.
If clays are sensitive and show significant strainsoftening in undrained shear (i.e. high sensitivity, St),strength loss can lead to significant deformations andinstability. Boulanger and Idriss (2007) stated thatthe magnitude of strain, or ground deformation, thatwill reduce the clay’s undrained shear strength (su)to its fully remolded value (sur) is currently diffi-cult to assess, but it is generally recognized that itwould require less deformation to remold very sensi-tive clays than more ductile relatively insensitive clays.Based on the assumption that the CPT sleeve friction(fs) measures the remolded shear strength of the soil
(i.e. sur = fs), it is possible to estimate the sensitivityof clays using CPT results (Robertson, 2008); where:
St = su/su(r) = 7.1/Fr (31)
It is also possible to estimate the remolded undra-ined shear strength ratio (sur/σ
′vo) using (Robertson,
2008):
sur/σ′vo = fs/σ
′vo = (Fr · Qtn)/100 (32)
As soil sensitivity increases, CPT data moves to theleft on the Qtn – Fr SBTn chart, as Fr decreases withincreasing St.
In a general sense, the FS(γ = 3%) is controlled bythe OCR and peak undrained shear strength of theclay (i.e. Qtn, equation 18) whereas the potential forstrength loss and large deformations is controlled bythe sensitivity of the clay (i.e. Fr, equation 31).
8 EVALUATION OF POST-EARTHQUAKEDEFORMATIONS USING CASE HISTORYOBSERVATIONS
Zhang et al. (2002; 2004) showed that CPT resultscould be used to provide reasonable estimates of post-earthquake reconsolidation settlements and lateralspread deformations. However, at that time there werelimited case history records that had CPT profiles. Theearthquakes in Turkey and Taiwan in 1999 have nowadded to the case history records with CPT profilesand recorded deformations. The following is a briefsummary of a comparison between shear deformationsobserved at sites in Taiwan and Turkey and those pre-dicted using the Zhang et al. (2004) CPT-based methodbut with the updates described in this paper. Four sitesexperienced lateral spreading during the Kocaeli earth-quake, Turkey in 1999, namely: Police Station, SoccerField, Yalova Harbour and Degirmendere Nose sites.Several sites also experienced lateral spreading duringthe Chi-Chi earthquake in Taiwan in 1999. As notedearlier the sites at Yalova Harbour and Soccer Fieldhave deposits of soft clay that would be predicted tohave been close to cyclic failure, but appear to have hadlittle influence on the lateral spread deformations dueto the low static shear stress at the depth of the soft clay.Hence, these sites do not assist in our estimate of prob-able post-earthquake shear strains in clays. Figure 11shows a summary of the predicted post-earthquakelateral displacements compared to the measured lat-eral displacements at the sites in Turkey and Taiwanbased on the Zhang et al. (2004) CPT-based methodwith the updates described in this paper. The updatedCPT-based method to estimate liquefaction and cyclicsoftening appears to provide reasonable estimates oflateral deformations.
The updated CPT-based method, including the addi-tion for estimating cyclic softening in clay-like soils,was used to re-evaluate the available case history
17
Figure 11. Measured post earthquake lateral displacementscompared to predicted values using Zhang et al. (2004)CPT-based method.
CPT records and showed that clay-like soils gener-ally play a minor role in almost all the available casehistory records. Although some clay-like soils likelyexperienced some cyclic softening during the earth-quake, they generally appear to contribute little tothe observed deformations, except the few cases wherehigh static shear stresses contributed to ground failure(Boulanger and Idriss, 2004). In a general sense, cyclicsoftening and ground failure during seismic loadingfor clay-like soils is confined to soft, normally tolightly overconsolidated and/or sensitive fine-grainedsoils.
