- Email: [email protected]
A study of the liquefaction risk potential at Yuanlin, Taiwan
Engineering Geology 71 (2003) 97 – 117 www.elsevier.com/locate/enggeo
A study of the liquefaction risk potential at Yuanlin, Taiwan Der-Her Lee a,*, Chih-Sheng Ku b, Haiming Yuan c a
Department of Civil Engineering, National Cheng Kung University, Tainan, Taiwan, ROC b Department of Civil Engineering, I-Shou University, Kaohsiung, Taiwan, ROC c Department of Civil Engineering, Clemson University, 133 Lowry Hall, Clemson, SC 29634, USA
Abstract The method by Iwasaki et al. [Iwasaki, T., Arakawa, T., Tokida, K., 1982. Simplified procedures for assessing soil liquefaction during earthquakes. Proceedings of the Conference on Soil Dynamics and Earthquake Engineering, Southampton, UK, pp. 925 – 939] for evaluating the liquefaction failure potential is widely used in Japan, Taiwan, and other countries due to its ease of use and general applicability. In this method, an index, called the Liquefaction Potential Index (IL), is calculated based on an integration of the calculated factor of safety ( Fs) over depth with a weighting function. Iwasaki et al. (1982) provided a set of criteria to interpret the calculated index IL based on a calibration with his dataset of field performance cases. However, in their method, the factor of safety ( Fs) is based on the liquefaction evaluation method adopted in the Japanese Highway Bridge Design Code [JSHE, 1990. Highway Bridge Design Guide Book. Japan Society of Highway Engineering, in Japanese]. Whether other liquefaction evaluation methods can be used in conjunction with the index IL or not needs further investigation. In this paper, the Cone Penetration Test (CPT) data from Yuanlin, Taiwan, the area that suffered the most from liquefaction in the 1999 Chi-Chi, Taiwan, earthquake, are analyzed. Three CPT-based methods are used for the calculation of the factor of safety for these cases derived from the Chi-Chi earthquake. These factors of safety are used to define the liquefaction potential and risk indexes. The calculated indexes are then used to construct the failure potential maps, and these maps are checked with the field observations. The study shows that the Liquefaction Risk Index (IR) defined in conjunction with the Juang et al. [Juang, C.H., Yuan, H., Lee, D.H., Lin, P.S., 2003. Simplified CPT-based method for evaluating liquefaction potential of soils. Journal of Geotechnical and Geoenvironmental Engineering, ASCE, vol. 129, no. 1, pp. 66 – 80] method yields the best result in interpreting field observations. D 2003 Elsevier Science B.V. All rights reserved. Keywords: Liquefaction; Ground failure; Cone Penetration Test; Probability; Chi-Chi earthquake
1. Introduction In the early morning (01:47 Taiwan time) of September 21, 1999, the largest earthquake of the century in Taiwan (Mw = 7.6, ML = 7.3) struck this * Corresponding author. Fax: +886-6-2757575-63156, 886-62763534. E-mail addresses: [email protected] (D.-H. Lee), [email protected] (C.-S. Ku), [email protected] (H. Yuan).
island country. The epicenter was near Chi-Chi, Nantou, and the hypocentral depth was about 8 km (CGSB, 1999). This earthquake killed more than 2400 persons and caused a great destruction to buildings, bridges, dams, highways, and railways. After this earthquake, sand boiling, serious settlement, and surface ruptures were widely observed in some areas (Lin et al., 1999). The large-scale liquefaction damages in Yuanlin attract the interest and attention of scholars all over the world.
0013-7952/$ - see front matter D 2003 Elsevier Science B.V. All rights reserved. doi:10.1016/S0013-7952(03)00128-5
98
D.-H. Lee et al. / Engineering Geology 71 (2003) 97–117
Being able to evaluate accurately liquefaction potential of soils is the first step toward the mitigation of the damages caused by liquefaction. Simplified methods using in-situ tests, originated by Seed and Idriss (1971), are widely used for this task. The in-situ tests commonly employed for liquefaction evaluation include the Standard Penetration Test (SPT), Cone Penetration Test (CPT) and Shear Wave Velocity Test (Vs). Compared with the other two tests, CPT has the advantages of greater accuracy and repeatability. In addition, with CPT, continuous profiles of soil parameters can be obtained, and thus it is suitable for studying soil characteristics and liquefaction potential. Ishihara (1993) pointed out that CPT is one of the best ways to investigate earthquake-induced liquefaction problems. In the past 15 years, many simplified methods based on CPT have been developed for evaluating soil liquefaction potential (Shibata and Teparaksa, 1988; Stark and Olson, 1995; Olsen, 1997; Robertson and Campanella, 1985; Robertson and Wride, 1998; Chen and Juang, 2000; Juang et al., 2000, in press). Among these methods, the Olsen (1997) method and the Robertson and Wride (1998) method are used more frequently. Lee and Ku (2002) applied these two methods to some liquefied sites in the Chi-Chi earthquake and concluded that both methods were quite
applicable for most cases with liquefaction observations. However, when they were applied to soils with high cone tip resistance ( qc), both methods underestimated liquefaction resistance and factor of safety. Though all of the above methods have their own advantages and disadvantages, and can be used for liquefaction potential evaluation, the factors of safety Fs obtained from different methods are incomparable. Here we take a CPT sounding in Yuanlin as an example. The CPT profiles in Fig. 1 show that except at the depth of 9 and 19 m where the soil is a clay, the soils at this CPT location mainly consist of sandy soils. Within the top 8 m, the factors of safety obtained from all three methods are less than 1 and there is no significant difference among these methods. However, at the depth between 12 and 17 m where the soil has higher cone tip resistance, the results obtained from these methods are quite different. In this layer, the factors of safety by the Olsen method are almost all less than 1, whereas the factors of safety by the Juang method are almost all greater than 1. The factors of safety by the Robertson and Wride method are either less than 1 or greater than 1. Direct comparison of the factors of safety obtained from these methods, however, is not very meaningful because these methods were characterized with differ-
Fig. 1. CPT measurements (YL-2, Yuanlin) and factor of safety against liquefaction.
D.-H. Lee et al. / Engineering Geology 71 (2003) 97–117
ent probabilities (Juang et al., 2002, in press). In other words, the meaning of the calculated factors of safety is not consistent, as some methods are more conservative than others. More important than the question of whether the soil at a given depth will liquefy is the potential of ground failure at a given site. Ishihara (1985) concluded that the potential for liquefaction-induced ground failure was related to the thickness of liquefied soil layers and non-liquefied soil layers. If the thickness of the overburden non-liquefied layer is smaller than the thickness of underlay liquefied layer, ground failure will occur. If the thickness of non-liquefied layer is greater than a threshold value, which depends on the magnitude of the peak horizontal ground acceleration, there will be no ground failure at this site. Iwasaki et al. (1982) proposed the Liquefaction Potential Index IL to evaluate the ground failure risk. The index IL is defined as follows: IL ¼
20 X
F1 W ðzÞdz
ð1Þ
0
where F1 is an index defined as: F1 = 1 Fs, if Fs V 1.0; and F1 = 0 if Fs>1.0. W(z) is a weight function of the depth, which is used to estimate the contribution of soil liquefaction at different depth to the failure of the ground. The weight function is assumed to be a linear function: W ðzÞ ¼ 10 0:5z
ð2Þ
where z is the depth from the ground surface in meters. Iwasaki et al. (1982) used the liquefaction evaluation method recommended in the Japanese Highway Bridge Design Code (JSHE, 1990) to calculate the factor of safety. Based on his analysis of a database of 64 liquefied sites and 23 non-liquefied sites from six earthquakes, Iwasaki et al. (1982) provided the following liquefaction risk criteria, referred to herein as the Iwasaki criteria: IL = 0, the liquefaction failure potential is extremely low; 0 < IL V 5, the liquefaction failure potential is low; 5 < IL V 15, the liquefaction failure potential is high; IL>15, the liquefaction failure potential is extremely high.
99
Due to its ease of use, the Iwasaki et al. (1982) method is widely adopted for the evaluation of the liquefaction failure risk in Japan and Taiwan. However, a question arises as to whether the factor of safety in Eq. (1), calculated from an SPT-based simplified method specified in the Japanese Highway Bridge Design Code (JSHE, 1990), can be replaced with those calculated from any of the three CPT-based methods mentioned previously. In addition, the applicability of the Iwasaki criteria as described previously is also questionable if a different method is used for the calculation of liquefaction resistance and factor of safety against liquefaction. In the present study, three CPT-based methods, the Robertson and Wride (1998) method, the Olsen (1997), and Juang et al. (in press) method, are used to analyze the sites that experienced liquefaction damages and those that did not. A total of 72 sites with CPT measurements are analyzed. These sites are located in Nantou and Yuanlin, the two towns that suffered the most from liquefaction in the Chi-Chi earthquake. The IL index is calculated in conjunction with the three-CPT based methods at this site, and recalibration of this index is carried out based on the results of the analyses. New criteria for interpreting the calculated liquefaction potential index are established. Liquefaction potential and risk contour maps are developed for the town of Yuanlin, based on the new criteria. Comparisons of the developed maps with field observations in the Chi-Chi earthquake are made, which provide a means of validating the new criteria.
2. Liquefaction damages in Yuanlin area According to Taiwan Central Weather Bureau, ground motions recorded at the Yuanlin Elementary School (Station TCU 110) were modest, with the vertical peak ground acceleration (PGA) of 116.28 gal, the S –N peak ground acceleration of 187.32 gal and the E – W peak ground acceleration of 178.24 gal. In this earthquake, the towns of Yuanlin, Nantou, and Wufeng suffered the most from liquefaction damages. Fig. 2 shows settlement of a building due to liquefaction of underlying soils. The maximum differential settlement of the building was about 36 cm and it caused a relatively upward movement of the rain shelter in the front of the building. Fig. 3 shows
100
D.-H. Lee et al. / Engineering Geology 71 (2003) 97–117
Fig. 2. Settlement of a building caused by liquefaction. The maximum differential settlement of 36 cm caused an upward movement of a rain shelter.
tilting of a building in Yuansui Road in Yuanlin. The steel frame shown in the figure provided a temporary support to the tilted building. Also shown in this figure is a 20-ton CPT truck that was in place to perform a CPT sounding. Many investigators performed field reconnaissance immediately after the Chi-Chi earthquake. Sites with liquefaction evidence such as sand boils
and significant settlement were identified and mapped. These maps provided a basis for validation of the methodology developed and presented in this paper. Field and lab tests were conducted by many investigators. In the present study, 102 CPTs were collected and analyzed. Among these, 72 CPTs were conducted at the sites where field liquefaction performance was observed and recorded immediately
Fig. 3. CPT test at a site where tilting of a building caused by liquefaction was performed. The steel frame shown was used to provide a temporary support.
D.-H. Lee et al. / Engineering Geology 71 (2003) 97–117
after the quake. The other 30 CPTs were additional measurements that were intended to supplement existing database for developing liquefaction damage potential maps.
