1 Introduction

In modern seismic design codes, the elastic response spectrum of earthquake ground motion is commonly adopted to determine the seismic design forces of buildings and other structures. However, the shapes of earthquake response spectrum are significantly influenced by site geology and local soil conditions. Thus, the pioneer work of Seed et al. (1976) proposed the first site-dependent spectra for earthquake resistant design. In their work, the seismic sites were classified into four kinds from rock, deep stiff soil to soft soil using qualitative descriptions. In the 1991 version of the National Earthquake Hazards Reduction Program (NEHRP) Provisions, the site classification was basically based on a semi-quantitative description similar to that of Seed et al. (1976). Thus, some engineering judgments are required for site classification (BSSC 1985, 1988, 1991). However, based on the research results of Dobry et al. (2000), the recent seismic design code of the FEMA (e.g., the 2015 version of the NEHRP Provisions) used a representative average shear wave velocity of the top 30 m of the subsurface profile (Vs30) to quantitatively consider the effect of local soil conditions on ground motion. According to the site’s velocity profile, the site can be classified into its own site class uniquely with its Vs30. Table 1 shows current site classification criteria with respect to Vs30 by BSSC (2003) and MOI (2011). Since Vs30 can provide unambiguous site classification and is also measurable by geophysical methods, thus the Vs30 approach becomes popular and is widely used in practice. In this approach, the site classification is determined first by its Vs30. Then, the site coefficients (Fa and Fv) are obtained according to its class. The design response spectrum is then derived by multiplying the reference spectrum with Fa and Fv at its short and long period parts as shown in Fig. 1. By this way, the effect of local soil conditions on the design seismic force of a structure is taken into account.

Table 1 Summary of the site categories by Vs30
Full size table
Fig. 1
figure 1

Illustration of spectral response acceleration subject to reference stratum and soils (modified from Dobry et al. 2000)

Full size image

Even though the definition of Vs30 is clear, there is still doubt whether different seismic methods yield the same results since the new site classification was introduced in 1994. Furthermore, whether the Vs30 value of a site may vary much due to the different Vs measurement methods, which could result in a different site class and design seismic force. It is the issue that this study wants to address.

Regarding the researches on the shear wave velocity of alluvial soils, the key influencing parameters such as geological age, void ratio, plasticity index, soil type, effective stress state, over-consolidation ratio, etc. have been studied and clarified by several early studies, and Sykora (1987) provided a comprehensive summary. Anbazhagan et al. (2016) studied the correlation of densities with Vs and SPT-N. Garofalo et al. (2016) reported the execution of the InterPACIFIC project and assessed the results of several Vs measurements at three different subsoil conditions. L’Heureux and Long (2017) used the data measured by several geophysical seismic methods and SCPT-u to compare the Vs differences due to measurement method and establish the Vs-qc correlation of Norwegian clays. Due to the demand of practical use, numerous correlation studies have been conducted to directly examine a relationship between Vs and SPT-N (Imai and Yoshimura 1975; Sykora and Stokoe 1983; Sykora and Koester 1988; Jafari et al. 2002; Anbazhagan and Sitharam 2006; Hasancebi and Ulusay 2007; Hanumantharao and Ramana 2008; Ulugergerli and Uyanık 2007; Unal 2009; Maheswari et al. 2010; Kumar et al. 2010; Akin et al. 2011; Tsiambaos and Sabatakakis 2011; Chaterjee and Chaudhury 2013; Marto et al. 2013, Jhinkwan and Jain 2016; Kirar et al. 2016). Wair et al. (2012) provided a summary of the above researches on the Vs estimation based on SPT data and gave the recommendation in estimating the Vs of the top 30 m of the soil profile (Vs30). Of course, it is better to conduct the in situ Vs measurement directly for each project, and evaluate the site classification according to the results of Vs measurement. However, in most of practical cases, only SPT borehole would be conducted and the Vs measurement is always ignored due to the limit of budget. Thus the use of some empirical Vs-N relationships for the evaluation of Vs becomes very popular in engineering practices. However, it should be noticed that the databases for the Vs-N regression analysis are generally collected from different Vs measurement methods. Although it is well known and proved by several studies that cross-hole testing is the most reliable Vs acquisition method. However, few researches used the databases based on cross-hole testing while conducting the Vs-N regression analysis. This is because cross-hole testing needs to bore two holes and the cost is expensive. Thus its data base is relatively insufficient.