9 SUMMARY
This paper has presented an update of the Robertsonand Wride (1998) CPT-based method to evaluateboth liquefaction and cyclic softening in cohesionlessand cohesive soils. Case history records have beencarefully reviewed to re-evaluate the CPT-based method.Where possible, the near continuous CPT records havebeen used in the re-evaluation. The original Robertsonand Wride (1998) method has been updated using anew stress normalization procedure that captures thechange in soil response with increasing overburdenstress and avoids the need for the Kσ correction forhigh overburden stresses. A transition zone detectionfeature has also been included to identify zones wherethe near continuous CPT data may incorrectly inter-pret soil type, due to rapid variation at soil boundaries.The method has also been extended to include cohe-sive clay-like soils using the concepts described byBoulanger and Idriss (2004). The extension into theclay-like region avoids the need for a SBTn Ic cut offto separate sand-like from clay-like soils.
Figure 12 presents a summary of the CPT SBTnQtn – Fr chart to identify zones of potential liquefac-tion and/or cyclic softening. The chart in Figure 12
Figure 12. CPT Soil Behavior Type (SBTn) chart forliquefaction and cyclic softening potential.
can be used as a guide for the choice of engineer-ing procedures to be used in evaluating potentialdeformation and strength loss in different types ofsoils during earthquakes. Zones A1 and A2 corre-spond to cohesionless or sand-like soils for which itis appropriate to use existing CPT case-history basedliquefaction correlations. Soils in Zones A1 and A2 areboth susceptible to cyclic liquefaction, while the loosersoils in zone A2 are more susceptible to substantialstrength loss. Zones B and C correspond to cohesiveor clay-like soils for which it is more appropriate touse procedures similar to, or modified from, thoseused to evaluate the undrained shear strength of clays(e.g., field vane tests, CPT, and shear strength tests onhigh-quality thin-walled tube samples). Soils in ZonesB and C are both susceptible to cyclic softening (e.g.accumulation of strains if the peak seismic stresses aresufficiently large), but the softer soils in Zone C aremore sensitive and susceptible to potential strengthloss. For moderate to high risk projects, undisturbedsampling of soils in Zones B and C is recommendedto determine soil response, since soils in these zonesare more suitable for conventional sampling and lab-oratory testing. Loose, saturated, non-plastic siltsoften fall in Zone C, however, their CRR is stronglycontrolled by undrained shear strength and the meth-ods described for clay-like soils also apply. However,the resulting shear and volumetric strains should beevaluated based on either, undisturbed sampling andlaboratory testing for moderate to high risk projects,or, assumed conservative values for low risk projects.For low risk projects, disturbed samples should beobtained for soils in Zones B and C to estimate if thesoils will respond either more sand-like or clay-like,based on Atterberg Limits and water content.
The CPT is a powerful in-situ test that can providecontinuous estimates of the potential for either lique-faction or cyclic softening and the resulting
18
post-earthquake deformations in a wide range of soils.However, the CPT-based approach is a simplifiedmethod that should be used appropriately depend-ing on the risk of the project. For low risk projects,the CPT-based method is appropriate when combinedwith selective samples to confirm soil type as well asconservative estimates of soil response. For moderaterisk projects, the CPT-based method should be com-bined with appropriate additional in-situ testing, aswell as selected undisturbed sampling and laboratorytesting, to confirm soil response, where thin-walledtube sampling is generally limited to fine-grainedsoils in Zones B and C. For high risk projects, theCPT-based method should be used as an initial screen-ing to indentify the extent and nature of potentialproblems, followed by additional in-situ testing andappropriate laboratory testing on high quality samples.Advanced numerical modeling is appropriate for highrisk projects where initial screening indicates a need.
Cohesionless soils (A1 & A2)—Evaluate potentialbehavior using CPT-based case-history liquefactioncorrelations.
A1 Cyclic liquefaction possible depending on leveland duration of cyclic loading.
A2 Cyclic liquefaction and post-earthquake strengthloss possible depending on loading and groundgeometry.
Cohesive soils (B & C)—Evaluate potential behaviorbased on in-situ or laboratory test measurements orestimates of monotonic and cyclic undrained shearstrengths.
