ground motions in the lower wabash - GeoScienceWorld

Evaluation of Linear Site-Response Methods for Estimating Higher-Frequency ( > 2 Hz) Ground Motions in the Lower Wabash River Valley of the Central United States Edward Woolery, Ron Street, and Phyllis Hart

Edward Woolery,1 Ron Street, 2 and Phyllis Hart 3

ABSTRACT

INTRODUCTION

The horizontal-to-vertical ratio of ambient noise (H/V*N) and the time-averaged S-wave velocity of the upper 30 m of soils and rock at a site (VS30) were evaluated for their suitability as techniques for estimating site effects in the lower Wabash River Valley area of southern Indiana and Illinois. We also calculated the horizontal-to-vertical ratio of the S wave (H/V*S) for the southwestern Indiana earthquake of 18 June 2002 and evaluated its effectiveness for estimating site effects. The database for the study consisted of new SH-wave seismic refraction/reflection profiles and ambient noise recordings at several blast monitor sites that recorded the M 4.5, 18 June 2002 earthquake, as well as the S-wave arrivals recorded by the blast monitors during the earthquake. Our results are limited to ground motions > 2 Hz because of the low-cut frequency of the velocity transducers used in the blast monitor systems. The shallow seismic refraction/reflection profiles allowed us to determine the S-wave velocities and soil-bedrock contact geometry, as well as the depth to bedrock, characteristic frequency, and VS30. These methods were compared with linear one-dimensional site amplification approximations. The results of the study are not encouraging. There is only a weak correlation between the linear 1-D amplification curves and the site effects predicted by the two horizontal-to-vertical ratios or the VS30 classification of the site. Likewise, there is poor correlation between the two horizontal-to-vertical techniques or between VS30 and the Modified Mercalli intensities reported for the 18 June 2002 earthquake. The results show that site effects in southern Indiana and southern Illinois are too complex for single-parameter characterization; consequently, a better evaluation likely requires a minimum of site-specific in situ seismic-velocity measurements, geotechnical index tests, and one-dimensional approximation.

Paleoseismological evidence, historical earthquake accounts, and contemporary earthquake records indicate that the Wabash River Valley (WRV) of southern Indiana and Illinois in the central United States has a considerable seismic hazard (e.g., Street 1980; Obermeier et al. 1991; Pond and Martin 1997; Munson et al. 1997; Pavlis et al. 2002; Kim 2003; Herrmann et al. 2008). The pre-instrumental and instrumental evidence have shown that small to moderate earthquakes occur in an area roughly coincident with the Wabash Valley fault system (Figure 1). The low rate of seismicity and historically sparse seismic network coverage has made correlating seismicity with specific geological structure problematic, however. It is also well documented that near-surface soils and sediments with low shear-wave velocity, along with uneven surface and underlying bedrock topography, can have a profound effect on the ground motion at a specific location; thus, this is an important aspect of the seismic hazard assessment to be considered. In the lower WRV, as elsewhere in the central United States, the sitespecific seismic-response amplification is often defined by the time-averaged shear-wave velocity of the upper 30 m of soils and rock at the site (VS30). Provisions in the National Earthquake Hazards Reduction Program (NEHRP) for seismic regulations of new buildings (Federal Emergency Management Agency 1997) classify the site, based on VS30, into one of six categories used to estimate a local site response. These provisions are a result of the work done by Borcherdt et al. (1978), Borcherdt (1994), and Martin and Dorby (1994) with earthquake data that were for the most part acquired in California. Estimating site-response amplification from the properties of the upper 30 m of soil and rock at a site has produced equivocal results in past studies. Although Wald and Mori (2000) concluded that in the Los Angeles area the time-averaged shear-wave velocities of the upper 30 m of soil and rock at sites roughly correlated with observed site responses, the scatter was large and the method was inadequate for predicting site amplification. Sun et al. (2005) found that in areas where the

1. University of Kentucky 2. 13813 Werth Rd., Hermosa, SD 57744 3. Indiana Department of Natural Resources, Division of Reclamation doi: 10.1785/gssrl.80.3.525

Seismological Research Letters Volume 80, Number 3 May/June 2009 525

▲▲ Figure 1. Location map that shows approximate location of significant historical and contemporary earthquakes in relation to the mapped structures of the Wabash Valley fault system (modified from Bear et al. 1997 and Woolery 2005). Dashed circles indicate uncertainty in the instrumentally derived epicenters (filled circles). Shaded circles represent historical epicenters from individual investigator’s interpretation of intensity reports.

depth to the bedrock is less than 30 m and the S-wave velocity of the bedrock is significantly greater than the S-wave velocities of the bedrock encountered in the studies by Borcherdt et al. (1978), Borcherdt (1994), and Martin and Dorby (1994), the amplification factors used in the NEHRP site classifications are too low. Castellaro et al. (2008) concluded that site amplification is too complex to be characterized by any single parameter, such as VS30. Conversely, Molnar et al. (2004) concluded that the site amplification based on NEHRP average shear-wave velocity in the greater Victoria, British Columbia, area is in agreement with the intensities observed for the 2001 Nisqually earthquake in Washington state. Two other techniques that have been used for estimating site effects are the horizontal-to-vertical spectral ratios of ambient noise, sometimes referred to as Nakamura’s (1989) technique, and the horizontal-to-vertical spectral ratio of earthquake-generated S waves. A basic assumption for both techniques is that the vertical component of the ground motions is relatively free of near-surface site effects, whereas the horizontal components are amplified in such a way as to reflect the resonance period at the site and the extent of the ground motion amplification at that period. The reason for using these two techniques is the relative ease in acquiring and