There are two purposes in this study. The first one is to clarify the difference of the measured Vs due to the different geophysical seismic methods in Holocene alluviums, and provides some comments on the discrepancy accordingly. The other one is to suggest a set of Vs-N regression curves based on the reliable databases collected from the several Vs measurement methods including the cross-hole seismic testing, SCPT, and suspension logging for practical use. The suggestion of Vs-N regression curve based on the cross-hole seismic method is believed to be useful and important for practical engineering use. This study first collected many high quality site investigation reports of important projects in Taiwan, and established a database including information on the soil types, the SPT-N-values, and the measured shear wave velocities (Vs) by different seismic measurement methods, such as suspension logging, SCPT, and cross-hole logging. The correlations between Vs and SPT-N were then built for different soil types and compared with each other. Based on the results, the measured Vs from suspension logging yielded significantly high shear wave velocity in comparison with the ones from SCPT and cross-hole logging. The reasons were discussed in the aspects of borehole treatment. Finally, the effect of the different Vs measurement methods on the site classification was then demonstrated with several examples.

2 The seismic design code for building structures in Taiwan

The seismic design code for buildings and other structures in Taiwan essentially follows the recent FEMA system (Construction and Planning Agency, MOI 2011), with some modifications according to the local conditions in Taiwan. For all surface structures, sites are classified as shown in Table 1. Note that the reference strata are different between the systems of FEMA and MOI. The reference strata defined by FEMA is the one with Vs30 > 760 m/s; however, the reference strata defined by MOI only has the Vs30 > 270 m/s. The site classification system of Taiwan originally followed the FEMA system. However, in Taiwan, the free-field seismic stations with Vs30 > 760 m/s is rare. Therefore, while conducting the probabilistic seismic hazard analysis to determine the seismic spectrum of reference stratum in a return period of 475 years, the seismic stations at the sites with Vs30 ranging from 270 to 760 m/s have to be included due to insufficient seismic records. This makes the site classification system of Taiwan has only 3 classes. The stratum with Vs30 > 270 m/s belongs to Site Class I; the one with Vs30 ranging from 180 to 270 m/s is Site Class II, and the one with Vs30 < 180 m/s belongs to Site Class III.

For site classification, Vs30 should be calculated as 30 m divided by the sum of the travel times for shear waves to travel through each layer. As shown in Eq. (1), the travel time for each layer is calculated as the layer thickness (d) divided by its Vs. For cases where measured Vs data is not available, the empirical relationship between SPT-N and Vs in sandy or fine-grained soils suggested by JRA (2002) would be commonly used to estimate Vs in terms of the SPT resistance.

$${\text{Vs}}30 = 30/\varSigma \left( {\text{d/Vs}} \right)$$
(1)

After that, the design spectral response acceleration parameters at the periods of 0.3 s (SDS) and 1 s (SD1) with adjustments for site effects can be calculated as follows:

$${\text{S}}_{{\rm DS}} = {\text{F}}_{{\rm a}} {\text{S}}_{{\rm s}}$$
(2)
$${\text{S}}_{{\rm D}1} = {\text{F}}_{{\rm v}} {\text{S}}_{ 1}$$
(3)

where Ss is the referenced stratum (Site Class I) spectral response acceleration at a short period, S1 is the referenced stratum (Site Class I) spectral response acceleration at a period of 1 s, and Fa and Fv are the site coefficients to consider site effect. A designer should refer to MOI (2011) for the Ss and S1 of each administrative division in Taiwan.

The framework of developing a local site spectrum is summarized in Fig. 1. In the figure, Ss, S1, Fa, and Fv can be derived from the MOI (2011), and Aa is the effective peak ground acceleration which is about Ss/2.5.