3. Characteristics of soils in the Yunalin area Fig. 4 shows a geologic map Yuanlin. The eastern part of Yuanlin consists of the Toukoshan Formation and terrace deposits. The Toukoshan Formation is distinguished into two lithofacies that are gradational to each other; one is the conglomerate facies (Houyenshan facies) and the other is the sandstone and shale facies (Hsiangshan facies). The terrace deposits are composed mainly of unconsolidated gravel with sandy or silty lenses, generally poorly stratified (Ho, 1988). West to the terrace deposits is alluvial deposits of clay, silt, sand, and gravels, which cover the plain of Yuanlin. Widespread liquefaction was observed in the plain in the 1999 ChiChi earthquake. Soils and geological conditions in Yuanlin are mainly characterized based on data from SPT borings and CPT soundings. Fig. 5 shows a CPT sounding and
101
a SPT borehole conducted at Yuansui Road, Yuanlin, where serious sand boiling and settlement were observed. The site is quite typical among the liquefied sites in the Yuanlin area. At this site, the thickness of top fill layer is about 0.7 m. Beneath the backfill to the depth of 8.1 m is mainly loose sandy soil with low cone tip resistance. The averages of tip resistance and side friction are 2.57 MPa and 19.62 kPa, respectively. From the depth of 8.1 to 10.3 m is a layer of cohesive soil with an average tip resistance of 0.91 MPa and an average side friction of 16.05 kPa. Beneath this cohesive soil layer to the depth of 17.6 m, there is a medium dense to dense sand with an average tip resistance of 10.96 MPa and an average side friction of 24.83 kPa. From 17.6 to 20 m in depth, there is a clay layer with an average tip resistance of 1.34 MPa and an average side friction of 37.25 kPa, respectively. Fig. 6 shows the particle size distributions of the samples collected from different depths at this site, as well as the sand boil samples collected at the ground surface. The particle size distribution curves show that the characteristics of the extruded soils are very similar to those of the soil at the depth of 4 m. This provided strong evidence that the soil at the depth of 4
Fig. 4. Simplified geologic map of Yuanlin.
102
D.-H. Lee et al. / Engineering Geology 71 (2003) 97–117
Fig. 5. CPT and SPT profile of a typical liquefied site in Yuanlin.
m had liquefied. The particle size distribution curves of the samples tested all fell within the ranges of liquefiable soils specified by the Japanese Port and Harbor Research Institute (JPHRI, 1989). These soils are all silty sands (SM) with fines content (FC) in the range of 15% to 17%.
4. CPT-based simplified methods CPT is generally considered a more consistent and repeatable in situ test than SPT (Lunne et al., 1997), and unlike SPT, it can provide a nearly continuous soil profile. This stratigraphic capability
Fig. 6. Particle size distributions of samples taken from a liquefied site.
D.-H. Lee et al. / Engineering Geology 71 (2003) 97–117
makes the CPT particularly advantageous for developing liquefaction-resistance profiles (Youd et al., 2001). However, each in situ test has its advantages and limitations, and thus CPT testing at a site should be verified with a few well-placed SPT boreholes to confirm soil types and the interpretation of liquefaction resistance. In this paper, three CPT-based methods are applied to evaluate liquefaction potential of soils. They are: the Robertson and Wride (1998) method, the Olsen’s (1997) simplified method, and the Juang et al. (2002, in press) method. For convenience, the three methods are referred to herein as the Robertson method, the Olsen method, and the Juang method. It is noted that all simplified methods that follow the general stressbased approach pioneered by Seed and Idriss (1971) require the determination of two variables, namely, cyclic stress ratio (CSR) and cyclic resistance ratio (CRR). To provide a basis for comparing the three CPT-based methods, use of exactly the same procedure for the determination of CSR is required. The sections that follow first describe the procedure for determining CSR, and then briefly summarize the procedure for determining CRR in each of the three methods. 4.1. Evaluation of cyclic stress ratio In this paper, the equation for CSR originally defined by Seed and Idriss (1971) is adjusted to the benchmark earthquake (moment magnitude Mw = 7.5): CSR7:5
rv amax ðrd Þ=MSF ¼ 0:65 rvV g
ð3Þ
rv = the vertical total stress of the soil at the depth studied, rvV= the vertical effective stress of the soil at the depth studied, amax = the peak horizontal ground surface acceleration, g = the acceleration of gravity, rd = the shear stress reduction factor, and MSF = the magnitude scaling factor. The variable rd is calculated as follows (Liao et al., 1988): rd ¼ 1:0 0:00765z; rd ¼ 1:174 0:0267z;
zV9:15 m 9:15 m < zV23 m
ð4Þ
103
The variable MSF is calculated as follows (Idriss, as cited in Youd et al., 2001): n Mw ð5Þ MSF ¼ 7:5 where Mw is the moment magnitude and n is an exponent. In the present study, n is set to be equal to 2.56. 4.2. CRR by the Olsen method This method is based on the Olsen’s soil classification chart-based method for evaluating liquefaction resistance (Olsen, 1988). In this simplified Olsen (1997) method, CRR is calculated as follows: CRR ¼ 0:00128½qc =ðrvVÞ0:7 0:025 þ 0:17Rf 0:028R2f þ 0:0016R3f
ð6Þ
where qc = cone penetration resistance in atm (1 atm is approximately equal to 100 kPa), rvV= effective stress in atm, and Rf = friction ratio in percent, defined as the sleeve friction fs divided by qc. The calculated CRR must be compared to CSR7.5 determined from Eq. (3), and the factor of safety determined as: Fs ¼ CRR=CSR7:5
ð7Þ
The simplified Olsen method is generally conservative. Note that in this method, CRR is strongly affected by the friction ratio (Rf). For a given normalized tip resistance of 100, the CRR determined from Eq. (6) would be 0.136, 0.181, and 0.247 for sands with a Rf value of 0.2, 0.5, and 1.0, respectively. Whether or not the effect of friction ratio should be so significant deserves further investigation. This issue is further complicated by the fact that in a Cone Penetration Test, the measurement of friction ratio is generally much less precise than that of cone tip resistance. Thus, the accuracy of the Olsen method may be compromised by the less precise measurement of the friction ratio. A typical site in Yuanlin with CPT measurements is used to examine the applicability of the Olsen method. YL-C31 is a CPT sounding conducted in
104
D.-H. Lee et al. / Engineering Geology 71 (2003) 97–117
the liquefied area of Yuanlin. The ground water table is 0.6 m below the ground surface. Fig. 7 shows results of the analysis using the Olsen method. The calculated factors of safety Fs along the depth are almost all less than 1.0. This suggests that the liquefaction potential of this site is extremely high, which appears to agree very well with the field observation. However, the soil in the layer of 13.5 to 18 m deep is a medium dense to dense sand with high cone tip resistance in the range of 8.5 to 15.6 MPa. Examination of the CPT soundings at liquefaction sites after the 1999 Chi-Chi earthquake revealed that qc values in the liquefied layers of loose sands after the quake were generally in the extremely low range of 1.13 to 2.57 MPa, suggesting that the strength of liquefied loose sands had been reduced by the occurrence of liquefaction. Thus, it is reasonable to assume that at deeper strata, the sands that retained relatively high qc values experienced no liquefaction in the earthquake. This view is supported by laboratory results by Kobayashi et al. (2001), which reported that the modulus of subgrade reaction of sands was reduced to about 20% of the initial value when the soil was liquefied through shaking, and such results were obtained regardless of
the initial relative density (either at 50% or 80%). However, the modulus of subgrade reaction of sands under shaking remained about the same if liquefaction did not occur. Based on the measured qc values and the above assertion, no liquefaction is judged to have occurred in the sand layer between the depth of 13.5 and 18 m. Thus, it appears that the Olsen method underestimates the liquefaction resistance of this medium dense to dense sand layer, as the average factor of safety by the Olsen method in this layer is 0.67. 4.3. Robertson method In the Robertson method, CRR is calculated as: CRR ¼ 0:833½qc1N;cs =1000 þ 0:05; if qc1N;cs < 50 CRR ¼ 93½qc1N;cs =10003 þ 0:08; if 50Vqc1N;cs < 160
ð8Þ
where qc1N,cs = clean-sand cone tip resistance normalized to a reference stress level of 1 atm (approx-
Fig. 7. Assessment of liquefaction potential at YL-C31 by the Olsen method.
D.-H. Lee et al. / Engineering Geology 71 (2003) 97–117
105
Fig. 8. Assessment of liquefaction potential at YL-C31 by the Robertson method.
imately, 100 kPa). The normalization process, which converts the measured cone tip resistance ( qc) to the normalized cone tip resistance ( qc1N), and the procedure to determine the clean-sand equivalence ( qc1N,cs) from qc1N are described in detail in Robertson and Wride (1998) and Youd et al. (2001). Again, in this method, CSR7.5 must be calculated from Eq. (3) and the factor of safety can be calculated from Eq. (7). The Robertson method follows more closely the SPT-based simplified method originated by Seed and Idriss (1971) and updated in Youd et al. (2001). Eq. (8) basically represents the boundary curve that encompasses liquefied data points in a plot of CRR versus qc1N,cs. Fig. 8 shows the analysis results of YL-C31 using the Robertson method. The Fs profile indicates that this site has a high liquefaction potential, as almost all of the calculated Fs are less than 1.0. This conclusion agrees well with field observations. Similar to the analysis with the Olsen method, the Robertson method also ‘‘predicts’’ liquefaction in the Chi-Chi event for the soil form 13.5 to 18.0 m, as the average factor of safety for this layer is 0.84.
Thus, the Robertson method also underestimates the liquefaction resistance of medium dense to dense sands. 4.4. Juang method Juang et al. (2000) pioneered a procedure with which the boundary curve, called limit state herein, may be derived in a more objective way. This method relies on artificial neural networks that are trained to extract the input –output relationship from a database of field liquefaction performance. The trained neural network is then used to establish limit state. The reader is referred to Juang et al. (2000, 2001) for details of this process. To facilitate the use of the neural network-generated limit state, Juang et al. (in press) performed least-square regression analyses of the data points generated by the neural network, and the following empirical equation was obtained as a whole: CRR ¼ Crexp½2:957 þ 1:264ðqc1N;cs =100Þ1:25 ð9Þ
106
D.-H. Lee et al. / Engineering Geology 71 (2003) 97–117
where Cr ¼ 0:016ðrvV=100Þ3 þ 0:178ðrvV=100Þ2 0:063ðrvV=100Þ þ 0:903
ð10Þ
With CRR calculated from Eq. (9) and CSR from Eq. (3), the factor of safety can be obtained from Eq. (7). Juang et al. (in press) further developed the following mapping function from which the probability of liquefaction for a given factor of safety (calculated from Eqs. (3), (7) and (9)) may be estimated:
qc1N;cs ¼ qc1N ð2:42914c 16:94313c þ 44:551Ic2
1
ð11Þ
PL ¼
qc1N ¼ 10½qc =ðrvVÞ0:5
ð12Þ
Ic ¼ ½ð3:47 logqc1N Þ2 þ ðlogF þ 1:22Þ2 0:5
ð13Þ
F ¼ fs =ðqc rv Þ 100%
ð14Þ
Fig. 9 shows results of the analysis when the case of YL-C31 is re-analyzed using the Juang method. From the derived profile of liquefaction probability, the soils within the top 11.7 m are liquefiable with a liquefaction probability generally greater than 0.85, which agrees well with field observations. Unlike the other two CPT-based methods described previously, however, the Juang method predicted little liquefaction potential (with an average probability of about 15%) in the layer between the depths of 13.5 and 18 m. This shows that the Juang method is more accurate in evaluating the medium dense to dense sandy soils than the Olsen method and the Robertson method, although all three methods correctly predicted liquefaction potential of loose sands in the Chi-Chi event.
51:497Ic þ 22:802Þ
In the above equations, cone tip resistance ( qc), sleeve friction ( fs), total overburden stress (rv), and effective overburden stress (rV v) must all be in the unit of kPa. The variables F, Ic, qc1N, qc1N,cs, and CRR are all dimensionless. Note that Eqs. (12), (13) and (14) were taken from Robertson and Wride (1998) with slight modification in Eq. (13).
1 þ ðFs =0:96Þ4:5
Fig. 9. Assessment of liquefaction potential at YL-C31 by the Juang method.