processing the necessary data; however, the results have been ambiguous. Nakamura’s technique has been successfully used by Lermo and Chavez-Garcia (1993), Chavez-Garcia et al. (1996), Seekins et al. (1996), and De Luca et al. (2005), among others. Other studies, however, have shown that it is possible to deduce the resonance site frequency using Nakamura’s technique, but the amplification factor is unreliable (e.g., Field and Jacob 1995; Lachet et al. 1996; Bard 1999; Horike et al. 2001). Still other studies have concluded that Nakamura’s technique fails either in general or under certain conditions. For example, Riepel et al. (1998) and Parolai et al. (2004) noted that the vertical component of the ambient noise is influenced by twoand three-dimensional site complexities, and consequently if there are such complexities, the technique fails to correctly estimate the site effects. Giacomo et al. (2005) determined that the technique underestimates the amplification of the ground motions in the presence of an inversion in the S-wave velocities of the soils, while Satoh et al. (2001) concluded that the peak amplitudes and frequencies of the maximum peaks defined by Nakamura’s technique are, in general, inconsistent with the site effects established by other means. The horizontal-to-vertical spectral ratio of the S-wave technique for characterizing site effects has also been the sub-

526 Seismological Research Letters Volume 80, Number 3 May/June 2009

DATA SETS

▲▲ Figure 2. Location of the sites investigated for this study (o) and the epicenter of the 18 June 2002 southwestern Indiana earthquake (★) ; from Street et al. 2005).

ject of many studies and, as with Nakumura’s technique, has exhibited inconsistent results. Chen and Atkinson (2002) concluded that the horizontal-to-vertical spectral ratios of S waves are a good indicator of site effects if averaged over many events at the site, and Murphy and Eaton (2005) found that the horizontal-to-vertical spectral ratios of S waves are a good indicator of site effects if the site conditions can be characterized as a thick layer (e.g., 70 m) of sediment over bedrock. Coutel and Mora (1998) evaluated the horizontal-to-vertical spectral ratios of S-waves at a number of sites in the Brisbane area of Australia and determined that the fundamental resonance frequency and peak amplitude of the ground motions were estimated to be within 10 percent of being correct at locations where the site conditions could be characterized as consisting of horizontal layering; but like many other investigators (e.g., Riepel et al. 1998; Yuncha and Luzón 2000; Parolai et al. 2004), Coutel and Mora (1998) went on to note that the results were unreliable when used at sites with two- and three-dimensional nearsurface complexities. Our study in the Wabash River Valley evaluated the timeaverage of the shear-wave velocities in the upper 30 m of soils/ rock (VS30) and the horizontal-to-vertical spectral ratio of the ambient noise (H/V*N) relative to the linear one-dimensional spectral response for a vertically propagating S wave (SHAKE91: Idriss and Sun 1992). We also evaluated the horizontal-to-vertical spectral ratio of the S-wave arrival (H/V* S) for the 18 June 2002, southwestern Indiana earthquake at several of the blast monitor sites where it was recorded (Figure 2). The motive for this study was the mixed success of the Wald and Mori (2000) results and the possibility that the higherthan-expected peak ground motions for the 18 June 2002 earthquake in the lower WRV (Street et al. 2005) are caused in large part by simple near-surface conditions.

Three data sets were used in this study: the blast monitor velocity records of the 18 June 2002 southwestern Indiana earthquake, SH-wave seismic refraction/reflection profiles, and ambient noise samples at 27 of the blast monitor sites (Figure 2). The velocity records for the earthquake were acquired with three-component geophones buried in soil 20 to 35 cm beneath the surface. Ground motions from the geophones were sampled at the blast monitors at a rate of either 512 or 1,024 sps. The response of the geophone/blast monitor system used in the acquisition of the data is flat over a frequency range of 3 Hz to 100 Hz, and down 10 to 15 percent at 2 Hz. For a more detailed description of the installation, instrumentation, and recordings for the 18 June 2002 earthquake, see Street et al. (2005). The 27 SH-wave seismic refraction/reflection profiles were acquired with a seismic hammer striking a steel I-beam as the energy source and an inline array of either 24 or 48 horizontally polarized 30-Hz geophones spaced at intervals of 2 m. Each seismic profile was developed from energy shots applied at the array ends and center. At site 16, the energy source was stepped out from both ends of the inline 48-geophone array to increase the depth of penetration of the refraction profile. The decision as to the length of the array and the need for additional offsets was based on the depth to bedrock as interpreted from the first-arrival data recorded from the array in the field. Although local boreholes in which to corroborate our depthto-bedrock measurements were not available for this study, our experience at controlled sites in the region has shown that we are capable of achieving depth measurement accuracy between 1 and 3 percent with good reflected data quality (e.g., Woolery and Street 2002; Woolery 2005 among others). Optimum vertical resolution ranges between approximately 0.75 meters in the near-surface and 2 meters at depth. Figure 3 shows a typical reversed SH-wave seismic refraction/reflection sounding and the interpreted S-wave velocity model. The site numbers referred to in this investigation are the same as those used in Street et al. (2005). The seismic data were acquired at a sampling rate of 0.25 ms and were processed with the interactive software VISTA 7.0 (Seismic Image Software Ltd. 1995). The S-wave velocity models are determined from the first-break arrival times and the seismic-refraction algorithm SIPT2, version 4.1 (Rimrock Geophysics Inc. 1995). When good reflections were observed, such as the one seen in the right panel of Figure 3A, the intercept times and RMS velocities were used to constrain the S-wave velocity model interpreted from the refractions. Good reflections from the soil-rock interface were generally found throughout the study area. Three 15-s ambient noise samples were acquired at the midpoints of the geophone arrays for each of the sites. The noise samples were recorded with a 1-Hz, Mark Products L-4C seismometer at a sampling rate of 250 sps. The recording system did not include a low-cut filter, but it did have an active, antialiasing high-cut filter that rolled off at 12 db at 15 Hz. The L-4C seismometer was sheltered from the wind, and obvious coherent noise sources (e.g., passing traffic, etc.) were avoided.


time (ms)

time (ms)

(A)

(B)

▲▲ Figure 3. (A) Reversed SH-wave seismic refraction/reflection sounding acquired at site 31. Data were collected using an inline spread of 24 geophones spaced at 2 m. A strong reflection is present on the right seismic panel. It has a 0.180-s intercept time, a 230m/s velocity, and a 20.7-m calculated depth. The site number corresponds to that used in Street et al. (2005). (B) S-wave velocity and depth model for site 31 derived from the first arrival times.