3 The database based on different Vs measurement methods

In practice, it is very common to adopt in situ geophysical seismic methods for estimating the shear wave velocities of geo-materials under a small strain level, and this has been incorporated into the site classification systems. In general, the geophysical methods can be divided into two categories: invasive and non-invasive. The main difference between them is the requirement of a drilling hole. Common invasive methods include downhole (DH) and cross-hole (CH) loggings, suspension logging (PS), and the seismic cone penetration test (SCPT). Non-invasive geophysical methods include spectral analysis of surface waves (SASW), seismic refraction, and seismic reflection. In this paper, only the cross-hole logging, suspension logging, and seismic cone penetration test (SCPT) Vs measurement methods are discussed, and the corresponding field measured results were collected from some important engineering projects in Taiwan; Fig. 2 shows schematic diagrams of these methods. Table 2 is a modification of that from Andrus et al. (2004) to compare and summarize the features of these methods.

Fig. 2
figure 2

Field arrangements used to perform intrusive seismic tests

Full size image
Table 2 Comparison of the concerned in situ Vs measurement methods (modified from Andrus et al. 2004)
Full size table

3.1 Sources used for the database

The database used in this study consists of field measured results based on the above Vs measurement methods collected from several important engineering projects located in the alluvium of western Taiwan over recent decades. The locations of these engineering sites are shown in Fig. 3. The basic information of the sites, as well as the references, is tabulated in Table 3. The information for each dataset consists of the depth, soil classification of the unified soil classification system (USCS), SPT-N value and the type of hammers, the measured shear wave velocity and the corresponding Vs measurement method, and the references. Note that while conducting SCPT sounding, because the SPT with the accompanying sampling process by split tube sampler was not used, neither the SPT-N nor the soil classification of the USCS (due to the lack of a soil sample for laboratory testing) were available. This required information is assumed to be the same as measurements from neighboring boreholes at the same depth for further regression work. In this study, the distance between the SCPT sounding and the reference SPT borehole were within about 20 m, and most of them are less than 10 m. In addition, the operation of the cross-hole seismic testing requires two boreholes, as shown in Fig. 2. In the collected cross-hole seismic testing cases, the standard penetration tests were carried out in all Vs-measuring boreholes, which means that a specific depth would have the results of one Vs measurement and two SPT-N values. Since each measured Vs at a specific depth would correspond to two SPT-N values from the two testing boreholes, in order to keep the origin of the raw data, the Vs measured by cross-hole seismic testing would be recorded in two datasets. Based on the above approach, the total number of documented datasets is 1189 (shown in Fig. 4). Among them, 317 datasets are from the recorded data using cross-hole seismic testing, 587 datasets are from the recorded data using SCPT, and the remaining 285 datasets are from the recorded data using the suspension logging testing. These datasets were screened based on selected criteria, and this is explained in the following section. The categorized datasets would then be used for the following regression work.

Fig. 3
figure 3

The locations of the engineering sites

Full size image
Table 3 Basic information for the sites
Full size table
Fig. 4
figure 4