ð15Þ
D.-H. Lee et al. / Engineering Geology 71 (2003) 97–117
5. Liquefaction failure potential 5.1. Liquefaction potential index Iwasaki et al. (1982) proposed the Liquefaction Potential Index (IL) to estimate the liquefaction damage or failure potential. Whereas liquefaction potential as defined in Seed and Idriss (1971) and Youd et al. (2001) is evaluated for a soil at a particular depth in the subsurface, the IL index is an integrated effect of the likely liquefaction over the entire depth of the profile (to be more specific, down to 20 m below the ground surface). Whereas this index is not without pitfalls, it is easy to use and particularly useful for mapping liquefaction damage or failure potential of an area. As mentioned previously, however, the index IL needs to be re-calibrated each time a different method is used for evaluating liquefaction resistance and potential because different degrees of conservatism were built into the existing empirical methods. Fig. 10 shows the distribution of the calculated IL values obtained by Iwasaki et al. (1982) based on the analysis of their database. More than 75% of non-liquefied cases had an IL value of less than 5, and almost all non-liquefied cases had an IL value of less than 15. Among the liquefied cases, 50% of them had an IL value of greater than 15. However, 20% of the liquefied cases were found to have an IL value of less than 5. Based on these results, Iwasaki et al. (1982) proposed four classes of liquefaction failure potential,
107
as described previously. The threshold IL value is 5 for high liquefaction failure potential, and is 15 for extremely high potential. In the present study, the Olsen method is first used to analyze the 72 CPT soundings each with observation of field performance (whether liquefactioninduced ground failure was observed). For each CPT profile, the factor of safety is calculated at each depth continuously over the depth of 20 m, as suggested by Iwasaki et al. (1982). In other words, a profile of factor of safety is obtained. This profile is then used to calculate the Liquefaction Potential Index IL according to Eq. (1). The distribution of the calculated IL values of the 72 CPTs is shown in Fig. 11. The results show 75% of liquefied cases with IL greater than 5, and 50% of liquefied cases with IL greater than 15. The results seem to agree well with the Iwasaki criteria. However, only 25% of nonliquefied cases have an IL value of less than 5. Recall that in database used by Iwasaki et al. (1982), 70% of non-liquefied cases were found to have an IL value of less than 5. This suggests that the Iwasaki criteria may not be applicable when the liquefaction resistance and factor of safety are calculated using the Olsen method. When applying the Olsen method without re-calibration, many of the non-liquefied sites would have been predicted to liquefy in the Chi-Chi earthquake according to the Iwasaki criteria. Another way to examine the Olsen method is to construct a contour map of liquefaction failure poten-
Fig. 10. Distribution of calculated IL values of liquefied group and non-liquefied group by Iwasaki et al. (1982).
108
D.-H. Lee et al. / Engineering Geology 71 (2003) 97–117
Fig. 11. Distribution of calculated IL values of liquefied group and non-liquefied group of the 72 CPTs analyzed by the Olsen method.
tial based on the calculated IL values. Fig. 12 shows a contour map of the calculated IL values in Yuanlin. Note that in this figure, the triangle symbol indicates the location of the CPT, and the shaded area with grids at the right belongs to the hill of the Central Mountain Range that consists mainly of gravelly cobble deposits
in which liquefaction is not an issue. The contour lines of IL = 5 and 15, the two bounds adopted by Iwasaki et al. (1982) in his criteria for liquefaction-induced failure potential, are drawn. The observed liquefaction sites in the Chi-Chi earthquake are marked with dark shades. Although many severe liquefaction sites are
Fig. 12. CPT locations and contour of IL values at Yuanlin by the Olsen method. The IL values of 5 and 15 are the lower and upper boundary values suggested by Iwasaki et al. (1982).
D.-H. Lee et al. / Engineering Geology 71 (2003) 97–117
109
Fig. 13. Distribution of calculated IL values of liquefied group and non-liquefied group of the 72 CPTs analyzed by the Robertson method.
located in the zone with IL>15, other areas without liquefaction sites also show IL>15. In addition, almost the entire town has an IL>5, indicating high liquefaction failure potential, which cannot explain many no-liquefaction sites. This again suggests that the Olsen method is too conservative when it is
used in conjunction with the Iwasaki criteria for assessing the liquefaction failure potential. On the other hand, it points to the need of establishing new criteria for assessing liquefaction failure potential in conjunction with the use of the Olsen method.
Fig. 14. CPT locations and contour of IL values at Yuanlin by the Robertson method. The IL values of 15 are the upper boundary values suggested by Iwasaki et al. (1982).
110
D.-H. Lee et al. / Engineering Geology 71 (2003) 97–117
Fig. 15. Distribution of calculated IL values of liquefied group and non-liquefied group of the 72 CPTs analyzed by the Juang method (Ic < 2.95).
The same 72 cases are analyzed using the Robertson method, and the index IL is calculated for each CPT profile. Fig. 13 shows the distribution of the calculated IL values. As shown in Fig. 12, only 10% of liquefied cases have an IL value of less than 5, whereas there are 50% of liquefied cases with IL
greater than 15. For liquefied cases, the results of the analysis using the Robertson method agree well with field observations. However, the results also show that the calculated IL values are generally too high. For example, 85% of non-liquefied cases have an IL value of greater than 5, and 30% of non-
Fig. 16. CPT locations and contour of IL values at Yuanlin by the Juang method (Ic < 2.95). The IL values of 15 are the upper boundary values suggested by Iwasaki et al. (1982).
D.-H. Lee et al. / Engineering Geology 71 (2003) 97–117
111
Fig. 17. Distribution of calculated IL values of liquefied group and non-liquefied group of the 72 CPTs analyzed by the Juang method (Ic < 2.4).
liquefied cases have an IL value of greater than 15. This indicates that the Iwasaki criteria as described previously are not applicable when the liquefaction resistance is calculated using the Robertson method. Fig. 14 shows a contour map of the calculated IL values in Yuanlin where the liquefaction resistance is calculated using the Robertson method. As in the analysis using the Olsen method, this contour map shows that the Robertson method is too conservative
when it is used in conjunction with the Iwasaki criteria for assessing the liquefaction failure potential. The need to establish new criteria for assessing liquefaction failure potential in conjunction with the use of the Robertson method is obvious. Fig. 15 shows the distribution of I L values obtained by using the Juang method. The results indicate that there is no clear-cut boundary between IL values of liquefied cases and those of non-liquefied
Table 1 Liquefaction potential index (IL) and liquefaction risk index (IR) at various sites CPT
NT-1 NT-C2 NT-C7 NT-C15 YL-2 YL-C22 YL-C24 YL-C25 YL-C31 YL-C43 YL-C7 YL-C8 YL-C9 YL-C10 YL-C11 YL-C13 YL-C15 YL-C16 YL-C28 YL-C36
IL (Juang)
IR (Juang)
Ic < 2.95
Ic < 2.4
Ic < 2.4
35 43 52 53 27 20 25 23 28 16 16 16 14 15 20 16 14 25 18 12
17 27 28 23 21 17 17 11 17 8 11 1 4 9 5 5 5 6 7 1
29 40 43 34 50 42 40 30 38 23 32 5 15 21 10 15 18 14 17 2
IL (Olsen)
IL (Robertson)
Liquefaction?
22 34 45 40 25 20 16 9 34 9 16 4 7 7 11 6 4 16 8 8
21 33 36 27 25 21 23 19 27 14 17 6 10 13 10 13 12 12 13 7
Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes No No No No No No No No No No
112
D.-H. Lee et al. / Engineering Geology 71 (2003) 97–117
cases. Fig. 16 shows the contour map of IL values in Yuanlin where the liquefaction resistance is calculated using the Juang method. As shown in Fig. 16, almost all IL values in the entire town of Yuanlin are above 15. This suggests that use of the IL values calculated with the Juang method to judge whether liquefactioninduced ground failure will occur is unlikely to succeed. An examination of the Juang method reveals that the upper bound of the soil classification index Ic for liquefied soil is 2.95 in this method. This may lead to an overestimate of liquefaction potential of silty soils and thus the liquefaction failure potential index IL. Ku (2001) studied the characteristics of the index Ic for the soils in the middle west of Taiwan, which covers all liquefaction sites in the Chi-Chi earthquake, and concluded that the upper bound for typical liquefied soils (SP, SP – SM, SM) in the Yuanlin area should be about 2.4. By placing an upper bound of Ic = 2.4 in the Juang method, and recalculating the IL values for all CPTs, a new distribution of IL is obtained, as shown in Fig. 17. In this figure, the difference of IL values between
liquefied and non-liquefied cases is obvious. About 50% of non-liquefied cases have an IL value of less than 5, and all non-liquefied cases are found to have an IL value of less than 15. For liquefied cases, only about 20% have an IL value of less than 5. The modified Juang method is shown to be compatible with the Iwasaki criteria for liquefaction failure potential evaluation. As mentioned previously, it is necessary to recalibrate the Iwasaki criteria if a new method for evaluating liquefaction resistance and potential is to be used. To this end, 20 CPTs that were identified with clear-cut liquefaction failure or no failure are examined in detail. At the locations of 10 of these CPTs, liquefaction induced failure were observed, and at the other 10, no failure was observed. Table 1 shows the calculated IL values in conjunction with the use of each of the three CPT-based methods for liquefaction resistance evaluation. With respect to the Olsen method, among the 10 liquefied cases, 80% were found to have IL>15, which is reasonable; and among the 10 non-liquefied cases, 20% were found to have IL < 5, which is not reasonable. Thus,
Fig. 18. Liquefaction potential map of Yuanlin by the Olsen method. The IL values of 8 and 16 are the lower and upper boundary values that provide the best delineation of liquefaction failure potential classes.
D.-H. Lee et al. / Engineering Geology 71 (2003) 97–117
in order to use the Olsen method, the two classification bounds, 5 and 15, in the Iwasaki criteria need to be adjusted. It is postulated that the higher bound should be taken as the one that encompasses (with IL equal or greater than) 70% of liquefied cases, and the lower bound should be taken as the one that encompasses (with IL equal or less than) 70% of nonliquefied cases. Under this postulation, the boundaries for high failure potential and low failure potential are 16 and 8, respectively. In other words, the ground failure potential is low if IL < 8, and the ground failure potential is extremely high if IL>16. The ground failure potential is high if 8 < IL V 16. Using this set of criteria, the contour map of IL values calculated with the Olsen method is redrawn, as shown in Fig. 18. From this figure, it can be seen that most liquefied sites are within the zone of IL>16 and there is little liquefied cases within the zone of IL < 8. The recalibrated IL index improves the accuracy of the Olsen method for such purpose. Similarly, the Iwasaki criteria are not applicable when the Robertson method is used for liquefaction resistance evaluation because all of the non-liquefied
113
cases have an IL value of greater than 5. Using the same postulation regarding the higher and lower bounds of IL, as in the case of the Olsen method, the two bounds are determined to 21 and 13, respectively. Thus, the ground failure potential is low if IL < 13, and the ground failure potential is extremely high if IL>21. The ground failure potential is high if 13 < IL V 21. With this new set of criteria, the contour map of IL values is redrawn, as shown in Fig. 19, and the results show that most liquefied sites are within the zone of IL>21 and there is little liquefied cases within the zone of IL < 13. The re-calibrated IL index improves the accuracy of the Robertson method for such purpose. With the new set of criteria, the Robertson method is able to predict fairly accurately the liquefaction-induced ground failure potential. The Iwasaki criteria are not applicable either when the Juang method is used for liquefaction resistance evaluation because there are only three non-liquefied cases with IL < 15. However, with the limit of Ic < 2.4, the modified Juang method is quite accurate without adjustment to the Iwasaki criteria. About 80% of
Fig. 19. Liquefaction potential map of Yuanlin by the Robertson method. The IL values of 13 and 21 are the lower and upper boundary values that provide the best delineation of liquefaction failure potential classes.