B Cyclic softening possible depending on level andduration of cyclic loading.
C Cyclic softening and post-earthquake strengthloss possible depending on soil sensitivity, loadingand ground geometry.
ACKNOWLEDGMENTS
This research could not have been carried out with-out the support, encouragement and input from JohnGregg, Kelly Cabal and other staff at Gregg Drillingand Testing Inc. Additional input and assistance wasprovided by John Ioannides.
REFERENCES
Ahmadi, M.M., and Robertson, P.K., 2005. Thin layer effectson the CPT qc measurement. Canadian GeotechnicalJournal, 42(9): 1302–1317.
Bray, J.D., and Sancio, R.B., 2006. Assessment of lique-faction susceptibility of fine-grained soils. Journal ofGeotechnical and Geoenvironmental Engineering, ASCE,132(9): 1165–177.
Boulanger, R.W., Idriss, I.M., and Mejia, L.H., 1995.Investigation and Evaluation of Liquefaction Relatedground displacements at Moss landing during 1989 LomaPrieta Earthquake. Report No. UCD/CGM-95/02. Center forGeotechnical Modeling, Department of Civil andEnvironmental Engineering, University of California,Davis.
Boulanger, R.W., and Idriss, I.M., 2004a. State normaliza-tion of penetration resistance and the effect of overburdenstress on liquefaction resistance. Proceedings 11thInternational Conference on Soil Dynamics and Earth-quake Engineering. Berkely, 484–491.
Boulanger, R.W., and Idriss, I.M., 2004b. Evaluating thepotential for liquefaction or cyclic failure of silts andclays. Report No. UCD/CGM-04/01, Center for Geotech-nical Modeling, Department of Civil and EnvironmentalEngineering, University of California, Davis.
Boulanger, R.W., and Idriss, I.M., 2006. Liquefaction sus-ceptibility criteria for silts and clays. Journal of Geotech-nical and Geoenvironmental Engineering, ASCE, 132(11):1413–426
Boulanger, R.W., and Idris, I.M., 2007. Evaluation of cyclicsoftening in silts and clays. Journal of Geotechnicaland Geoenvironmental Engineering, ASCE, 133(6):641–52.
Cetin, K.O., and Isik, N.S., 2007. Probabilistic assessment ofstress normalization for CPT data. Journal of Geotech-nical and Geoenvironmental Engineering, ASCE, Vol.133(7): 887–897.
Chu, D.B., Stewart, J.P., Youd, T.L., and Chule, B.L., 2006.Liquefaction induced lateral spreading in near-fault regionsduring the 1999 Chi-Chi, Taiwan earthquake. Journal ofGeotechnical and Geoenvironmental Engineering, ASCE,Vol. 132(12): 1549–565.
CLiq, 2008, Geologismiki: Geotechnical liquefaction soft-ware at http://w.w.w.geologismiki.gr/
Douglas, B.J., and Olsen, R.S., 1981. Soil classification usingelectric cone penetrometer. In Proceedings of Symposiumon Cone Penetration Testing and Experience, Geotech-nical Engineering Division, ASCE. St. Louis, Missouri,October 1981, pp. 209–227.
FEMA 356. Federal Emergency Management Agency,Washington, DC, Nov., 2008.
Fiegal, G.L., Kutter, B.L., and Idriss, I.M., 1998. Earth-quakeinduced settlement of soft clay. Proceedings Cen-trifuge 98. Balkema, Rotterdam (1): 231–36.
Hight, D., and Leroueil, S., 2003. Characterization of soils forengineering purposes. Characterization and EngineeringProperties of Natural Soils, Vol.1, Swets and Sitlinger,Lisse, pp. 255–360.
Idriss, I.M. and Boulanger, R.W., 2004. Semi-empirical pro-cedures for evaluating liquefaction potential during earth-quakes. Proceedings 11th International Conference onSoil Dynamics and Earthquake Engineering. Berkeley,32–56.