The database established by this study

Full size image

3.2 Data processing

It is recognized that geologic considerations can aid in the estimation of Vs. The datasets shown in Fig. 4 were first divided by soil classification based on the USCS. In this study, the geologic units were divided into two groups, sandy soil and fine-grained soil. Note that the N60-values of the datasets included in the sandy soil group were then screened out if the N60-values were larger than 50. For the fine-grained soil group, the N60-values of the datasets were screened out if the N60-values were larger than 25. The consideration of this screening criterion is because the mechanical behaviors of “the sandy soils” with N60 > 50 and “the fine-grained soils” with N60 > 25 are similar to soft rock instead of soil based on the local experiences of Taiwan and Japan. Since the study focused on the N60-Vs regressions of the alluvium soils, the datasets with N60 higher than the above threshold values are therefore deleted. The ones with SC, SM, SP, or SW of the USCS belong to the sandy soil group. In the sandy soil group, most of the soils do not have plasticity index (PI) except clayey sand (SC). The N60 of the soils ranges from 1 to 49 and the average value is about 18. According to Terzaghi and Peck (1967), the relative density of the soils ranges from about 5 ~ 85% and the degree of compactness is from very loose to dense. The other ones with CL, CH, ML, or CL-ML of the USCS belong to the fine-grained soil group. In the fine-grained soil group, the plasticity index (PI) ranges from 0 to about 47. The N60 of the soils ranges from 1 to 25 and the average value is about 10. According to Terzaghi and Peck (1967), the consistency of the soils ranges from very soft to very stiff. The N60-value represents the blow count number during performing standard penetration test. The subscript of N60 denotes the amount of energy transmitted to the sampler divided by the theoretical maximum SPT energy (64.5 kg dropped from a height of 76 cm) in percentage. In an attempt to minimize variability, the N-values were converted to a uniform reference energy ratio of 60% of the theoretical SPT energy (N60). Due to the lack of energy measurement data for the different types of SPT hammers in the collected database, this study assumed a hammer energy ratio of 60% for rope-cathead hammer system, which is the most common type in Taiwan, and 72% for an automatic hammer system based on Japanese engineering practice. In order to facilitate the following correlations and comparisons, all SPT-N values were converted to the basis of energy ratio of 60%. The SPT-N value with the energy ratio other than 60% should be adjusted. For example, the SPT-N value done by automatic hammer system with the energy ratio of 72% should be multiplied by 1.2(72/60). After the data processing based on the above criteria, there were 6 categories (listed in Table 4). Note that some data were screened out, and the total number of datasets in the table is not the same as the original database. Data points of Vs versus N60 with respect to the different Vs measurement methods for the sandy soil and fine-grained soil groups (the 6 categorized datasets) as well as regression curves are shown in Figs. 5 and 6, respectively. According to the visual inspection on the figures, it can be seen that the cross-hole seismic testing datasets have a relatively small variation at any specific N-value when compared with the other two Vs measurement methods. The SCPT datasets typically had a significant amount of scatter in the measurements (as evident in Figs. 5b and 6b), which may result from the locations of the measured Vs differing from the locations of the SPT. In Figs. 5c and 6c, the Vs measured by suspension logging testing also has a small amount of scatter, and the measured Vs is similar to the results from cross-hole seismic testing when the N60-values are higher than a threshold value (approximately N60 = 20 for sandy soil and N60 = 15 for fine-grained soil). However, on the average, the measured Vs by suspension logging testing seems to be slightly larger than the measurements by cross-hole seismic testing when the SPT-N values are small. In view of the coefficient of determination (R2), the cross-hole seismic testing in this study has the highest R2; the suspension logging testing ranks the second highest R2; the data pairs by SCPT and the referred N-value are the most scattered and has the lowest R2, which is probably due to the spatially varied soil conditions between the SCPT and SPT.

Table 4 Details of the datasets from this study
Full size table
Fig. 5
figure 5

Vs versus N60 and correlation equations for the sandy soil group

Full size image
Fig. 6
figure 6

Vs versus N60 and correlation equations for the fine-grained soil group

Full size image

3.3 A comparison with the Japanese database

In Japan, the measurement of shear wave velocity and the studies for establishing the relationship between the N-value and Vs have been ongoing since the 1960s (Kanai 1966; Shibata 1970; Ohba and Toriuma 1970; Ohsaki and Iwasaki 1973; JSSMFE 1981; Imai and Tonouchi 1982). Among these studies, the database provided by Mr. Tsuneo Imai and the associated regression results recorded by JSSMFE (1981) have been commonly referred by the Japanese Road Association (JRA) for many decades. The database has 183 datasets with information on clayey (fine-grained) soils, and 151 datasets with information on sandy soils in alluvium. The regression equations, with a little modification to the regressed parameters for the purpose of practical use, were adopted by the JRA (2002) and implemented in their design codes. The database was updated by Imai and Tonouchi (1982), and the number of the datasets of clayey (fine-grained) soils became 325, and the number of the datasets of sandy soils became 294. The database established by Imai and Tonouchi (1982) (hereinafter referred to as IT database) are mainly measured by downhole seismic testing and suspension logging testing in alluvium. It is believed that the updated database should have a certain degree of reliability and could be a reference to compare with the database used in this study. Figure 7 displays the Vs versus N60-value distribution of the IT database. Note that the information of the SPT energy ratio, as well as the hammer type, are not available in Imai and Tonouchi (1982). According to Nishizawa et al. (1980), the energy ratio in Japanese hammer systems is from 63 to 72% for rope-cathead system and from 80 to 90% for automatic hammer system. An energy ratio of 72% was assumed, which is about average of the rope-cathead and automatic hammer systems in Japan, and the N-values of the data pairs in IT database were converted to N60-values accordingly. Besides, the data pairs with N60 larger than 50 in sandy soil group and 25 in fine-grained soil group do not show in the figure. From Figs. 5, 6, and 7, it can be seen that the data pairs collected from the SCPT and referred N-value are relatively scattered. Other than that, the Vs measurement results among the discussed methods are comparable. The difference among them would be discussed more in following section.