The earthquake records, SH-wave refraction/reflection data, and ambient noise samples enabled us to calculate the time-averaged S-wave velocity of the upper 30 m of soil and rock (VS30), the characteristic frequency of the soils, and the 1-D linear site response, as well as the site-specific horizontal-to-vertical spectral ratios of the S wave of the earthquake (H/V* S) and ambient noise (H/V* N) for the 27 sites listed in Table 1. The characteristic frequency ( fc) of the soils is the timeaveraged S-wave velocity of the soil(s) at the site divided by four times the total thickness of the soil layer.

RESULTS FOR PREDICTIVE GROUND-MOTION TECHNIQUES VS30 and SHAKE91 Ground Motion Comparison Figure 4 shows the shear-wave velocity columns interpreted from the SH-wave seismic refraction/reflection data. Contacts between velocity layers are shown as horizontal, slanted, wavy, or slanted-wavy lines to indicate horizontal, sloping, irregular, and sloping irregular interfaces, respectively. In accordance with NEHRP provisions, bedrock is assumed to be the upper-


TABLE 1 Amplification (Amp) and frequencies (f) at which the Fourier amplitude spectrum peaked for the southwestern Indiana earthquake of 18 June 2002, as well as the amplifications and site frequencies estimated by SHAKE91. Also included in the table are the VS30 velocities, NEHRP site classifications, and Modified Mercalli intensities determined for the sites. Location

Observed

Site1 No.

°N

°W

1 2 3 4 5 6 7 12 13 14 15 16 19 20 21 25 30 31 32 33 34 35 36 37 40 41 42

37.746 37.750 37.756 37.756 37.757 37.758 37.759 38.128 38.121 38.227 38.225 38.089 38.092 38.253 38.365 38.285 38.306 38.339 38.354 38.352 38.352 38.354 38.354 38.168 38.804 38.805 38.605

88.337 88.361 88.368 88.369 88.374 88.375 88.377 87.35 87.358 87.391 87.393 87.285 87.275 87.350 87.365 87.364 87.349 87.341 87.293 87.283 87.278 87.279 87.276 86.946 87.491 87.474 87.009

Amp2 2.8 3.2 3.6 4.0 4.8 4.0 4.4 3.0 3.0 10.8 8.4 2.7 2.7 8.5 15.3 10.2 8.1 5.8 12.2 12.8 15.1 17.1 11.7 2.0 4.2 2.6 4.4

SHAKE91 f3

Amp4

f5

VS30 (m/s)

NEHRP Site Class

MMI6

7.8 9 12 12 12 12 5.5 20 11.5 19 19 10 10 15 2.1 9.1 5.8 3 22 13.3 18.5 7 18.5 11 11.5 6 12.1

11.5 7.3 6.7 4.7m 14.8 5m 6.1m 6 3.6/3.1 5 3.1/3 4.2m 6.9/3.5 17.7 4.3/5 5.1 3.7/3.2 2.7 4/3.9 4.2/3.5 3.9 5.2 2.7 3.9 2.8 5.5 2.6 m

6.7 8.9 2.1 5.1 6.6 5.8 6.3 7.1 6.3/13.2 9.3 5.3/15 10.1 7.1/9.1 14.4 14.8/17 2.8 2.7/14.7 2.0 7.0/21 5.0/12.8 12.7 6.6 12.9 14.0 11.9 11.8 2.4

675 755 755 755 840 840 590 580 760 740 740 900 900 935 250 740 450 305 800 755 930 435 925 1065 680 570 885

C C C C B B C C C C C B B B D C C D B C B C B B C C B

IV IV IV IV IV IV IV IV IV V V IV IV V V V V V V V V V V IV IV IV IV

1. Site numbers are the same as those used in Street et al. (2005). 2. Amplifications are the ratios of the PGVs reported in Street et al. (2005) to those predicted by the Atkinson and Boore (1995) model. 3. Frequencies of the peak amplitude as determined by the Fourier spectrum of the S wave. 4. Peak amplitude(s) of the ground motions determined from 1-D linear approximation. Multiple peaks (m) are listed for sites if the peaks resulting from the groundmotion modeling are approximately equal in amplitude. 5. Frequencies at which the peak amplitudes occur in the spectra of the ground motion modeling. 6. Modified Mercalli intensities from Street et al. (2005).

most layer having an S-wave velocity of ≥760 m/s. The average S-wave velocity of bedrock found in this study is 1,268 (± 402) m/s. The variation in bedrock velocity may in part be due to the stratigraphic variation of the underlying Pennsylvanian sandstones, siltstones, and shales, as well as their degree of weathering. Listed at the bottom of each velocity column is the distance in meters between the location of the blast monitor that had recorded the 18 June 2002 earthquake and the midpoint of the geophone array (Δ), the site characteristic frequency ( fc), and the time-averaged S-wave velocity (m/s) of the upper 30 m

at the site (VS30). The fc and VS30 given in Figure 4 are based on the S-wave velocities of the soils and rock at the array midpoint. One approach to a seismic hazard study is to estimate a bedrock ground-motion (e.g., Atkinson and Boore 1995), and utilizing VS30 from the refraction/reflection profile, apply a site-specific amplification coefficient to the bedrock ground motion parameter (e.g., peak velocity, acceleration, or a damped response spectral amplitude). Alternatively, the amplification coefficients for a site could be chosen on the basis of the peak amplitudes or spectral response determined from a 1-D linear


