REFRACTION MICROTREMOR AND OPTIMIZATION METHODS AS ALTERNATIVES TO BOREHOLES FOR SITE STRENGTH AND EARTHQUAKE HAZARD ASSESSMENTS.

Transcription

1 REFRACTION MICROTREMOR AND OPTIMIZATION METHODS AS ALTERNATIVES TO BOREHOLES FOR SITE STRENGTH AND EARTHQUAKE HAZARD ASSESSMENTS John N. Louie and Robert E. Abbott, University of Nevada, Reno, NV Satish Pullammanappallil, Optim LLC, Reno, NV Abstract A thorough assessment of shallow shear velocity is important to both earthquake-hazard assessment and efficient foundation design. The only standard procedure for determining shear velocity, crosshole seismic (ASTM D8), is not much used as it requires two boreholes with high-precision positional logs. Downhole shear-wave profiles in a single hole are adequately accurate, but still too expensive for many projects. We tested two methods for inexpensively estimating shallow shear velocities with seismic refraction equipment, at the sites of several boreholes in California and Nevada. The sites ranged from hard to soft (NEHRP hazard classes A to D). The first method, refraction microtremor, records ambient ground noise on simple seismic refraction equipment (as in ASTM D5777). Wavefield analysis of the noise allows picking of Rayleigh-wave phase velocities. It works well in dense urban areas and transportation corridors. The second method, by Optim LLC, takes standard (ASTM D5777) seismic refraction arrival picks and finds an optimized -d P-velocity model. The shear velocities estimated from both methods are just as effective as borehole velocities for two purposes: estimating 30- meter depth-averaged shear velocity for foundation design; and estimating the seismic spectrum input for earthquake-hazard evaluations of sites. Shear-Velocity Estimation Methods A thorough assessment of shallow shear velocity is important to earthquake-hazard assessment. Frankel et al. (199) developed seismic-hazard maps for the United States that account for the geographic patterns of earthquake occurrence, but not for local site conditions. To account for path and site effects on earthquake shaking in the Los Angeles region, the Southern California Earthquake Center (SCEC) developed a "Community Velocity Model" (CVM) by summarizing the results of interdisciplinary investigations (Magistrale et al., 000). The development of the SCEC CVM focused on obtaining information at two scales: within important sedimentary basins; and for a "geotechnical layer" within 00 meters of the surface. Seekins et al. (199), Steidl et al. (199), Bonilla et al. (1997), Satoh et al. (001b), and others have found geotechnical velocities helpful in resolving differences among siteamplification estimation techniques. Shear velocity is a fundamental base parameter in understanding nonlinear site effects as well, as in Ni et al. (1997). Shallow shear velocity can also guide efficient foundation design. Shear velocity is easily related to the bulk shear modulus of the soil or rock at a foundation site. In the design of drilled shaft foundations to carry large axial loads, for example, a downhole assessment of shear velocity has proven effective in reducing expensive over-design (Bowman et al., 001). As such foundations typically are at least several meters in dimension, the axial capacity in question is also a bulk average over distances of several meters. The American Society for Testing and Materials (ASTM) describes only one standard procedure for determining shear velocity, the crosshole seismic method (ASTM D8). Crosshole seismic provides as little as 3% velocity error in the absence of lateral heterogenity. The method is not much used, however, as it requires at least two and preferably three boreholes, each having high-precision positional logs. The SCEC CVM (Magistrale et al., 000), for example, encompasses very few crosshole seismic measurements. Downhole shear-wave profiles are far more common than crosshole results in summaries such as the SCEC CVM and Boore and Joyner (1997). The downhole shear-wave technique is not described by an ASTM standard, but practitioners generally hold to the applicable parts of ASTM D5753 for borehole geophysical logging and ASTM D5777 for P-wave seismic refraction surveys. A downhole profile is measured in a single borehole and does not require positional logging, so is much less expensive than 1