Jefferies, M.G., and Davies, M.O., 1991. Soil Classifica-tion by the cone penetration test: discussion. CanadianGeotechnical Journal, 28(1): 173–176.
Jefferies, M.G., and Davies, M.P., 1993. Use of CPTU to esti-mate equivalent SPT N60. Geotechnical Testing Journal,ASTM, 16(4): 458–468.
19
Juang, C.H., Yuan, H., Lee, D.H., and Lin, P.S., 2003. Simpli-fied Cone Penetration Test-based Method for EvaluatingLiquefaction Resistance of Soils. Journal of Geotechnicaland Geoenvironmental Engineering, ASCE, Vol. 129(1):66–80.
Kishida, H. 1966. Damage to reinforced concrete build-ings in Niigata City and Special reference to foundationengineering. Soils and Foundations. Japanese Societyof Soil Mechanics and Foundation Engineering, 6(1),71–86.
Kulhawy, F.H., and Mayne, P.H., 1990. Manual on estimat-ing soil properties for foundation design, Report EL-6800.Electric Power Research Institute, EPRI, August 1990.
Lacasse, S., and Nadim, F., 1998. Risk and reliability ingeotechnical engineering. State-of-the-Art paper. Pro-ceedings of 4th International Conference on Case His-tories in Geotechnical Engineering. St. Louis, Missouri,Paper No. 50.A.-S.
Ladd, C.C., 1991. Stability evaluation during staged con-struction (22nd Terzaghi Lecture). Journal of Geotechni-cal Engineering, 117(4): 540–615.
Lunne, T., Robertson, P.K., and Powell, J.J.M., 1997. Conepenetration testing in geotechnical practice. BlackieAcademic, EF Spon/Routledge Publ., New York, 1997,312 p.
Matsuda, H., and Ohara, S., 1991. Settlement calculations ofclay layers induced by earthquake. Proceedings 2nd Inter-national Conference on Recent Advances in GeotechnicalEarthquake Engineering and Soil Dynamics. St. Louis,M.O. pp. 473–79.
Moss, R., 2003. CPT-based Probabilistic Assessment of Seis-mic Soil liquefaction initiation. PhD. Thesis. Universityof California, Berkeley.
Moss, R.E.S., Seed, R.B., and Olsen, R.S., 2006.Normalizing the CPT for overburden stress. Journalof Geotechnical and Geoenvironmental Engineering,132(3): 378–387.
NBC 2005. National Building Code of Canada. NationalResearch Council Canada.
Ohara, S., and Matsuda, H., 1988. Study on the settlementof saturated clay layer induced by cyclic shear. Soil andFoundations. Japanese Society for Soil Mechanics andFoundation Engineering 28(3): 103–13.
Olsen, R.S., and Malone, P.G., 1988. Soil classification andsite characterization using the cone penetrometer test.Penetration Testing 1988, ISOPT-1, Edited by De Ruiter,Balkema, Rotterdam, Vol. 2, pp. 887–893.
Olsen, R.S., and Koester, J.P. 1995. Prediction of liquefactionresistance using the CPT. Proceedings of the InternationalSymposium on Cone Penetration Testing, Vol. 2, SwedishGeotechnical Society, Linkoping, pp. 257–262.
Robertson, P.K., 1990. Soil classification using the conepenetration test. Canadian Geotechnical Journal, 27(1):151–158.
Robertson, P.K., 1998. Risk-based site investigation.Geotechnical News: 45–47, September 1998.
Robertson, P.K., 2008. Interpretation of the CPT: a uni-fied approach. Manuscript submitted to the CanadianGeotechnical Journal.
Robertson, P.K., and Campanella, R.G., 1983. Interpreta-tion of cone penetration tests—Part I (sand). CanadianGeotechnical Journal, 20(4): 718–733.
Robertson, P.K. and Campanella, R.G., 1985. LiquefactionPotential of sands using the cone penetration test. Journal
of Geotechnical Engineering Division, ASCE, 111(3):384–406.