▲▲ Figure 4. S-wave velocity and depth models derived during this study. Straight lines indicate planar contacts that are horizontal or dipping; wavy lines are used to indicate irregular contacts. The S-wave velocities (m/s) are shown in the interiors of the columns, and scaling for the depths (m) is shown along the left side of the figure. At the bottom of the velocity columns are: (1) the distance between the midpoint of the geophone array and the blast monitor site (Δ), (2) the characteristic site frequency (fc ), and (3) the time-averaged shear-wave velocity of the upper 30 m of the soil and rock (VS30). The quantities VS30 and fC are based on the site conditions defined at the midpoints of the shear-wave velocity profiles. The characteristic site frequency is four times the total thickness of the soils divided by the time-averaged velocity of the soils.

transfer function. We note that the transfer function is the relationship between the soil properties and the amplification factor. Specifically, it is defined by soil properties such as damping ratio, layer thickness, shear modulus, unit weight, and seismic wave velocity. The geotechnical index parameters were not available at the sites in our investigation, thus introducing a source of uncertainty in the resultant SHAKE91 calculation. However, the pertinent aspect is the relative comparison of the bedrock and surface motions (i.e., amplification). Table 1 lists the VS30 site classification, the observed peak amplitudes and resonance frequencies at each site, and those determined from the SHAKE91 calculations. Figures 5A and 5B show the results using the Atkinson and Boore (1995) ground-motion model (i.e., M 4.5 earthquake corresponding to the location of the 18 June 2002 southwestern Indiana earthquake), along with the amplification coefficients (Table 1) derived from the VS30 and SHAKE91 calculations, respectively. The groundmotion parameter plotted in Figure 5 is peak ground veloc-

ity (PGV) because the blast monitors are particle velocity devices. The PGVs are indicated by either pluses or open circles, depending on whether the site was in Modified Mercalli Intensity (MMI) IV or V zone, respectively (Table 1). Ideally, the pluses and open circles in the plots of Figure 5 would plot in separate areas, with the open circles corresponding to the larger PGVs since they are from the blast monitors in the areas where MMI V effects were reported. In Figure 5A, the pluses and open circles cluster together, whereas in Figure 5B, the pluses and open circles, with the exception of the symbols for sites 6 and 32, are separate, with the open circles associated with the higher PGVs and the pluses associated with the lower PGVs. The dashed horizontal lines in the plots shown in Figure 5 indicate the mean PGV values for MMI IV and V suggested by Atkinson and Kaka (2007; Equation 1) for use in central United States ShakeMap applications. Based on the separation of the observed Modified Mercalli intensities, the amplified ground motions derived by the 1-D linear approximations are


(B)

(A)

▲▲ Figure 5. (A) Peak ground velocities predicted for the sites in the study estimated by assuming Atkinson and Boore’s (1995) groundmotion model and the amplification coefficients for NEHRP site classes B (o), C (+), and D (◊). The dashed lines indicate the mean PGV values for the Modified Mercalli intensity levels of IV and V in the central United States (Atkinson and Kaka 2007). The vertical arrows show the ± 1 standard deviation for the Modified Mercalli intensity levels, respectively. (B) Predicted peak ground velocities estimated by assuming Atkinson and Boore’s (1995) ground motion model and a vertically propagating S wave.

more apt to predict the correct MMI level than the amplified ground motions derived using the VS30 NEHRP methodology for this earthquake. H/V and SHAKE91 Spectral Comparison Figure 6A shows the results of the various spectral techniques used to characterize the site effects at site 31. Included in the figure are the 1-D linear site response approximations and the H/V spectral ratios from the earthquake record and ambient noise samples. The H/V* S spectral ratio was determined from a window centered on the arrival of the earthquake S wave. The window length used to calculate the H/V* S at the sites varied between 1 and 3.4 s, depending on the duration of the amplitude of the vertical component of the S wave with respect to amplitude of the background noise. The amplitude of the vertical component was, in all cases, the least energetic component of the blast monitor records. At site 31, as with all sites, the window was visually picked so that it began shortly before the onset of the S wave and continued until that point on the vertical trace at which the signal-to-noise ratio was less than 2. A 2.28-s window, typical for all sites, was chosen for site 31. The windowed three-component traces were then tapered with a 5 percent cosine curve and padded at both ends for a total of 8,192 points. Following customary practice, the individual horizontal and vertical components were transformed and smoothed and the square root of the sum of the squares of the transforms of the horizontal components was divided by two times the smoothed transform of the vertical component. Spectra were smoothed with the logarithmic window function (b = 20) proposed by Konno and Ohmachi (1998). The purpose of smoothing is to dampen large deviations in the ratios that could result from narrow-band spurious peaks and troughs in the spectra.