114
D.-H. Lee et al. / Engineering Geology 71 (2003) 97–117
liquefied cases have IL values greater than 15, and 60% of non-liquefied cases have IL values of less than 5. Fig. 20 shows the contour of IL values that are obtained with the modified Juang method. As may be seen from this figure, almost all liquefied sites are within the zone of IL>15, and non-liquefied sites are located in the zone of IL < 5. This suggests that the modified Juang method is able to predict quite accurately the ground failure potential without adjustment to the Iwasaki criteria. 5.2. Liquefaction risk index The liquefaction potential index IL is defined with the profile of the factor of safety, and only those with Fs < 1 contribute to the index IL. As noted previously, the factor of safety from different methods essentially has different meaning since each method was developed with different degree of conservatism. In addition, the liquefaction potential is not linearly proportional to the factor of safety; rather, it is linearly proportional to the probability of liquefaction. Thus,
a new index, called Liquefaction Risk Index (IR) is defined using the probability of liquefaction: IR ¼
20 X
PL wðzÞdz
ð16Þ
0
where PL is the probability of liquefaction obtained from Eq. (15); W(z) is the weight function as defined in Eq. (2). Eq. (16) is seen as a variable form of Eq. (1). Re-analysis of the 72 CPTs with the modified Juang method (Ic = 2.4 as the upper bound of liquefied cases) coupling with the definition of IR in Eq. (16) yields distributions of IR for liquefied cases and for non-liquefied cases, as shown in Fig. 21. The distributions are deemed more reasonable than any of the previously obtained distributions by other methods. Following the same postulation in which 70 percentile in the IR distribution is used to determine lower and higher bounds for classification, the two bounds are determined to be about 20 and 30, respectively. Thus, the ground failure potential is
Fig. 20. Liquefaction potential map of Yuanlin by the Juang method (Ic < 2.4). The IL values of 5 and 15 are the lower and upper boundary values that provide the best delineation of liquefaction failure potential classes, which are the same as in the Iwasaki criteria.
D.-H. Lee et al. / Engineering Geology 71 (2003) 97–117
115
Fig. 21. Distribution of calculated IR values of liquefied group and non-liquefied group of the 72 CPTs analyzed by the Juang method (Ic < 2.4).
low if IR < 20, and the ground failure potential is extremely high if IR>30. The ground failure potential is high if 20 < IR V 30. Fig. 22 shows a contour map of the calculated IR values. Almost all of the liquefied sites locate in the zone with IR>30, and almost all of the non-liquefied sites are in the zone of IR < 20. The contour map shows that the formulation
of IR and the criteria for assessing liquefactioninduced ground failure are reasonable. In summary, three CPT-based methods have been examined for their applicability of being integrated into the liquefaction potential index (IL) defined by Iwasaki et al. (1982). The study concludes that recalibration of the calculated failure potential index IL
Fig. 22. Contour map of liquefaction risk index (IR) at Yuanlin by the Juang method.
116
D.-H. Lee et al. / Engineering Geology 71 (2003) 97–117
is needed, regardless of which CPT-based method is used for determining liquefaction resistance. The modified Juang method is further used to define the Liquefaction Risk Index (IR), which provides the best overall results in assessing liquefaction-induced failure potential. The contour maps of IL (or IR) along with field observations provide an effective means to validate the methodology and procedures for assessing liquefaction-induced failure potential of a site presented in this paper. As a final note, the accuracy of the contour maps may be improved by additional well-placed CPTs in the Yuanlin area, although the indexes IL and IR and the associated criteria established in this paper will remain the same, since no field observations will be available with these CPTs.
6. Conclusions Based on the analysis of limited CPT data presented in this paper, the following conclusions are reached: (1) All three CPT-based methods examined in this paper can predict accurately liquefied cases in the 1999 Chi-Chi earthquake. The Juang method, however, is shown to be more accurate in evaluating liquefaction resistance of the medium dense to dense sand with high cone resistance, the soil that is generally considered non-liquefiable. (2) When the CPT-based methods are integrated into the framework of the Liquefaction Potential Index (IL) defined by Iwasaki et al. (1982), the criteria for assessing the liquefaction failure potential need to be re-calibrated. Once re-calibrated, each method, be it the Robertson method, the Olsen method, or the Juang method, may be used to assess the liquefaction failure potential. For the Olsen method, the ground failure potential is low if IL < 8, and the ground failure potential is extremely high if IL>16. For the Robertson method, the ground failure potential is low if IL < 13, and the ground failure potential is extremely high if IL>21. However, for the Juang method with liquefaction upper bound of Ic = 2.4, the established criteria are consistent with the Iwasaki criteria (low if IL < 5; extremely high if IL>15).
(3) The Liquefaction Risk Index (IR) developed in this paper shows an improvement over the Liquefaction Potential Index (IL) for predicting the liquefaction-induced failure potential. The term IR is defined with the modified Juang method, and the following criteria are established for assessing the liquefactioninduced failure potential: (1) the failure potential is high if IR z 30, (2) the failure potential is medium if 20 z IR>30, and (3) the failure potential is low if IR < 20. The contour map of IR values prepared for the town of Yuanlin can identify accurately the liquefaction and no-liquefaction sites in the Chi-Chi earthquake.
Acknowledgements The study on which this paper is based was sponsored by National Science Council, Taipei, Taiwan through Grant No. NSC 90-2611-E-006-029. This financial support is greatly appreciated. The authors are indebted to Dr. C. Hsein Juang of Clemson University who first communicated to the authors the need of re-calibration of the Iwasaki criteria, and provided many helpful suggestions during the course of this study. Dr. John Christian, a peer-reviewer, and another anonymous reviewer are thanked for their constructive comments that help sharpen this paper.
References CGSB, 1999. Investigation Report of the 1999, Chi-Chi Earthquake. Central Geological Survey Bureau (CGSB), Republic of China, Taipei, Taiwan. In Chinese. Chen, C.J., Juang, C.H., 2000. Calibration of SPT- and CPTbased liquefaction evaluation methods. In: Mayne, P., Hryciw, R. (Eds.), Innovations Applications in Geotechnical Site Characterization. Geotechnical Special Publication, vol. 97. ASCE, Reston, VA, pp. 49 – 64. Ho, C.S., 1988. An introduction to the geology of Taiwan and explanatory text of the geologic map of Taiwan. Central Geological Survey. The Ministry of Economic Affairs, Taipei, Taiwan. Ishihara, K., 1985. Stability of natural deposits during earthquake. Proceedings of the Eleventh International Conference on Soil Mechanics and Foundation Engineering, San Francisco, California, vol. 1, pp. 321 – 376. Ishihara, K., 1993. Liquefaction and flow failure during earthquakes. Ge´otechnique, The Institution of Civil Engineers 43 (3), 351 – 415. Iwasaki, T., Arakawa, T., Tokida, K., 1982. Simplified procedures
D.-H. Lee et al. / Engineering Geology 71 (2003) 97–117 for assessing soil liquefaction during earthquakes. Proceedings of the Conference on Soil Dynamics and Earthquake Engineering, Southampton, UK, pp. 925 – 939. JPHRI, 1989. Port and Harbor Facilities Design Guide Book. Japan Port and Harbor Research Institute, Tokyo. In Japanese. JSHE, 1990. Highway Bridge Design Guide Book. Japan Society of Highway Engineering, Tokyo. In Japanese. Juang, C.H., Chen, C.J., Tang, W.H., Rosowsky, D.V., 2000. CPTbased liquefaction analysis: Part I. Determination of limit state function. Ge´otechnique, The Institution of Civil Engineers 50 (5), 583 – 592. Juang, C.H., Chen, C.J., Jiang, T., 2001. Probabilistic framework for liquefaction potential by shear wave velocity. Journal of Geotechnical and Geoenvironmental Engineering, ASCE 127 (8), 670 – 678. Juang, C.H., Yuan, H., Lee, D.H., Ku, C.S., 2002. Assessing CPTbased methods for liquefaction evaluation with emphasis on the cases from the Chi-Chi, Taiwan, earthquake. International Journal of Soil Dynamics and Earthquake Engineering 22 (3), 241 – 258. Juang, C.H., Yuan, H., Lee, D.H., Lin, P.S., 2003. A simplified CPT-based method for evaluating liquefaction potential of soils. Journal of Geotechnical and Geoenvironmental Engineering 129 (1), 66 – 80. Kobayashi, H., Morimitsu, T., Azuma, T., Tamura, K., 2001. Experimental study on reduction of the horizontal subgrade reaction due to liquefaction. Journal of Japan Civil Engineering 43 (2), 46 – 51. In Japanese. Ku, C.S., 2001. ‘‘A study on soil classification and liquefaction evaluation based on CPT.’’ PhD Dissertation. National Cheng Kung University, Tainan, Taiwan. In Chinese. Lee, D.H., Ku, C.S., 2002. Application of CPT in the Liquefaction Investigation in the central Taiwan. Proceeding, Conference on Liquefaction Potential Investigation in Central Taiwan, Tainan, Taiwan. In Chinese. Liao, S.C.C., Veneziano, D., Whitman, R.V., 1988. Regression models for evaluating liquefaction probability. Journal of Geotechnical Engineering, ASCE 114 (4), 389 – 411. Lin, M.L., Liao, H.J., Ueng, T.S., 1999. A study on the geotechnical-
117
related damages in the Chi-Chi Earthquake. Civil and Hydraulic Engineering 26 (3), 60 – 71 (in Chinese). Lunne, T., Robertson, P.K., Powell, J.J.M., 1997. Cone Penetration Testing. Blackie Academic and Professional, London, UK. Olsen, R.S., 1988. Using the CPT for dynamic site response characterization. In: Von Thum, J.L. (Ed.), Proceedings, Earthquake Engineering and Soil Dynamics II Conference. Geotechnical Special Publication, vol. 2. ASCE, New York, pp. 374 – 388. Olsen, R.S., 1997. Cyclic liquefaction based on the cone penetration test. In: Youd, T.L., Idriss, I.M. (Eds.), Proceedings of the NCEER Workshop of Evaluation of Liquefaction Resistance of Soils, Technical report NCEER-97-0022. National Center for Earthquake Engineering Research, State University of New York at Buffalo, Buffalo, NY, pp. 225 – 276. Robertson, P.K., Campanella, R.G., 1985. Liquefaction potential of sands using the CPT. Journal of Geotechnical Engineering, ASCE 111 (3), 298 – 307. Robertson, P.K., Wride, C.E., 1998. Evaluating cyclic liquefaction potential using the cone penetration test. Canadian Geotechnical Journal 35 (3), 442 – 459. Seed, H.B., Idriss, I.M., 1971. Simplified procedure for evaluating soil liquefaction potential. Journal of the Soil Mechanics and Foundation Division, ASCE 97 (9), 1249 – 1273. Shibata, T., Teparaksa, W., 1988. Evaluation of liquefaction potential of soils using cone penetration tests. Soils and Foundations 28 (2), 49 – 60. Stark, T.D., Olson, S.M., 1995. Liquefaction resistance using CPT and field case histories. Journal of Geotechnical Engineering, ASCE 121 (GT 12), 856 – 869. Youd, T.L., Idriss, I.M., Andrus, R.D., Arango, I., Castro, G., Christian, J.T., Dobry, R., Liam Finn, W.D., Harder Jr., L.F., Hynes, M.E., Ishihara, K., Koester, J.P., Laio, S.S.C., Marcuson III, W.F., Martin, G.R., Mitchell, J.K., Moriwaki, Y., Power, M.S., Robertson, P.K., Seed, R.B., Stokoe II, K.H., 2001. Liquefaction resistance of soils: summary report from the 1996 NCEER and 1998 NCEER/NSF workshops on evaluation of liquefaction resistance of soils. Journal of Geotechnical and Geoenvironmental Engineering 127 (10), 817 – 833.