Fig. 7
figure 7

The databases of Imai and Tonouchi (1982) as well as regression curves

Full size image

4 SPT-based Vs correlations

Standard Penetration Test (SPT) has historically been the most widely used in situ geotechnical test throughout the world. In order to easily estimate shear wave velocity of soils by SPT-N value, many researchers have studied the relationship between the Vs and SPT-N value. From Figs. 5 and 6, it is obvious that different Vs measurement methods can give quite different values of Vs, but few studies addressed this difference and investigated the rationale of Vs measurements in alluvium. The Vs difference caused by measurement methods also results in different Vs versus N-value correlation equations. To address this, the study used the above built database and separately established correlation equations for each of the dataset categories. The relationship of Vs versus N-value follows the form of Eq. (4):

$${\text{Vs = a N}}_{ 6 0}^{\text{b}}$$
(4)

where “a” and “b” are the parameters to be determined by the regression analysis. The parameter “a” is a scaling factor, moving the values of Nb60 up or down, and the parameter “b”, called the power, determines the increasing rate of Vs as N value increases.

Note that these datasets typically contain a significant amount of scatter in the measurements (as evident in Figs. 5 and 6). To develop a better correlation, the “LINEST” function of Microsoft Excel was used to perform a best fit of the data in regression analysis. The coefficients of determination (R2) were used to assess the strength of the relationships between variables. Higher correlation coefficients indicate greater correlation between two variables. Perfect correlation between variables would result in a “R2” of 1.0. Further discussions of the regressed correlation equations, as well as the comparison with Japanese equations, are presented in the following sections.

4.1 Sandy soils

A summary of Vs versus N60-value correlation equations for sandy soils, as well as the equations from IT database of sandy soils (Imai and Tonouchi 1982), is presented in Table 5a. In view of the coefficient of determination (R2) shown in Fig. 5, the cross-hole seismic testing in this study and IT database have the highest R2, which demonstrates the reliability of the measurements of the shear wave velocity; the suspension logging testing in this study ranks the second most R2; the data pairs by SCPT and the referred N-value are the most scattered. On the basis of the similar Vs measurement method, the results measured by the suspension logging testing in this study are still much scattered than those in IT database. From the regression results based on all analyzed categories, basically, it can be seen that when parameter “a” is larger, parameter “b” will be smaller. Figure 5d shows the yielded regression curves, as well as the result by Imai and Tonouchi (1982). All regression curves are comparable. The results show that the predicted Vs (the red dotted curve in Fig. 5d) based on the data pairs of suspension logging testing is the least conservative. The one based on the data pairs measured by SCPT is the second. In contrast, the predicted Vs based on the regression curves by the data pairs of cross-hole seismic testing (the blue curve in Fig. 5d) are the most conservative, which is very close to the one derived by Imai and Tonouchi (1982).