The horizontal-to-vertical spectral ratio of ambient noise for each site was calculated by taking the average of three 15-s windows. Like the S-wave samples, the traces for each noise sample were tapered with a 5 percent cosine curve, padded for a total sample length of 8,192 points, transformed, and smoothed with the logarithmic window function (b = 20) proposed by Konno and Ohmachi (1998). The ratio was then determined by taking the square root of the sum of the squares of the transforms of the horizontals and dividing the result by two times the transform of the vertical. The curves for the three spectral techniques shown in Figure 6A for site 31 are similar in shape and amplitude, and the amplitudes of the spectra are very close to the site amplification of the PGV from the earthquake. Two other sites, 7 and 30, have spectra from the three techniques in which the shape and amplification of the curves and the PGV are in general agreement. Figure 6B shows the spectra for site 7. The amplitude for the SHAKE91 response is high compared with the H/V techniques and observed amplification based on the PGV for the earthquake; however, there is a general agreement in the shape of the three spectra at approximately 7 Hz, the fundamental resonance frequency. The higher-than-expected amplification exhibited by the response spectrum for the 1-D linear model is likely the result of the uncertainty in the assumed geotechnical index properties and/or the nonhorizontal interfaces. In a sitespecific design-level investigation that includes geotechnical sampling and testing, the material properties of the soils would be better defined, and the amplitude of the resulting response spectra for the site may be more similar to those estimated by the H/V techniques at these sites. Sites 7 and 31 are two of the three sites in the study that showed general agreement between the response spectra; site


(A)

(B)

▲▲ Figure 6. (A) Comparison of horizontal-to-vertical ratios of the ambient noise (dashed line), S wave (dot-dash line), and SHAKE91 (solid line) spectra at site 31. The characteristic site frequency (fc ) is 3 Hz. The ratio of the recorded PGV to that predicted by Atkinson and Boore (1995) is 5.8 as indicated by the horizontal line. The peaks and troughs in the H/V spectral ratios are approximately 3 and 6.5 Hz. (B) Spectra and PGV amplification for site 7.


▲▲ Figure 7. The similar in shape but shifted in frequency relationship between the spectra resulting from the SHAKE91 and H/V*S for site 1. Similar results between the two sets of spectra were obtained for sites 1, 5, 14, 19, 20, and 40. The characteristic site frequency (fc ) is 7.8 Hz. The amplification of the PGV, which we define as the ratio of the recorded PGV to that predicted by Atkinson and Boore (1995), is 2.8 as indicated by the horizontal line.

30 (not shown) is the other site. At sites 1, 5, 14, 19, 20, and 40, we observed a general correlation between the spectral results of the SHAKE91 models and H/V* S. We assume that differences in the peak spectral frequencies for the two techniques were due to differences in the soil geometries and depth to bedrock at the actual blast monitor locations and the recording locations (i.e., midpoint of the seismic refraction/reflection arrays). The SHAKE91 spectral responses and H/V* S at site 1 are shown in Figure 7 and are similar to the fit between the two spectra observed at sites 5, 14, 19, 20, and 40. Specifically, the two spectra are similar in shape, but shifted in frequency, with respect to one another. Outside the spectral results for sites 7, 30, and 31, there is little similarity between the H/V*N, the SHAKE91 models, or the H/V*S spectra. An additional correlation exception is the H/V*N and SHAKE91 spectra at site 42 (Figure 8A). An observed correlation between the shape, not the amplitude, of the H/V*N and SHAKE91 spectra is found at sites 2, 5, and 25. Site 2 is an example of this type of correlation (Figure 8B).

DISCUSSION The objective of this study was to evaluate the VS30, H/V*N, and H/V*S field techniques for predicting earthquake-generated ground motions of >2 Hz in the lower Wabash River Valley area. Based on the ground motions observed for the

18 June 2002 southwestern Indiana earthquake and our site investigations (Figure 4), we conclude that the VS30 and H/V*N techniques are not reliable enough to be used as predictors of linear ground motions (> 2 Hz) resulting from moderate earthquakes in the lower WRV. The H/V*S technique shows promise because the spectra exhibit a general correlation with the spectra calculated by the 1-D linear approximation; however, given infrequent earthquakes, the sparseness of instrumented locations and the variability in the site conditions, this field technique is not a viable reconnaissance tool for the area. There are several potential reasons why the ground motions estimated using the VS30 and H/V*N techniques might fail to agree with those observed for the 18 June 2002 earthquake. One possibility is that many of the intra-soil interfaces, as well as the soil-bedrock interfaces, are not horizontal but sloping and/or irregular (Figure 4). Such features are a potential source of scattering, focusing, and defocusing, particularly when considering higher frequency (>2 Hz) waves in which the wavelengths of some of the seismic waves are on the same order of magnitude as the irregularities (e.g., Ohtsuki and Harumi 1983; Aki 1988; Harmsen 1997). In addition, the extent of the apparent sloping and irregular contacts in the subsurface is likely more pervasive than suggested by the seismic-refraction profiles, because they are two-dimensional images interpreted from single inline seismic spreads. Multiple seismic profiles at the blast monitor sites would have defined the true dip angle


(A)

(B)

▲▲ Figure 8. (A) The SHAKE91 (solid line) and ambient noise (dashed line) spectra determined for site 42. The characteristic site frequency (fc ) is 12 Hz, and the ratio of the recorded PGV to that predicted by Atkinson and Boore (1995) is 4.4, indicated by the horizontal line. (B) The SHAKE91 (solid line) and ambient noise (dashed line) spectra determined for site 2, showing a similarity in shape but not in amplitude. The characteristic site frequency (fc ) is 9 Hz, and the ratio of the recorded PGV to that predicted by Atkinson and Boore (1995) is 3.2, indicated by the horizontal line.