A study of the liquefaction risk potential at Yuanlin, Taiwan Der-Her Lee a,*, Chih-Sheng Ku b, Haiming Yuan c a
Department of Civil Engineering, National Cheng Kung University, Tainan, Taiwan, ROC b Department of Civil Engineering, I-Shou University, Kaohsiung, Taiwan, ROC c Department of Civil Engineering, Clemson University, 133 Lowry Hall, Clemson, SC 29634, USA
Abstract The method by Iwasaki et al. [Iwasaki, T., Arakawa, T., Tokida, K., 1982. Simplified procedures for assessing soil liquefaction during earthquakes. Proceedings of the Conference on Soil Dynamics and Earthquake Engineering, Southampton, UK, pp. 925 – 939] for evaluating the liquefaction failure potential is widely used in Japan, Taiwan, and other countries due to its ease of use and general applicability. In this method, an index, called the Liquefaction Potential Index (IL), is calculated based on an integration of the calculated factor of safety ( Fs) over depth with a weighting function. Iwasaki et al. (1982) provided a set of criteria to interpret the calculated index IL based on a calibration with his dataset of field performance cases. However, in their method, the factor of safety ( Fs) is based on the liquefaction evaluation method adopted in the Japanese Highway Bridge Design Code [JSHE, 1990. Highway Bridge Design Guide Book. Japan Society of Highway Engineering, in Japanese]. Whether other liquefaction evaluation methods can be used in conjunction with the index IL or not needs further investigation. In this paper, the Cone Penetration Test (CPT) data from Yuanlin, Taiwan, the area that suffered the most from liquefaction in the 1999 Chi-Chi, Taiwan, earthquake, are analyzed. Three CPT-based methods are used for the calculation of the factor of safety for these cases derived from the Chi-Chi earthquake. These factors of safety are used to define the liquefaction potential and risk indexes. The calculated indexes are then used to construct the failure potential maps, and these maps are checked with the field observations. The study shows that the Liquefaction Risk Index (IR) defined in conjunction with the Juang et al. [Juang, C.H., Yuan, H., Lee, D.H., Lin, P.S., 2003. Simplified CPT-based method for evaluating liquefaction potential of soils. Journal of Geotechnical and Geoenvironmental Engineering, ASCE, vol. 129, no. 1, pp. 66 – 80] method yields the best result in interpreting field observations. D 2003 Elsevier Science B.V. All rights reserved. Keywords: Liquefaction; Ground failure; Cone Penetration Test; Probability; Chi-Chi earthquake
1. Introduction In the early morning (01:47 Taiwan time) of September 21, 1999, the largest earthquake of the century in Taiwan (Mw = 7.6, ML = 7.3) struck this * Corresponding author. Fax: +886-6-2757575-63156, 886-62763534. E-mail addresses: [email protected] (D.-H. Lee), [email protected] (C.-S. Ku), [email protected] (H. Yuan).
island country. The epicenter was near Chi-Chi, Nantou, and the hypocentral depth was about 8 km (CGSB, 1999). This earthquake killed more than 2400 persons and caused a great destruction to buildings, bridges, dams, highways, and railways. After this earthquake, sand boiling, serious settlement, and surface ruptures were widely observed in some areas (Lin et al., 1999). The large-scale liquefaction damages in Yuanlin attract the interest and attention of scholars all over the world.
0013-7952/$ - see front matter D 2003 Elsevier Science B.V. All rights reserved. doi:10.1016/S0013-7952(03)00128-5
98
D.-H. Lee et al. / Engineering Geology 71 (2003) 97–117
Being able to evaluate accurately liquefaction potential of soils is the first step toward the mitigation of the damages caused by liquefaction. Simplified methods using in-situ tests, originated by Seed and Idriss (1971), are widely used for this task. The in-situ tests commonly employed for liquefaction evaluation include the Standard Penetration Test (SPT), Cone Penetration Test (CPT) and Shear Wave Velocity Test (Vs). Compared with the other two tests, CPT has the advantages of greater accuracy and repeatability. In addition, with CPT, continuous profiles of soil parameters can be obtained, and thus it is suitable for studying soil characteristics and liquefaction potential. Ishihara (1993) pointed out that CPT is one of the best ways to investigate earthquake-induced liquefaction problems. In the past 15 years, many simplified methods based on CPT have been developed for evaluating soil liquefaction potential (Shibata and Teparaksa, 1988; Stark and Olson, 1995; Olsen, 1997; Robertson and Campanella, 1985; Robertson and Wride, 1998; Chen and Juang, 2000; Juang et al., 2000, in press). Among these methods, the Olsen (1997) method and the Robertson and Wride (1998) method are used more frequently. Lee and Ku (2002) applied these two methods to some liquefied sites in the Chi-Chi earthquake and concluded that both methods were quite
applicable for most cases with liquefaction observations. However, when they were applied to soils with high cone tip resistance ( qc), both methods underestimated liquefaction resistance and factor of safety. Though all of the above methods have their own advantages and disadvantages, and can be used for liquefaction potential evaluation, the factors of safety Fs obtained from different methods are incomparable. Here we take a CPT sounding in Yuanlin as an example. The CPT profiles in Fig. 1 show that except at the depth of 9 and 19 m where the soil is a clay, the soils at this CPT location mainly consist of sandy soils. Within the top 8 m, the factors of safety obtained from all three methods are less than 1 and there is no significant difference among these methods. However, at the depth between 12 and 17 m where the soil has higher cone tip resistance, the results obtained from these methods are quite different. In this layer, the factors of safety by the Olsen method are almost all less than 1, whereas the factors of safety by the Juang method are almost all greater than 1. The factors of safety by the Robertson and Wride method are either less than 1 or greater than 1. Direct comparison of the factors of safety obtained from these methods, however, is not very meaningful because these methods were characterized with differ-
Fig. 1. CPT measurements (YL-2, Yuanlin) and factor of safety against liquefaction.
D.-H. Lee et al. / Engineering Geology 71 (2003) 97–117
ent probabilities (Juang et al., 2002, in press). In other words, the meaning of the calculated factors of safety is not consistent, as some methods are more conservative than others. More important than the question of whether the soil at a given depth will liquefy is the potential of ground failure at a given site. Ishihara (1985) concluded that the potential for liquefaction-induced ground failure was related to the thickness of liquefied soil layers and non-liquefied soil layers. If the thickness of the overburden non-liquefied layer is smaller than the thickness of underlay liquefied layer, ground failure will occur. If the thickness of non-liquefied layer is greater than a threshold value, which depends on the magnitude of the peak horizontal ground acceleration, there will be no ground failure at this site. Iwasaki et al. (1982) proposed the Liquefaction Potential Index IL to evaluate the ground failure risk. The index IL is defined as follows: IL ¼
20 X
F1 W ðzÞdz
ð1Þ
0
where F1 is an index defined as: F1 = 1 Fs, if Fs V 1.0; and F1 = 0 if Fs>1.0. W(z) is a weight function of the depth, which is used to estimate the contribution of soil liquefaction at different depth to the failure of the ground. The weight function is assumed to be a linear function: W ðzÞ ¼ 10 0:5z
ð2Þ
where z is the depth from the ground surface in meters. Iwasaki et al. (1982) used the liquefaction evaluation method recommended in the Japanese Highway Bridge Design Code (JSHE, 1990) to calculate the factor of safety. Based on his analysis of a database of 64 liquefied sites and 23 non-liquefied sites from six earthquakes, Iwasaki et al. (1982) provided the following liquefaction risk criteria, referred to herein as the Iwasaki criteria: IL = 0, the liquefaction failure potential is extremely low; 0 < IL V 5, the liquefaction failure potential is low; 5 < IL V 15, the liquefaction failure potential is high; IL>15, the liquefaction failure potential is extremely high.
99
Due to its ease of use, the Iwasaki et al. (1982) method is widely adopted for the evaluation of the liquefaction failure risk in Japan and Taiwan. However, a question arises as to whether the factor of safety in Eq. (1), calculated from an SPT-based simplified method specified in the Japanese Highway Bridge Design Code (JSHE, 1990), can be replaced with those calculated from any of the three CPT-based methods mentioned previously. In addition, the applicability of the Iwasaki criteria as described previously is also questionable if a different method is used for the calculation of liquefaction resistance and factor of safety against liquefaction. In the present study, three CPT-based methods, the Robertson and Wride (1998) method, the Olsen (1997), and Juang et al. (in press) method, are used to analyze the sites that experienced liquefaction damages and those that did not. A total of 72 sites with CPT measurements are analyzed. These sites are located in Nantou and Yuanlin, the two towns that suffered the most from liquefaction in the Chi-Chi earthquake. The IL index is calculated in conjunction with the three-CPT based methods at this site, and recalibration of this index is carried out based on the results of the analyses. New criteria for interpreting the calculated liquefaction potential index are established. Liquefaction potential and risk contour maps are developed for the town of Yuanlin, based on the new criteria. Comparisons of the developed maps with field observations in the Chi-Chi earthquake are made, which provide a means of validating the new criteria.
2. Liquefaction damages in Yuanlin area According to Taiwan Central Weather Bureau, ground motions recorded at the Yuanlin Elementary School (Station TCU 110) were modest, with the vertical peak ground acceleration (PGA) of 116.28 gal, the S –N peak ground acceleration of 187.32 gal and the E – W peak ground acceleration of 178.24 gal. In this earthquake, the towns of Yuanlin, Nantou, and Wufeng suffered the most from liquefaction damages. Fig. 2 shows settlement of a building due to liquefaction of underlying soils. The maximum differential settlement of the building was about 36 cm and it caused a relatively upward movement of the rain shelter in the front of the building. Fig. 3 shows
100
D.-H. Lee et al. / Engineering Geology 71 (2003) 97–117
Fig. 2. Settlement of a building caused by liquefaction. The maximum differential settlement of 36 cm caused an upward movement of a rain shelter.
tilting of a building in Yuansui Road in Yuanlin. The steel frame shown in the figure provided a temporary support to the tilted building. Also shown in this figure is a 20-ton CPT truck that was in place to perform a CPT sounding. Many investigators performed field reconnaissance immediately after the Chi-Chi earthquake. Sites with liquefaction evidence such as sand boils
and significant settlement were identified and mapped. These maps provided a basis for validation of the methodology developed and presented in this paper. Field and lab tests were conducted by many investigators. In the present study, 102 CPTs were collected and analyzed. Among these, 72 CPTs were conducted at the sites where field liquefaction performance was observed and recorded immediately
Fig. 3. CPT test at a site where tilting of a building caused by liquefaction was performed. The steel frame shown was used to provide a temporary support.
D.-H. Lee et al. / Engineering Geology 71 (2003) 97–117
after the quake. The other 30 CPTs were additional measurements that were intended to supplement existing database for developing liquefaction damage potential maps.