Table 5 Vs versus N60-value correlation equations
Full size table

4.2 Fine-grained soils

A summary of Vs versus N60-value correlation equations for fine-grained soils, as well as the equations from IT database of cohesive soils (Imai and Tonouchi 1982), is presented in Table 5b. In view of the coefficient of determination (R2) shown in Fig. 6, the data pairs by cross-hole seismic testing in this study and those of IT database have the highest R2, which could demonstrate the better reliability of the Vs measurements; the R2 from the database of the suspension logging testing in this study ranks the second; the data pairs by SCPT and the referred N-value are the most scatter. The trend of the R2 is similar to that of sandy soils, except with a lower R2 in cohesive soils than that of sandy soils for the case of suspension logging testing. It is interesting that a lower R2 occurs when applying the suspension logging seismic testing to measure the shear wave velocity of fine-grained soils in this study. The reason will be discussed in the following section. From the regression results for all analyzed categories, a larger “a” parameter, will have a smaller “b” parameter. Figure 6d shows the regression curves as well as the result by Imai and Tonouchi (1982). All regression curves are basically comparable. From the figure, most of the predicted Vs from regression curve by the data pairs of cross-hole seismic testing is relatively conservative (the blue curve in Fig. 6d). However, the predicted Vs by the regression curve from the data pairs of suspension logging testing (the red dotted curve in Fig. 6d) is obviously higher than the other regression curves whereas the IT database is also mainly based on the suspension logging testing. The other two regression curves fall between.

Table 6 The predicted Vs for each site
Full size table

4.3 All soils

The Vs versus N60-value correlation equations for all soils used in this study is also presented in Table 5c. The parameters of “a” and “b” are 110.9 and 0.227, respectively. The coefficient of determination (R2) is 0.28. The engineers can use the equation to evaluate Vs of a soil layer if only the information of SPT-N value is available. However, it should be noted that the prediction would be definitely more uncertain than the one with the information of soil type.

4.4 Discussion

Among the measurement results in this study, cross-hole seismic testing is the most reliable method in view of the evaluation based on R-square. This result also agrees with comments by Andrus et al. (2004) in Table 2. In comparison with the results by Imai and Tonouchi (1982), the results between two are quite comparable in view of the evaluated R2 and the distribution of the data pairs. The data pairs by SCPT and the referred N-value show the most scatter. This may be the result of differing locations of the Vs and SPT measurements. Even so, the regression results based on these data pairs are still comparable with the ones based on the database of the cross-hole seismic testing. However, the SCPT measured Vs seems to be slightly higher than the one from cross-hole seismic testing.

As for the suspension logging testing, it is obvious that the regression curves based on the data measured by it in this study and in Imai and Tonouchi (1982) are different. The distribution of the data pairs is more scattered and the regression curve is non-conservative than the one proposed by Imai and Tonouchi (1982). A possible reason may be due to the difference in the preparation of the testing borehole. In Taiwan, the borehole for suspension logging has to be cased before measurement to avoid borehole collapse in alluvium sites. Based on the consultation with the Japanese engineers, the comments on the installation of the PVC casing for suspension logging testing are not totally consistent. Some suggest that the effect of the installation of the PVC casing is slight. In contrast, some think the installation of the PVC casing does affect the measurement results and should not be used as much as possible in order to derive better measurement. The consensus is that, in soft ground, the stability of the borehole should be evaluated first and the installation of PVC casing may be necessary in order not to stick the apparatus in the borehole. Once a borehole is cased, it is necessary to fill the gap space between ground formation and casing tube to ensure seismic waves propagate directly into the sensor without pollution of tube waves. Therefore, the borehole is always cased, and the PVC casing must be properly installed and grouted (see Fig. 8). Note that the shear wave propagation in suspension logging testing travels in the vertical direction, which means that the seismic wave may travel through grouted soils. Theoretically, the grout material should be as soft as the soils surrounding the borehole. In practices, it is really difficult to control the hardness of the grout accurately. In most cases, the hardness of grout is always greater than that of soft formations (i.e., lower N-value condition). This could cause overestimation of the shear wave velocity. Indirect evidence of the possible overestimation of Vs by the suspension logging test may be found by comparing the regression curves in Figs. 5 and 6 with the data pairs based on the SCPT. In SCPT, the travel path of the seismic wave is also in vertical direction, but it does not require the installation of the PVC casing and grouting. The record of the use of any casing in suspension logging testing is very important in order to let the designer make a reasonable engineering judgement. It is doubtful whether the problem of Vs overestimation also exists when conducting cross-hole seismic testing, where casing and the grouting of the testing boreholes are also required. Fortunately, the possible overestimation of Vs due to the effect of soil improvement could be largely reduced since the horizontal travel path of the shear wave propagation between the two testing boreholes (Fig. 2) is much longer than the grout-effected zones.