in the various soil contacts, as well as the soil-bedrock interface. The VS30 and H/V*N techniques assume, at least to some extent, horizontal layering of the soils and top of the bedrock (as does the one-dimensional linear modeling). Another potential reason that the VS30 criteria may not adequately predict ground motions in this area is the variability of the S-wave bedrock velocities between sites. We acquired sufficient seismic data to define the S-wave velocity models to a depth of 30 m. Results show the S-wave velocities of the soils and bedrock at the sites are highly variable; the soil velocities range between 113 and 721 m/s and the bedrock velocities range between 766 and 2,318 m/s. This is not unique: the S-wave velocities of the soils and bedrock in the seismically active regions of the western United States are likewise highly variable (Abrahamson 1995; Joyner 1995). In such areas, the impedance contrasts within the soil column, at the soil-bedrock interface, and within the rock formations can all have a pronounced influence on the surface ground motions through such processes as selective amplification and scattering. It may be that the VS30 criteria are too generalized to adequately predict site-specific ground motions in areas where the S-wave velocities of the soils and bedrock change rapidly over short distances. The variability in the soils and bedrock S-wave velocities, as well as potential sloping and irregular contacts, suggest that site conditions can change appreciably over distances of as little as a few tens of meters. Wald and Mori (2000) observed significant variations in the site effects at two sites less than 400 m apart, and Hartzell et al. (1996) noted that site effects can vary by as much as a factor of 2 over a distance of 200 m, even when the geologic units are the same. This also indicates the possibility that part of the resultant dissimilarity between the various field techniques maybe due to the distance (Δ) between the blast monitor locations and the field measurements. This possibility is supported by the results of the near-surface seismic profiling at site 21. The blast monitor at site 21 is located in a field and is approximately 200 m from two orthogonal publicaccess roads. The velocity column shown in Street et al. (2005) was interpreted from a seismic refraction/reflection profile located in a floodplain along the north-south road, whereas a second velocity column was interpreted for this study from data acquired along a loess ridgeline on the perpendicular eastwest road. The elevation difference is less than 10 meters and the distance between the ends of the two profiles is approximately 500 m. The reversed seismic characteristics and their interpretations are very different, however (Figure 9). The topof-rock elevation is similar when the profiles are adjusted for elevation differences, but these results clearly show that site conditions in the study area can be very different over distances as small as 0.5 km. There is also the potential that ground motions are affected as the result of rock formations below 30 m. Anderson et al. (1996), among others, has pointed out that layering and formations within the rock well below 30 m can contribute to site effects. Hartzell et al. (1997) arrived at a similar conclusion, as did Wald and Mori (2000), who concluded that significant site

effects can result from complexities in the propagation path, including materials deeper than 30 m. Given the wide range of S-wave velocities of the near-surface (≤ 30 m) bedrock in this area, it is likely that material property variations and structural features (i.e., impedance contrasts, velocity gradients, and three-dimensional structures) present in rock deeper than that sampled in this study play a role in the site effects. The quality and quantity of the data available for determining site effects via the H/V*S may also be problematic. The seismogram quality at some sites was not optimal, primarily because of the low amplitude of the vertical component at onset of the S wave. Several sites that recorded the earthquake and are listed in Street et al. (2005) are not included in this study because of the low amplitude of the vertical component of the ground motions. Perhaps a larger earthquake, or one in which the vertical component of the ground motions at the onset of the S wave was more energetic, might lead to improved results. As suggested by Chen and Atkinson (2002), H/V*S may be a valid technique for estimating site effects if the results are averaged over many events. Since approximately only one event of M 4.5 or greater occurs every 10 to 15 years in the lower Wabash River Valley area, it is unlikely that H/V*S is a technique that can be of any practical use in the near future if the technique requires H/V*S to be averaged over many events.

CONCLUSIONS The ambient noise and S-wave (earthquake) horizontal-to-vertical methods, as well as the VS30 field method, were evaluated for estimating ground-motion site effects (>2 Hz) at locations in southern Indiana and southern Illinois where the 18 June 2002 earthquake was recorded on blast monitors. The goal was to determine if any of these easy-to-use methods were effective for estimating site effects in this area. The results indicate the horizontal-to-vertical ratio of the ambient noise method is too inconsistent to use as a site-specific seismic hazard tool and, similar to Wald and Mori (2000), show that there is large scatter in the VS30 method that makes it unreliable as well. The horizontal-to-vertical ratio of the S-wave method for estimating site effects might have potential if a sufficient number of events can be recorded and averaged at a site (Chen and Atkinson 2002), but based on our results using this single event, it too fails as a site-specific seismic hazard tool for this area. The results show that site effects for locations in the lower Wabash River Valley are site-specific and too complex to be evaluated by any single parameter. It is likely that site effects in this area are best estimated using a combination of conventional in situ seismic velocity measurements, geotechnical index tests, and the 1-D linear approximation. We also recommend that the subsurface seismic-velocity sampling at a site be extended deeper than 30 meters, because there is sufficient evidence in our data set to suggest that the site response also depends to some extent on the impedance and geometry of deeper materials (see also, for example, Anderson et al. 1996; Hartzell et al. 1997; Wald and Mori 2000).


time (ms)

time (ms)

(A)

Multiple

Multiple

(C)

time (ms)

time (ms)

(B)

▲▲ Figure 9. (A) Site 21 reversed SH-wave seismic refraction/reflection data acquired along the east-west road as part of this study. (B) SH-seismic refraction/reflection data collected by Street et al. (2005) along an adjacent north-south road ~0.5 km from the former. Acquisition and processing parameters for the two sets of seismic lines are the same, except that the interval spacing used along the east-west road is 4 m and the interval spacing used along the north-south road is 2 m. (C) Site 21 velocity model derived by Street et al. (2005).