3. Characteristics of soils in the Yunalin area Fig. 4 shows a geologic map Yuanlin. The eastern part of Yuanlin consists of the Toukoshan Formation and terrace deposits. The Toukoshan Formation is distinguished into two lithofacies that are gradational to each other; one is the conglomerate facies (Houyenshan facies) and the other is the sandstone and shale facies (Hsiangshan facies). The terrace deposits are composed mainly of unconsolidated gravel with sandy or silty lenses, generally poorly stratified (Ho, 1988). West to the terrace deposits is alluvial deposits of clay, silt, sand, and gravels, which cover the plain of Yuanlin. Widespread liquefaction was observed in the plain in the 1999 ChiChi earthquake. Soils and geological conditions in Yuanlin are mainly characterized based on data from SPT borings and CPT soundings. Fig. 5 shows a CPT sounding and
101
a SPT borehole conducted at Yuansui Road, Yuanlin, where serious sand boiling and settlement were observed. The site is quite typical among the liquefied sites in the Yuanlin area. At this site, the thickness of top fill layer is about 0.7 m. Beneath the backfill to the depth of 8.1 m is mainly loose sandy soil with low cone tip resistance. The averages of tip resistance and side friction are 2.57 MPa and 19.62 kPa, respectively. From the depth of 8.1 to 10.3 m is a layer of cohesive soil with an average tip resistance of 0.91 MPa and an average side friction of 16.05 kPa. Beneath this cohesive soil layer to the depth of 17.6 m, there is a medium dense to dense sand with an average tip resistance of 10.96 MPa and an average side friction of 24.83 kPa. From 17.6 to 20 m in depth, there is a clay layer with an average tip resistance of 1.34 MPa and an average side friction of 37.25 kPa, respectively. Fig. 6 shows the particle size distributions of the samples collected from different depths at this site, as well as the sand boil samples collected at the ground surface. The particle size distribution curves show that the characteristics of the extruded soils are very similar to those of the soil at the depth of 4 m. This provided strong evidence that the soil at the depth of 4
Fig. 4. Simplified geologic map of Yuanlin.
102
D.-H. Lee et al. / Engineering Geology 71 (2003) 97–117
Fig. 5. CPT and SPT profile of a typical liquefied site in Yuanlin.
m had liquefied. The particle size distribution curves of the samples tested all fell within the ranges of liquefiable soils specified by the Japanese Port and Harbor Research Institute (JPHRI, 1989). These soils are all silty sands (SM) with fines content (FC) in the range of 15% to 17%.
4. CPT-based simplified methods CPT is generally considered a more consistent and repeatable in situ test than SPT (Lunne et al., 1997), and unlike SPT, it can provide a nearly continuous soil profile. This stratigraphic capability
Fig. 6. Particle size distributions of samples taken from a liquefied site.
D.-H. Lee et al. / Engineering Geology 71 (2003) 97–117
makes the CPT particularly advantageous for developing liquefaction-resistance profiles (Youd et al., 2001). However, each in situ test has its advantages and limitations, and thus CPT testing at a site should be verified with a few well-placed SPT boreholes to confirm soil types and the interpretation of liquefaction resistance. In this paper, three CPT-based methods are applied to evaluate liquefaction potential of soils. They are: the Robertson and Wride (1998) method, the Olsen’s (1997) simplified method, and the Juang et al. (2002, in press) method. For convenience, the three methods are referred to herein as the Robertson method, the Olsen method, and the Juang method. It is noted that all simplified methods that follow the general stressbased approach pioneered by Seed and Idriss (1971) require the determination of two variables, namely, cyclic stress ratio (CSR) and cyclic resistance ratio (CRR). To provide a basis for comparing the three CPT-based methods, use of exactly the same procedure for the determination of CSR is required. The sections that follow first describe the procedure for determining CSR, and then briefly summarize the procedure for determining CRR in each of the three methods. 4.1. Evaluation of cyclic stress ratio In this paper, the equation for CSR originally defined by Seed and Idriss (1971) is adjusted to the benchmark earthquake (moment magnitude Mw = 7.5): CSR7:5
rv amax ðrd Þ=MSF ¼ 0:65 rvV g
ð3Þ
rv = the vertical total stress of the soil at the depth studied, rvV= the vertical effective stress of the soil at the depth studied, amax = the peak horizontal ground surface acceleration, g = the acceleration of gravity, rd = the shear stress reduction factor, and MSF = the magnitude scaling factor. The variable rd is calculated as follows (Liao et al., 1988): rd ¼ 1:0 0:00765z; rd ¼ 1:174 0:0267z;
zV9:15 m 9:15 m < zV23 m
ð4Þ
103
The variable MSF is calculated as follows (Idriss, as cited in Youd et al., 2001): n Mw ð5Þ MSF ¼ 7:5 where Mw is the moment magnitude and n is an exponent. In the present study, n is set to be equal to 2.56. 4.2. CRR by the Olsen method This method is based on the Olsen’s soil classification chart-based method for evaluating liquefaction resistance (Olsen, 1988). In this simplified Olsen (1997) method, CRR is calculated as follows: CRR ¼ 0:00128½qc =ðrvVÞ0:7 0:025 þ 0:17Rf 0:028R2f þ 0:0016R3f
ð6Þ
where qc = cone penetration resistance in atm (1 atm is approximately equal to 100 kPa), rvV= effective stress in atm, and Rf = friction ratio in percent, defined as the sleeve friction fs divided by qc. The calculated CRR must be compared to CSR7.5 determined from Eq. (3), and the factor of safety determined as: Fs ¼ CRR=CSR7:5
ð7Þ
The simplified Olsen method is generally conservative. Note that in this method, CRR is strongly affected by the friction ratio (Rf). For a given normalized tip resistance of 100, the CRR determined from Eq. (6) would be 0.136, 0.181, and 0.247 for sands with a Rf value of 0.2, 0.5, and 1.0, respectively. Whether or not the effect of friction ratio should be so significant deserves further investigation. This issue is further complicated by the fact that in a Cone Penetration Test, the measurement of friction ratio is generally much less precise than that of cone tip resistance. Thus, the accuracy of the Olsen method may be compromised by the less precise measurement of the friction ratio. A typical site in Yuanlin with CPT measurements is used to examine the applicability of the Olsen method. YL-C31 is a CPT sounding conducted in
104
D.-H. Lee et al. / Engineering Geology 71 (2003) 97–117
the liquefied area of Yuanlin. The ground water table is 0.6 m below the ground surface. Fig. 7 shows results of the analysis using the Olsen method. The calculated factors of safety Fs along the depth are almost all less than 1.0. This suggests that the liquefaction potential of this site is extremely high, which appears to agree very well with the field observation. However, the soil in the layer of 13.5 to 18 m deep is a medium dense to dense sand with high cone tip resistance in the range of 8.5 to 15.6 MPa. Examination of the CPT soundings at liquefaction sites after the 1999 Chi-Chi earthquake revealed that qc values in the liquefied layers of loose sands after the quake were generally in the extremely low range of 1.13 to 2.57 MPa, suggesting that the strength of liquefied loose sands had been reduced by the occurrence of liquefaction. Thus, it is reasonable to assume that at deeper strata, the sands that retained relatively high qc values experienced no liquefaction in the earthquake. This view is supported by laboratory results by Kobayashi et al. (2001), which reported that the modulus of subgrade reaction of sands was reduced to about 20% of the initial value when the soil was liquefied through shaking, and such results were obtained regardless of
the initial relative density (either at 50% or 80%). However, the modulus of subgrade reaction of sands under shaking remained about the same if liquefaction did not occur. Based on the measured qc values and the above assertion, no liquefaction is judged to have occurred in the sand layer between the depth of 13.5 and 18 m. Thus, it appears that the Olsen method underestimates the liquefaction resistance of this medium dense to dense sand layer, as the average factor of safety by the Olsen method in this layer is 0.67. 4.3. Robertson method In the Robertson method, CRR is calculated as: CRR ¼ 0:833½qc1N;cs =1000 þ 0:05; if qc1N;cs < 50 CRR ¼ 93½qc1N;cs =10003 þ 0:08; if 50Vqc1N;cs < 160
ð8Þ
where qc1N,cs = clean-sand cone tip resistance normalized to a reference stress level of 1 atm (approx-
Fig. 7. Assessment of liquefaction potential at YL-C31 by the Olsen method.
D.-H. Lee et al. / Engineering Geology 71 (2003) 97–117
105
Fig. 8. Assessment of liquefaction potential at YL-C31 by the Robertson method.
imately, 100 kPa). The normalization process, which converts the measured cone tip resistance ( qc) to the normalized cone tip resistance ( qc1N), and the procedure to determine the clean-sand equivalence ( qc1N,cs) from qc1N are described in detail in Robertson and Wride (1998) and Youd et al. (2001). Again, in this method, CSR7.5 must be calculated from Eq. (3) and the factor of safety can be calculated from Eq. (7). The Robertson method follows more closely the SPT-based simplified method originated by Seed and Idriss (1971) and updated in Youd et al. (2001). Eq. (8) basically represents the boundary curve that encompasses liquefied data points in a plot of CRR versus qc1N,cs. Fig. 8 shows the analysis results of YL-C31 using the Robertson method. The Fs profile indicates that this site has a high liquefaction potential, as almost all of the calculated Fs are less than 1.0. This conclusion agrees well with field observations. Similar to the analysis with the Olsen method, the Robertson method also ‘‘predicts’’ liquefaction in the Chi-Chi event for the soil form 13.5 to 18.0 m, as the average factor of safety for this layer is 0.84.
Thus, the Robertson method also underestimates the liquefaction resistance of medium dense to dense sands. 4.4. Juang method Juang et al. (2000) pioneered a procedure with which the boundary curve, called limit state herein, may be derived in a more objective way. This method relies on artificial neural networks that are trained to extract the input –output relationship from a database of field liquefaction performance. The trained neural network is then used to establish limit state. The reader is referred to Juang et al. (2000, 2001) for details of this process. To facilitate the use of the neural network-generated limit state, Juang et al. (in press) performed least-square regression analyses of the data points generated by the neural network, and the following empirical equation was obtained as a whole: CRR ¼ Crexp½2:957 þ 1:264ðqc1N;cs =100Þ1:25 ð9Þ
106
D.-H. Lee et al. / Engineering Geology 71 (2003) 97–117
where Cr ¼ 0:016ðrvV=100Þ3 þ 0:178ðrvV=100Þ2 0:063ðrvV=100Þ þ 0:903
ð10Þ
With CRR calculated from Eq. (9) and CSR from Eq. (3), the factor of safety can be obtained from Eq. (7). Juang et al. (in press) further developed the following mapping function from which the probability of liquefaction for a given factor of safety (calculated from Eqs. (3), (7) and (9)) may be estimated:
qc1N;cs ¼ qc1N ð2:42914c 16:94313c þ 44:551Ic2
1
ð11Þ
PL ¼
qc1N ¼ 10½qc =ðrvVÞ0:5
ð12Þ
Ic ¼ ½ð3:47 logqc1N Þ2 þ ðlogF þ 1:22Þ2 0:5
ð13Þ
F ¼ fs =ðqc rv Þ 100%
ð14Þ
Fig. 9 shows results of the analysis when the case of YL-C31 is re-analyzed using the Juang method. From the derived profile of liquefaction probability, the soils within the top 11.7 m are liquefiable with a liquefaction probability generally greater than 0.85, which agrees well with field observations. Unlike the other two CPT-based methods described previously, however, the Juang method predicted little liquefaction potential (with an average probability of about 15%) in the layer between the depths of 13.5 and 18 m. This shows that the Juang method is more accurate in evaluating the medium dense to dense sandy soils than the Olsen method and the Robertson method, although all three methods correctly predicted liquefaction potential of loose sands in the Chi-Chi event.
51:497Ic þ 22:802Þ
In the above equations, cone tip resistance ( qc), sleeve friction ( fs), total overburden stress (rv), and effective overburden stress (rV v) must all be in the unit of kPa. The variables F, Ic, qc1N, qc1N,cs, and CRR are all dimensionless. Note that Eqs. (12), (13) and (14) were taken from Robertson and Wride (1998) with slight modification in Eq. (13).
1 þ ðFs =0:96Þ4:5
Fig. 9. Assessment of liquefaction potential at YL-C31 by the Juang method.