Fig. 8
figure 8

The installation of the PVC casing and grouting before conducting suspension logging testing in Taiwan

Full size image

Other than the above possible reasons, it should be noted that Sites A to K only had Vs measurements by cross-hole seismic testing and SCPT, which may be the reason that the regression curves based on the two Vs measurement methods were comparable. In contrast, Sites L to Y only had the Vs measurements by suspension logging testing. In the database of this study, none of the sites included the Vs measurements based on the three methods of interest. The variation of site characteristics may be another reason that the regression curves based on the suspension logging testing were different from the others.

5 Case study

In this case study, the influence of different regressed Vs-N curves deduced by different measurement methods on site classification will be demonstrated. Totally, four sets of regression curves were used. Three sets of regression curves were established by cross-hole logging, SCPT, and suspension logging respectively in this study and the other one set was directly cited from Imai and Tonouchi (1982). Each set had two regression curves, one for sandy soils and the other for fine-grained soils. In this demonstration, 10 sites located at the alluvium in central and south Taiwan were selected. They are sites M, N, O, R, S, U, V, W, X, and Y which are shown in Fig. 3 and denoted with new symbols as CHY004, CHY012, CHY016, CHY027, CHY031, CHY044, CHY049, CHY059, CHY066, and CHY112, accordingly. The Vs profile of each site was evaluated by SPT-N according to the 4 sets of the Vs-N regression curves. Table 6 shows the SPT results from the boreholes of the 10 sites, and the predicted Vs based on different sets of regression curves. With the shear wave velocities along the borehole, Vs30 could then be calculated according to Eq. (1), and the site classification could be determined by Table 2. The Vs30 of all sites are shown in Fig. 9. It can be found that all the 10 sites are classified into Site Class III by regression curves from cross-hole logging and Site Class II by regression curves from SCPT and suspension logging. If using regression curves suggested by Imai and Tonouchi (1982), four sites are classified into Site Class III and six sites are Site Class II. Therefore, it can be concluded that the influence of Vs measurement methods on site classification is at most one class difference. Following the above results, the subsequent influence on the design response spectra can be evaluated by comparing the site coefficients of Fa and Fv corresponding to the two evaluated site classes (Site Classes II and III). For a site located in the seismic region with Ss = 0.6 and S1 = 0.35 in Taiwan, for example, the site coefficients of Fa and Fv at the Class II site are 1.1 and 1.4 respectively while the Fa and Fv at the Site Class III are 1.2 and 1.7 respectively, according to MOI (2001). The ratios of Fa and Fv between the Site Class II and III are about 1.09 and 1.21 respectively. This means the design seismic force acting on a building located in Class III site would be larger than that of the same building located in Class II site by a difference of 9 ~ 21% which depends on the period of the building considered.

Fig. 9
figure 9

The predicted Vs30 for each site

Full size image

6 Conclusion

  1. 1.

    The established Vs-N database by cross-hole logging, SCPT, and suspension logging is believed to be reliable and provide a good reference for the relevant research work.

  2. 2.

    The discrepancies of the Vs-N correlations based on the database produced by different methods are found to be about tens of meters per second. The influence of Vs measurement methods on site classification is at most one class difference.

  3. 3.

    The predicted results of the site classification did vary if different sets of regression curves were adopted. The curves regressed from the datasets of cross-hole seismic testing are the most conservative in determining design seismic forces. Care must be taken when using less conservative regression curves by other measurement methods while conducting design work.

  4. 4.

    From the comparison study, it was found that the measured Vs by suspension logging testing were non-conservative than the other methods for shallow soft alluviums. This is because seismic waves produced by the suspension logging travel vertically along the surrounding disturbed region of borehole. The grouting between PVC case and borehole wall may affect the measurement results significantly.