ACKNOWLEDGMENTS We wish to thank Messrs. F. A. Rutledge, W. F. Reid, D. M. Vance, and D. X. Wang for their considerable effort in the field activities. The authors credit the valuable technical reviews from Martin Chapman and Rob Williams. Meg Smath and Collie Rulo of the Kentucky Geological Survey also improved the manuscript with editorial and graphical support. We want to express our appreciation to Greg Steiner (University of Memphis, retired) who assisted us with the ambient noise instrumentation. The research was supported by U.S. Geological Survey grant 04HQGR0095 and assistance from the Kentucky Geological Survey. The views and conclusions presented in this paper are those of the authors and do not represent the official policies, either expressed or implied, of the U.S. government or the Commonwealth of Kentucky.

REFERENCES Abrahamson, N. (1995). Rock should be the reference condition. In Proceedings of National Earthquake Ground Motion Mapping Workshop, 27–31. Redwood City, CA: Applied Technology Council, ATC-35-2; http://www.atcouncil.org. Aki, K. (1988). Local site effects on strong motion. In Earthquake Engineering and Soil Dynamics II, ed. J. L. Von Thun, 103–155. Geotechnical Special Publication no. 20. New York: American Society of Civil Engineers. Anderson, J. G, Y. Lee, Y. Zeng, and S. Day (1996). Control of strong motion by the upper 30 meters. Bulletin of the Seismological Society of America 86, 1,749–1,759. Atkinson, G. M., and D. M. Boore (1995). Ground-motion relations for eastern North America. Bulletin of the Seismological Society of America 85, 17–30. Atkinson, G. M., and S. I. Kala (2007). Relationships between felt intensity and instrumental ground motion: Implications for ShakeMaps in the central U.S. and other regions. Bulletin of the Seismological Society of America 97, 497–510. Bard, P. Y. (1999). The effects of surface geology on seismic motion. In Microtremor Measurements: A Tool for Site Effect Estimation? ed. K. Irikara, K. Kudo, H. Okada, and T. Sasatani, 1,251–1,279. Rotterdam: Balkema. Bear, G. W., J. A. Rupp, and A. J. Rudman (1997). Seismic interpretation of the deep structure of the Wabash Valley fault system. Seismological Research Letters 68, 624–640. Borcherdt, R. D. (1994). Estimates of site-dependent response spectra for design (methodology and justification). Earthquake Spectra 10, 617–653. Borcherdt, R. D., J. F. Gibbs, and T. E. Fumal (1978). Progress on ground motion predictions for the San Francisco Bay region, California. In Proceedings, Second International Conference on Microzonation for Safer Construction— Research and Application, Appendix 2.1, 13–25. San Francisco, CA: International Conference on Microzonation for Safer Construction. Castellaro, S., F. Mulargia, and P. L. Rossi (2008). Vs30 Proxy for seismic amplification? Seismological Research Letters 79, 540–543. Chavez-Garcia, F. J., L. R. Sanchez, and D. Hatzfeld (1996). Topographic site effects and HVSR, a comparison between observation and theory. Bulletin of the Seismological Society of America 86, 1,559–1,573. Chen, S.-Z., and G. M. Atkinson (2002). Global comparisons of earthquake source spectra. Bulletin of the Seismological Society of America 92, 885–895. Coutel, F., and P. Mora (1998). Simulation-based comparison of four site response estimation techniques. Bulletin of the Seismological Society of America 88, 30–42.

De Luca, G., S. Marcucci, G. Milana, and T. Sanò (2005). Evidence of low-frequency amplification in the city of L’Aquila, central Italy, through a multidisciplinary approach including strong- and weakmotion data, ambient noise, and numerical modeling. Bulletin of the Seismological Society of America 95, 1,469–1,481. Federal Emergency Management Agency (1997). 1997 NEHRP Recommended Provisions for Seismic Regulations for New Buildings. Washington, DC: Building Seismic Safety Council/Federal Emergency Management Agency. Field, E. H., and K. H. Jacob (1995). A comparison and test of site response estimation techniques, including three that are non-reference-site dependent. Bulletin of the Seismological Society of America 85, 1,127–1,143. Giacomo, D. D., M. R. Gallipoli, M. Mucciarelli, S. Parolai, and S. M. Richwalski (2005). Analysis and modeling of HVSR in the presence of a velocity inversion: The case of Venosa, Italy. Bulletin of the Seismological Society of America 95, 2,364–2,372. Harmsen, C. (1997). Determination of site amplification in Los Angeles urban area from inversion of strong-motion records. Bulletin of the Seismological Society of America 87, 866–887. Hartzell, S., A. Leeds, A. Frankel, and J. Michael (1996). Site response for urban Los Angeles using aftershocks of the Northridge earthquake. Bulletin of the Seismological Society of America 86 (1B), 168–192. Hartzell, S., E. Cranswick, A. Frankel, D. Carver, and M. Meremonte (1997). Variability of site response in the Los Angeles urban area. Bulletin of the Seismological Society of America 87, 1,377–1,400. Herrmann, R. B., M. Withers, and H. Benz (2008). The April 18, 2008 southern Illinois earthquake: An ANSS monitoring success. Seismological Research Letters 79, 830–843. Horike, M., B. Zhao, and H. Kawase (2001). Comparison of site response characteristics inferred from microtremors and earthquake shear waves. Bulletin of the Seismological Society of America 91, 1,526–1,536. Idriss, M., and J. I. Sun (1992). User’s Manual for SHAKE91: A Computer Program for Conducting Equivalent Linear Seismic Response Analysis of Horizontally Layered Soil Deposits. Davis, CA: Center for Geotechnical Modeling, University of California, Davis. Joyner, W. B. (1995). Soil is the most appropriate reference site condition. In Proceedings of National Earthquake Ground Motion Mapping Workshop, 33–44. Redwood City, CA: Applied Technology Council, ATC-35-2; http://www.atcouncil.org. Kim, W.-Y. (2003). The 18 June 2002 Caborn, Indiana, earthquake: Reactivation of ancient rift in the Wabash Valley seismic zone. Bulletin of the Seismological Society of America 93, 2,201–2,211. Konno, K., and T. Ohmachi (1998). Ground-motion characteristics estimated from spectral ratio between horizontal and vertical components of microtremor. Bulletin of the Seismological Society of America 88, 228–241. Lachet, C., D. Hatzfeld, P.-Y. Bard, N. Theodulidis, C. Papaioannou, and A. Savvaidis (1996). Site effects and microzonation in the city of Thessaloniki (Greece): Comparison of different approaches. Bulletin of the Seismological Society of America 86, 1,692–1,703. Lermo, J., and F. J. Chavez-Garcia (1993). Site effects evaluation using spectral ratios with one station. Bulletin of the Seismological Society of America 83, 1,574–1,594. Martin, G. R., and R. Dobry (1994). Earthquake site response and seismic code provisions. National Center for Earthquake Engineering Research Bulletin 8, 1–6. Molnar, S., J. F. Cassidy, and S. E. Dosso (2004). Comparing intensity variation of the 2001 Nisqually earthquake with geology in Victoria, British Columbia. Bulletin of the Seismological Society of America 94, 2,229–2,238. Munson, P. J., S. F. Obermeier, C. A. Munson, and E. R. Hajic (1997). Liquefaction evidence for Holocene and latest Pleistocene seismicity in the southern halves of Indiana and Illinois: A preliminary overview. Seismological Research Letters 68, 521–536.