ð15Þ
D.-H. Lee et al. / Engineering Geology 71 (2003) 97–117
5. Liquefaction failure potential 5.1. Liquefaction potential index Iwasaki et al. (1982) proposed the Liquefaction Potential Index (IL) to estimate the liquefaction damage or failure potential. Whereas liquefaction potential as defined in Seed and Idriss (1971) and Youd et al. (2001) is evaluated for a soil at a particular depth in the subsurface, the IL index is an integrated effect of the likely liquefaction over the entire depth of the profile (to be more specific, down to 20 m below the ground surface). Whereas this index is not without pitfalls, it is easy to use and particularly useful for mapping liquefaction damage or failure potential of an area. As mentioned previously, however, the index IL needs to be re-calibrated each time a different method is used for evaluating liquefaction resistance and potential because different degrees of conservatism were built into the existing empirical methods. Fig. 10 shows the distribution of the calculated IL values obtained by Iwasaki et al. (1982) based on the analysis of their database. More than 75% of non-liquefied cases had an IL value of less than 5, and almost all non-liquefied cases had an IL value of less than 15. Among the liquefied cases, 50% of them had an IL value of greater than 15. However, 20% of the liquefied cases were found to have an IL value of less than 5. Based on these results, Iwasaki et al. (1982) proposed four classes of liquefaction failure potential,
107
as described previously. The threshold IL value is 5 for high liquefaction failure potential, and is 15 for extremely high potential. In the present study, the Olsen method is first used to analyze the 72 CPT soundings each with observation of field performance (whether liquefactioninduced ground failure was observed). For each CPT profile, the factor of safety is calculated at each depth continuously over the depth of 20 m, as suggested by Iwasaki et al. (1982). In other words, a profile of factor of safety is obtained. This profile is then used to calculate the Liquefaction Potential Index IL according to Eq. (1). The distribution of the calculated IL values of the 72 CPTs is shown in Fig. 11. The results show 75% of liquefied cases with IL greater than 5, and 50% of liquefied cases with IL greater than 15. The results seem to agree well with the Iwasaki criteria. However, only 25% of nonliquefied cases have an IL value of less than 5. Recall that in database used by Iwasaki et al. (1982), 70% of non-liquefied cases were found to have an IL value of less than 5. This suggests that the Iwasaki criteria may not be applicable when the liquefaction resistance and factor of safety are calculated using the Olsen method. When applying the Olsen method without re-calibration, many of the non-liquefied sites would have been predicted to liquefy in the Chi-Chi earthquake according to the Iwasaki criteria. Another way to examine the Olsen method is to construct a contour map of liquefaction failure poten-
Fig. 10. Distribution of calculated IL values of liquefied group and non-liquefied group by Iwasaki et al. (1982).
108
D.-H. Lee et al. / Engineering Geology 71 (2003) 97–117
Fig. 11. Distribution of calculated IL values of liquefied group and non-liquefied group of the 72 CPTs analyzed by the Olsen method.
tial based on the calculated IL values. Fig. 12 shows a contour map of the calculated IL values in Yuanlin. Note that in this figure, the triangle symbol indicates the location of the CPT, and the shaded area with grids at the right belongs to the hill of the Central Mountain Range that consists mainly of gravelly cobble deposits
in which liquefaction is not an issue. The contour lines of IL = 5 and 15, the two bounds adopted by Iwasaki et al. (1982) in his criteria for liquefaction-induced failure potential, are drawn. The observed liquefaction sites in the Chi-Chi earthquake are marked with dark shades. Although many severe liquefaction sites are
Fig. 12. CPT locations and contour of IL values at Yuanlin by the Olsen method. The IL values of 5 and 15 are the lower and upper boundary values suggested by Iwasaki et al. (1982).
D.-H. Lee et al. / Engineering Geology 71 (2003) 97–117
109
Fig. 13. Distribution of calculated IL values of liquefied group and non-liquefied group of the 72 CPTs analyzed by the Robertson method.
located in the zone with IL>15, other areas without liquefaction sites also show IL>15. In addition, almost the entire town has an IL>5, indicating high liquefaction failure potential, which cannot explain many no-liquefaction sites. This again suggests that the Olsen method is too conservative when it is
used in conjunction with the Iwasaki criteria for assessing the liquefaction failure potential. On the other hand, it points to the need of establishing new criteria for assessing liquefaction failure potential in conjunction with the use of the Olsen method.
Fig. 14. CPT locations and contour of IL values at Yuanlin by the Robertson method. The IL values of 15 are the upper boundary values suggested by Iwasaki et al. (1982).
110
D.-H. Lee et al. / Engineering Geology 71 (2003) 97–117
Fig. 15. Distribution of calculated IL values of liquefied group and non-liquefied group of the 72 CPTs analyzed by the Juang method (Ic < 2.95).
The same 72 cases are analyzed using the Robertson method, and the index IL is calculated for each CPT profile. Fig. 13 shows the distribution of the calculated IL values. As shown in Fig. 12, only 10% of liquefied cases have an IL value of less than 5, whereas there are 50% of liquefied cases with IL
greater than 15. For liquefied cases, the results of the analysis using the Robertson method agree well with field observations. However, the results also show that the calculated IL values are generally too high. For example, 85% of non-liquefied cases have an IL value of greater than 5, and 30% of non-
Fig. 16. CPT locations and contour of IL values at Yuanlin by the Juang method (Ic < 2.95). The IL values of 15 are the upper boundary values suggested by Iwasaki et al. (1982).
D.-H. Lee et al. / Engineering Geology 71 (2003) 97–117
111
Fig. 17. Distribution of calculated IL values of liquefied group and non-liquefied group of the 72 CPTs analyzed by the Juang method (Ic < 2.4).
liquefied cases have an IL value of greater than 15. This indicates that the Iwasaki criteria as described previously are not applicable when the liquefaction resistance is calculated using the Robertson method. Fig. 14 shows a contour map of the calculated IL values in Yuanlin where the liquefaction resistance is calculated using the Robertson method. As in the analysis using the Olsen method, this contour map shows that the Robertson method is too conservative
when it is used in conjunction with the Iwasaki criteria for assessing the liquefaction failure potential. The need to establish new criteria for assessing liquefaction failure potential in conjunction with the use of the Robertson method is obvious. Fig. 15 shows the distribution of I L values obtained by using the Juang method. The results indicate that there is no clear-cut boundary between IL values of liquefied cases and those of non-liquefied
Table 1 Liquefaction potential index (IL) and liquefaction risk index (IR) at various sites CPT
NT-1 NT-C2 NT-C7 NT-C15 YL-2 YL-C22 YL-C24 YL-C25 YL-C31 YL-C43 YL-C7 YL-C8 YL-C9 YL-C10 YL-C11 YL-C13 YL-C15 YL-C16 YL-C28 YL-C36
IL (Juang)
IR (Juang)
Ic < 2.95
Ic < 2.4
Ic < 2.4
35 43 52 53 27 20 25 23 28 16 16 16 14 15 20 16 14 25 18 12
17 27 28 23 21 17 17 11 17 8 11 1 4 9 5 5 5 6 7 1
29 40 43 34 50 42 40 30 38 23 32 5 15 21 10 15 18 14 17 2
IL (Olsen)
IL (Robertson)
Liquefaction?
22 34 45 40 25 20 16 9 34 9 16 4 7 7 11 6 4 16 8 8
21 33 36 27 25 21 23 19 27 14 17 6 10 13 10 13 12 12 13 7
Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes No No No No No No No No No No
112
D.-H. Lee et al. / Engineering Geology 71 (2003) 97–117
cases. Fig. 16 shows the contour map of IL values in Yuanlin where the liquefaction resistance is calculated using the Juang method. As shown in Fig. 16, almost all IL values in the entire town of Yuanlin are above 15. This suggests that use of the IL values calculated with the Juang method to judge whether liquefactioninduced ground failure will occur is unlikely to succeed. An examination of the Juang method reveals that the upper bound of the soil classification index Ic for liquefied soil is 2.95 in this method. This may lead to an overestimate of liquefaction potential of silty soils and thus the liquefaction failure potential index IL. Ku (2001) studied the characteristics of the index Ic for the soils in the middle west of Taiwan, which covers all liquefaction sites in the Chi-Chi earthquake, and concluded that the upper bound for typical liquefied soils (SP, SP – SM, SM) in the Yuanlin area should be about 2.4. By placing an upper bound of Ic = 2.4 in the Juang method, and recalculating the IL values for all CPTs, a new distribution of IL is obtained, as shown in Fig. 17. In this figure, the difference of IL values between
liquefied and non-liquefied cases is obvious. About 50% of non-liquefied cases have an IL value of less than 5, and all non-liquefied cases are found to have an IL value of less than 15. For liquefied cases, only about 20% have an IL value of less than 5. The modified Juang method is shown to be compatible with the Iwasaki criteria for liquefaction failure potential evaluation. As mentioned previously, it is necessary to recalibrate the Iwasaki criteria if a new method for evaluating liquefaction resistance and potential is to be used. To this end, 20 CPTs that were identified with clear-cut liquefaction failure or no failure are examined in detail. At the locations of 10 of these CPTs, liquefaction induced failure were observed, and at the other 10, no failure was observed. Table 1 shows the calculated IL values in conjunction with the use of each of the three CPT-based methods for liquefaction resistance evaluation. With respect to the Olsen method, among the 10 liquefied cases, 80% were found to have IL>15, which is reasonable; and among the 10 non-liquefied cases, 20% were found to have IL < 5, which is not reasonable. Thus,
Fig. 18. Liquefaction potential map of Yuanlin by the Olsen method. The IL values of 8 and 16 are the lower and upper boundary values that provide the best delineation of liquefaction failure potential classes.
D.-H. Lee et al. / Engineering Geology 71 (2003) 97–117
in order to use the Olsen method, the two classification bounds, 5 and 15, in the Iwasaki criteria need to be adjusted. It is postulated that the higher bound should be taken as the one that encompasses (with IL equal or greater than) 70% of liquefied cases, and the lower bound should be taken as the one that encompasses (with IL equal or less than) 70% of nonliquefied cases. Under this postulation, the boundaries for high failure potential and low failure potential are 16 and 8, respectively. In other words, the ground failure potential is low if IL < 8, and the ground failure potential is extremely high if IL>16. The ground failure potential is high if 8 < IL V 16. Using this set of criteria, the contour map of IL values calculated with the Olsen method is redrawn, as shown in Fig. 18. From this figure, it can be seen that most liquefied sites are within the zone of IL>16 and there is little liquefied cases within the zone of IL < 8. The recalibrated IL index improves the accuracy of the Olsen method for such purpose. Similarly, the Iwasaki criteria are not applicable when the Robertson method is used for liquefaction resistance evaluation because all of the non-liquefied
113
cases have an IL value of greater than 5. Using the same postulation regarding the higher and lower bounds of IL, as in the case of the Olsen method, the two bounds are determined to 21 and 13, respectively. Thus, the ground failure potential is low if IL < 13, and the ground failure potential is extremely high if IL>21. The ground failure potential is high if 13 < IL V 21. With this new set of criteria, the contour map of IL values is redrawn, as shown in Fig. 19, and the results show that most liquefied sites are within the zone of IL>21 and there is little liquefied cases within the zone of IL < 13. The re-calibrated IL index improves the accuracy of the Robertson method for such purpose. With the new set of criteria, the Robertson method is able to predict fairly accurately the liquefaction-induced ground failure potential. The Iwasaki criteria are not applicable either when the Juang method is used for liquefaction resistance evaluation because there are only three non-liquefied cases with IL < 15. However, with the limit of Ic < 2.4, the modified Juang method is quite accurate without adjustment to the Iwasaki criteria. About 80% of
Fig. 19. Liquefaction potential map of Yuanlin by the Robertson method. The IL values of 13 and 21 are the lower and upper boundary values that provide the best delineation of liquefaction failure potential classes.