Robertson, P.K. and Wride, C.E., 1998. Evaluating cyclicliquefaction potential using the cone penetration test.Canadian Geotechnical Journal, Ottawa, 35(3): 442–459.
Robertson, P.K., Campanella, R.G., Gillespie, D., andRice, A., 1986. Seismic CPT to measure in-situ shear wavevelocity. Journal of Geotechnical Engineering Division,ASCE, 112(8): 791–803.
Sangrey, D.A., Castro, G., Poulos, S.J., and France, J.W.,1978. Cyclic loading of sands, silts and clays. Proceedingsof ASCE Specialty Conference Earthquake Engineeringand Soil Dynamics, Pasadena, Vol. 2 836–851.
SEAOC 1995 VISION 2000. Structural Engineers Associa-tion of California (SEAOC).
Seed, H.B., Tokimatsu, K., Harder, L.F. Jr., and Chung, R.,1984. The influence of SPT Procedures on Soil Liq-uefaction Resistance Evaluation. Report No.UCB/EERC-84/15, Earthquake Engineering ResearchCenter, University of California, Berkeley.
Seed, H.B., and Idriss, I.M., 1971. Simplified Procedurefor evaluating soil liquefaction potential. Journal SoilMechanics and Foundations Div., ASCE 97(SM9),1249–273.
Seed, R.B., Cetin, K.O., Moss, R.E.S., Kammerer, A.,Wu., J., Pestana, J., Riemer, M., Sancio, R.B., Bray,J.D., Kayen, R., and Faris, A., 2003. Recent advancesin soil liquefaction engineering: a unified and consistentframework. Keynote presentation, 26th Annual ASCE LosAngeles Geotechnical Spring Seminar, Long Beach, CA.
Suzuki, Y., Tokimatsu, K., Taya, Y., and Kubota, Y., 1995.Correlation between CPT data and dynamic propertiesof insitu frozen samples. Proceedings 3rd InternationalConf. on Recent Advances in Geotechnical EarhquakeEngineering and Soil Dynamics. Vol. 1, St. Louis, M.O.
Wroth, C.P., 1984. The interpretation of in-situ soil tests.Rankine Lecture, Geotechnique (4).
Wijewickreme, D., and Sanin, M.V., 2007. Effect of Plasticityon the laboratory cyclic shear resistance of fine-grainedsoils. Proceedings 4th International Conf. on EarthquakeGeotechnical Engineering, Thessaloniki, Greece.
Youd, T.L., Idriss, I.M., Andrus, R.D., Arango, I.,Castro, G., Christian, J.T., Dobry, R., Finn, W.D.L.,Harder, L.F., Hynes, M.E., Ishihara, K., Koester, J.P.,Liao, S.S.C., Marcuson, W.F., Martin, G.R., Mitchell,J.K., Moriwaki, Y., Power, M.S., Robertson, P.K., Seed,R.B., and Stokoe, K.H., 2001. Liquefaction Resistanceof soils: summary report from 1996 NCEER and 1998NCEER/NSF workshop on evaluation of liquefactionresistance of soils. J. Geotechnical and GeoenvironmentalEng., ASCE 127(10): 817–33.
Zhang, G., Robertson, P.K., and Brachman, R.W.I., 2002,Estimating Liquefaction induced Ground SettlementsFrom CPT for Level Ground, Canadian GeotechnicalJournal, 39(5): 1168–1180.
Zhang, G., Robertson, P.K., and Brachman, R.W.I., 2004.Estimating Liquefaction induced Lateral Displace-ments using the SPT or CPT. J. Geotechnical andGeoenvironmental Eng., ASCE 130(8): 861–71.
Zhou, S., 1980. Evaluation of the liquefaction of sandby static cone penetration tests. Proceedings 7th WorldConference on Earthquake Engineering, Istanbul, Turkey,Vol. 3, 156–162.