Murphy, C., and D. Eaton (2005). Empirical site response for POLARIS stations in southern Ontario, Canada. Seismological Research Letters 76, 99–109. Nakamura, Y. (1989). A method for dynamic characteristics estimation of subsurface using microtremor on the ground surface. Quarterly Report of Railway Technology Research Institute 30 (1), 25–33. Obermeier, S. F., N. K. Bleuer, C. A. Munson, P. J. Munson, W. S. Martin, K. M. McWilliams, D. A. Tabaczynski, J. K. Odum, M. Rubin, and D. L. Eggert (1991). Evidence of strong earthquake shaking in the lower Wabash Valley from prehistoric liquefaction features. Science 251, 1,061–1,063. Ohtsuki, A., and K. Harumi (1983). Effect of topography and subsurface inhomogeneities on seismic SV waves. Earthquake Engineering and Structural Dynamics 11, 441–462. Parolai, D., D. Bindi, M. Baumbach, H. Grosser, C. Milkereit, S. Karakisa, and S. Zünbül (2004). Comparison of different site response techniques using aftershocks of the 1999 Izmit earthquake. Bulletin of the Seismological Society of America 94, 1,096–1,108. Pavlis, G. L., A. L. Rudman, B. M. Pope, M. W. Hamburger, G. W. Bear, and A. S. Haydar (2002). Seismicity of the Wabash Valley seismic zone based on a temporary seismic array equipment. Seismological Research Letters 73, 751–761. Pond, E. C., and J. R. Martin (1997). Estimated magnitudes and accelerations associated with prehistoric earthquakes in the Wabash Valley region of the central United States. Seismological Research Letters 68, 611–623. Riepel, J., P.-Y. Bard, D. Hatzfeld, C. Papaioannou, and S. Nechtschein (1998). Detailed evaluation of site-response estimation methods across and along the sedimentary valley of Vovi (EURO-SEISTEST). Bulletin of the Seismological Society of America 88, 488–502. Rimrock Geophysics Inc. (1995). SIPT2 Version 4.1 (and other programs), 56 pps. Satoh, T., H. Kawase, and S. Matsushima (2001). Differences between site characteristics obtained from microtremors, S waves, P waves, and codas. Bulletin of the Seismological Society of America 91, 313– 334.

Seekins, L. C., L. Wennerberg, L. Margheriti, and H.-P. Liu (1996). Site amplifications at five locations in San Francisco, California: A comparison of S waves, codas, and microtremor. Bulletin of the Seismological Society of America 86, 627–635. Seismic Image Software Ltd. (1995). VISTA 7.0, notes, 155 pps. Street, R. (1980). The southern Illinois earthquake of September 27, 1891. Bulletin of the Seismological Society of America 70, 915–920. Street, R., J. Wiegand, E. W. Woolery, and P. Hart (2005). Groundmotion parameters of the southwestern Indiana earthquake of 18 June 2002 and the disparity between the observed and predicted values. Seismological Research Letters 76, 512–530. Sun, C. G., D. S. Kim, and C. K. Chung (2005). Geologic site conditions and site coefficients for estimating earthquake ground motions in the inland areas of Korea. Engineering Geology 81, 446–469. Wald, L. A., and J. Mori (2000). Evaluation of methods for estimating linear site-response amplifications in the Los Angeles region. Bulletin of the Seismological Society of America 90 (6B), S32–S42. Woolery, E. W., and R. Street (2002). Quaternary fault reactivation in the Fluorspar area fault complex of western Kentucky. Seismological Research Letters 73, 628–639. Woolery, E. W. (2005). Geophysical and geological evidence of neotectonic deformation along the Hovey Lake fault, lower Wabash Valley fault system, central United States. Bulletin of the Seismological Society of America 95, 1,193–1,201. Yuncha, Z. A., and F. Luzón (2000). On the horizontal-to-vertical spectral ratio in sedimentary basins. Bulletin of the Seismological Society of America 90, 1,101–1,106.

Department of Earth and Environmental Sciences University of Kentucky Lexington, Kentucky 40506-0053 U.S.A.


[email protected]

(E. W.)