114
D.-H. Lee et al. / Engineering Geology 71 (2003) 97–117
liquefied cases have IL values greater than 15, and 60% of non-liquefied cases have IL values of less than 5. Fig. 20 shows the contour of IL values that are obtained with the modified Juang method. As may be seen from this figure, almost all liquefied sites are within the zone of IL>15, and non-liquefied sites are located in the zone of IL < 5. This suggests that the modified Juang method is able to predict quite accurately the ground failure potential without adjustment to the Iwasaki criteria. 5.2. Liquefaction risk index The liquefaction potential index IL is defined with the profile of the factor of safety, and only those with Fs < 1 contribute to the index IL. As noted previously, the factor of safety from different methods essentially has different meaning since each method was developed with different degree of conservatism. In addition, the liquefaction potential is not linearly proportional to the factor of safety; rather, it is linearly proportional to the probability of liquefaction. Thus,
a new index, called Liquefaction Risk Index (IR) is defined using the probability of liquefaction: IR ¼
20 X
PL wðzÞdz
ð16Þ
0
where PL is the probability of liquefaction obtained from Eq. (15); W(z) is the weight function as defined in Eq. (2). Eq. (16) is seen as a variable form of Eq. (1). Re-analysis of the 72 CPTs with the modified Juang method (Ic = 2.4 as the upper bound of liquefied cases) coupling with the definition of IR in Eq. (16) yields distributions of IR for liquefied cases and for non-liquefied cases, as shown in Fig. 21. The distributions are deemed more reasonable than any of the previously obtained distributions by other methods. Following the same postulation in which 70 percentile in the IR distribution is used to determine lower and higher bounds for classification, the two bounds are determined to be about 20 and 30, respectively. Thus, the ground failure potential is
Fig. 20. Liquefaction potential map of Yuanlin by the Juang method (Ic < 2.4). The IL values of 5 and 15 are the lower and upper boundary values that provide the best delineation of liquefaction failure potential classes, which are the same as in the Iwasaki criteria.
D.-H. Lee et al. / Engineering Geology 71 (2003) 97–117
115
Fig. 21. Distribution of calculated IR values of liquefied group and non-liquefied group of the 72 CPTs analyzed by the Juang method (Ic < 2.4).
low if IR < 20, and the ground failure potential is extremely high if IR>30. The ground failure potential is high if 20 < IR V 30. Fig. 22 shows a contour map of the calculated IR values. Almost all of the liquefied sites locate in the zone with IR>30, and almost all of the non-liquefied sites are in the zone of IR < 20. The contour map shows that the formulation
of IR and the criteria for assessing liquefactioninduced ground failure are reasonable. In summary, three CPT-based methods have been examined for their applicability of being integrated into the liquefaction potential index (IL) defined by Iwasaki et al. (1982). The study concludes that recalibration of the calculated failure potential index IL
Fig. 22. Contour map of liquefaction risk index (IR) at Yuanlin by the Juang method.
116
D.-H. Lee et al. / Engineering Geology 71 (2003) 97–117
is needed, regardless of which CPT-based method is used for determining liquefaction resistance. The modified Juang method is further used to define the Liquefaction Risk Index (IR), which provides the best overall results in assessing liquefaction-induced failure potential. The contour maps of IL (or IR) along with field observations provide an effective means to validate the methodology and procedures for assessing liquefaction-induced failure potential of a site presented in this paper. As a final note, the accuracy of the contour maps may be improved by additional well-placed CPTs in the Yuanlin area, although the indexes IL and IR and the associated criteria established in this paper will remain the same, since no field observations will be available with these CPTs.
6. Conclusions Based on the analysis of limited CPT data presented in this paper, the following conclusions are reached: (1) All three CPT-based methods examined in this paper can predict accurately liquefied cases in the 1999 Chi-Chi earthquake. The Juang method, however, is shown to be more accurate in evaluating liquefaction resistance of the medium dense to dense sand with high cone resistance, the soil that is generally considered non-liquefiable. (2) When the CPT-based methods are integrated into the framework of the Liquefaction Potential Index (IL) defined by Iwasaki et al. (1982), the criteria for assessing the liquefaction failure potential need to be re-calibrated. Once re-calibrated, each method, be it the Robertson method, the Olsen method, or the Juang method, may be used to assess the liquefaction failure potential. For the Olsen method, the ground failure potential is low if IL < 8, and the ground failure potential is extremely high if IL>16. For the Robertson method, the ground failure potential is low if IL < 13, and the ground failure potential is extremely high if IL>21. However, for the Juang method with liquefaction upper bound of Ic = 2.4, the established criteria are consistent with the Iwasaki criteria (low if IL < 5; extremely high if IL>15).
(3) The Liquefaction Risk Index (IR) developed in this paper shows an improvement over the Liquefaction Potential Index (IL) for predicting the liquefaction-induced failure potential. The term IR is defined with the modified Juang method, and the following criteria are established for assessing the liquefactioninduced failure potential: (1) the failure potential is high if IR z 30, (2) the failure potential is medium if 20 z IR>30, and (3) the failure potential is low if IR < 20. The contour map of IR values prepared for the town of Yuanlin can identify accurately the liquefaction and no-liquefaction sites in the Chi-Chi earthquake.
Acknowledgements The study on which this paper is based was sponsored by National Science Council, Taipei, Taiwan through Grant No. NSC 90-2611-E-006-029. This financial support is greatly appreciated. The authors are indebted to Dr. C. Hsein Juang of Clemson University who first communicated to the authors the need of re-calibration of the Iwasaki criteria, and provided many helpful suggestions during the course of this study. Dr. John Christian, a peer-reviewer, and another anonymous reviewer are thanked for their constructive comments that help sharpen this paper.
References CGSB, 1999. Investigation Report of the 1999, Chi-Chi Earthquake. Central Geological Survey Bureau (CGSB), Republic of China, Taipei, Taiwan. In Chinese. Chen, C.J., Juang, C.H., 2000. Calibration of SPT- and CPTbased liquefaction evaluation methods. In: Mayne, P., Hryciw, R. (Eds.), Innovations Applications in Geotechnical Site Characterization. Geotechnical Special Publication, vol. 97. ASCE, Reston, VA, pp. 49 – 64. Ho, C.S., 1988. An introduction to the geology of Taiwan and explanatory text of the geologic map of Taiwan. Central Geological Survey. The Ministry of Economic Affairs, Taipei, Taiwan. Ishihara, K., 1985. Stability of natural deposits during earthquake. Proceedings of the Eleventh International Conference on Soil Mechanics and Foundation Engineering, San Francisco, California, vol. 1, pp. 321 – 376. Ishihara, K., 1993. Liquefaction and flow failure during earthquakes. Ge´otechnique, The Institution of Civil Engineers 43 (3), 351 – 415. Iwasaki, T., Arakawa, T., Tokida, K., 1982. Simplified procedures
D.-H. Lee et al. / Engineering Geology 71 (2003) 97–117 for assessing soil liquefaction during earthquakes. Proceedings of the Conference on Soil Dynamics and Earthquake Engineering, Southampton, UK, pp. 925 – 939. JPHRI, 1989. Port and Harbor Facilities Design Guide Book. Japan Port and Harbor Research Institute, Tokyo. In Japanese. JSHE, 1990. Highway Bridge Design Guide Book. Japan Society of Highway Engineering, Tokyo. In Japanese. Juang, C.H., Chen, C.J., Tang, W.H., Rosowsky, D.V., 2000. CPTbased liquefaction analysis: Part I. Determination of limit state function. Ge´otechnique, The Institution of Civil Engineers 50 (5), 583 – 592. Juang, C.H., Chen, C.J., Jiang, T., 2001. Probabilistic framework for liquefaction potential by shear wave velocity. Journal of Geotechnical and Geoenvironmental Engineering, ASCE 127 (8), 670 – 678. Juang, C.H., Yuan, H., Lee, D.H., Ku, C.S., 2002. Assessing CPTbased methods for liquefaction evaluation with emphasis on the cases from the Chi-Chi, Taiwan, earthquake. International Journal of Soil Dynamics and Earthquake Engineering 22 (3), 241 – 258. Juang, C.H., Yuan, H., Lee, D.H., Lin, P.S., 2003. A simplified CPT-based method for evaluating liquefaction potential of soils. Journal of Geotechnical and Geoenvironmental Engineering 129 (1), 66 – 80. Kobayashi, H., Morimitsu, T., Azuma, T., Tamura, K., 2001. Experimental study on reduction of the horizontal subgrade reaction due to liquefaction. Journal of Japan Civil Engineering 43 (2), 46 – 51. In Japanese. Ku, C.S., 2001. ‘‘A study on soil classification and liquefaction evaluation based on CPT.’’ PhD Dissertation. National Cheng Kung University, Tainan, Taiwan. In Chinese. Lee, D.H., Ku, C.S., 2002. Application of CPT in the Liquefaction Investigation in the central Taiwan. Proceeding, Conference on Liquefaction Potential Investigation in Central Taiwan, Tainan, Taiwan. In Chinese. Liao, S.C.C., Veneziano, D., Whitman, R.V., 1988. Regression models for evaluating liquefaction probability. Journal of Geotechnical Engineering, ASCE 114 (4), 389 – 411. Lin, M.L., Liao, H.J., Ueng, T.S., 1999. A study on the geotechnical-
117
related damages in the Chi-Chi Earthquake. Civil and Hydraulic Engineering 26 (3), 60 – 71 (in Chinese). Lunne, T., Robertson, P.K., Powell, J.J.M., 1997. Cone Penetration Testing. Blackie Academic and Professional, London, UK. Olsen, R.S., 1988. Using the CPT for dynamic site response characterization. In: Von Thum, J.L. (Ed.), Proceedings, Earthquake Engineering and Soil Dynamics II Conference. Geotechnical Special Publication, vol. 2. ASCE, New York, pp. 374 – 388. Olsen, R.S., 1997. Cyclic liquefaction based on the cone penetration test. In: Youd, T.L., Idriss, I.M. (Eds.), Proceedings of the NCEER Workshop of Evaluation of Liquefaction Resistance of Soils, Technical report NCEER-97-0022. National Center for Earthquake Engineering Research, State University of New York at Buffalo, Buffalo, NY, pp. 225 – 276. Robertson, P.K., Campanella, R.G., 1985. Liquefaction potential of sands using the CPT. Journal of Geotechnical Engineering, ASCE 111 (3), 298 – 307. Robertson, P.K., Wride, C.E., 1998. Evaluating cyclic liquefaction potential using the cone penetration test. Canadian Geotechnical Journal 35 (3), 442 – 459. Seed, H.B., Idriss, I.M., 1971. Simplified procedure for evaluating soil liquefaction potential. Journal of the Soil Mechanics and Foundation Division, ASCE 97 (9), 1249 – 1273. Shibata, T., Teparaksa, W., 1988. Evaluation of liquefaction potential of soils using cone penetration tests. Soils and Foundations 28 (2), 49 – 60. Stark, T.D., Olson, S.M., 1995. Liquefaction resistance using CPT and field case histories. Journal of Geotechnical Engineering, ASCE 121 (GT 12), 856 – 869. Youd, T.L., Idriss, I.M., Andrus, R.D., Arango, I., Castro, G., Christian, J.T., Dobry, R., Liam Finn, W.D., Harder Jr., L.F., Hynes, M.E., Ishihara, K., Koester, J.P., Laio, S.S.C., Marcuson III, W.F., Martin, G.R., Mitchell, J.K., Moriwaki, Y., Power, M.S., Robertson, P.K., Seed, R.B., Stokoe II, K.H., 2001. Liquefaction resistance of soils: summary report from the 1996 NCEER and 1998 NCEER/NSF workshops on evaluation of liquefaction resistance of soils. Journal of Geotechnical and Geoenvironmental Engineering 127 (10), 817 – 833.