2 crosshole work. Downhole shear velocities may be just as accurate as crosshole estimates in the presence of strong lateral heterogeneity. However, while the crosshole method provides volume averaging over the distance between the holes, the downhole method is essentially a point sample at one location. Adjacent downhole surveys often show very different velocity profiles. This limited geometric extent is also a weakness of the OYO downhole shear-velocity logging tool. While the downhole method will at least provide a very accurate average velocity between the surface and the bottom of the hole, the logging tool never averages over distances greater than several meters. We tested two methods for estimating shallow shear velocities using surface seismic refraction data, without a borehole. In our experience (Louie, 001), surface surveys never provide the detailed velocity accuracy of borehole profiles. But, surface data also average over larger volumes, and thus suffer less variability at laterally heterogeneous sites. Our main motivation is to provide engineers with much quicker and less expensive velocity assessment techniques. We present here the results of two types of surface-based seismic velocity estimation methods. We compare the surface methods against borehole results at several sites in California and Nevada. The sites ranged from hard to soft (shear velocities from 3000 m/s down to 00 m/s). Since we wish to compare the methods on a basis of how engineers make use of velocity results, we will compare the amplitude spectra of seismic impulses as they are affected by the velocity estimates. Seismic spectra are somewhat sensitive to vertical averages of shear velocity, in addition to the details of the velocity profiles. The first surface method, refraction microtremor (Louie, 001), records ambient ground noise on simple seismic refraction equipment (as in ASTM D5777). Ambient ground noise in urban areas is largely Rayleigh waves resulting from the passage of vehicles over uneven road surfaces, and coupling of tall structures with wind and air waves (Horike, 1985; Liu et al., 000; Satoh et al., 001a,b). Wavefield analysis of the noise (as in figure 1, grayscale images) clearly shows dispersion of Rayleigh-wave phase velocities with frequency (Louie, 001). We model the phase-velocity dispersion with a shear-velocity profile. Refraction microtremor works better than methods such as SASW and MASW, that require active seismic sources, in dense urban areas and transportation corridors. However, a refraction array at least 00 meters long must be used. Thus, refraction microtremor provides a reconnaissance velocity profile that averages over an area hundreds of meters wide. The second method, by Optim LLC, takes standard seismic refraction arrival picks and finds an optimized -d P-velocity model. The optimization of first-arrival time picks follows the method published by Pullammanappallil and Louie (199) The P-wave refraction data are collected according to ASTM standard D5777. SeisOpt -d P-velocity sections (as in figure 1, color images) have proved very effective in several geotechnical applications, including in predicting ripability for excavation engineering, and in mapping lateral velocity variations across faults and discontinuities. For comparison against the 1-d velocity results from downhole and refraction microtremor analyses, we average the -d SeisOpt models horizontally. We also apply the Poisson-solid assumption to the P-wave velocities, converting the SeisOpt profiles to shear-velocity profiles by dividing velocities by the square root of 3. Borehole and Refraction Measurements Figure shows all velocity profiles available from three hard-rock sites in southern California and Nevada. Our refraction measurements at Keenwild and the Pinon Flat Observatory (PFO) were sponsored by SCEC; the Nevada Seismological Lab sponsored our measurements on the Yucca Mountain crest. Downhole shear-wave profiles at Keenwild and PFO were provided by F. Vernon (pers. comm. 001). The hole N7 downhole profile was measured by B. Redpath (G. Biasi, pers. comm. 001). The refraction-microtremor dispersion modeling was conducted completely independently from any knowledge of the downhole velocities. At all three sites the 30-meter depth-averaged velocities match well, within 15% of the downhole velocities. These hard-rock sites have a high degree of lateral variability between adjacent borehole measurements (not shown). Refraction microtremor, being a surface technique, distinguishes vertically averaged velocities much better than it can distinguish the depth of an interface from the velocity contrast across it (Louie, 001). The SeisOpt profiles of figure show the vertical averaging explicitly. Since they are lateral averages across 100-meter distances, the smooth profile results from combining the lateral variations of layer depth and velocity across a large area. If the Keenwild or PFO site were to truly have a significant

3 velocity contrast at one constant depth, that would appear as a discontinuity on the SeisOpt profiles. The absence of any such constant-depth interface is direct evidence of just how subject the downhole profiles are to lateral variations. Figure also plots for comparison the Generic Rock summary of Boore and Joyner (1997), and the B-C Boundary model of Frankel et al. (199). It is important to note that the most important feature distinguishing the B-C boundary model is a lower velocity at the surface. The southern California rock sites also show strong velocity gradients in the upper ten meters. The Yucca Mountain crest does not show this feature in the refraction microtremor model. That result shows almost no velocity gradient near the surface. In that data set we were able to observe obvious Rayleigh-wave dispersion up to a frequency of 0 Hz. The observed dispersion showed no evidence whatever for low velocities at the surface. This conclusion only holds, of course, for lateral averages of velocities over 100-meter distances. A comparison of velocity profiles at a hard-rock site near Reno in northern Nevada shows the point-sampling effect on downhole velocities. Extension of Interstate Highway 580 to link Carson City to the Interstate system requires construction of a concrete-arch bridge with 300-meter span. To design the arch abutments to bear the huge loads, Black Eagle Consultants, Reno, contracted B. Redpath to log holes 30 meters deep at the foundation sites. Despite being only 300 m apart, figure 3 shows that the downhole logs suggest the north abutment site has velocities higher by almost a factor of two over the south abutment. Other logs only 50 m away show similar heterogeneity (Bowman et al., 001). The refraction microtremor measurements were made with arrays 00 m long. Each array ends at one of the logged boreholes, and the centers of the arrays are about 500 m apart. The velocities modeled from the refraction-microtremor dispersion curves are high, and these sites are in hard volcanic rock. The spacing has detected the significantly different 30-m depth-averaged velocities between the north and south abutment sites. However, the averaging over the lateral variations by the arrays (figure 3, thick lines) has ameliorated much of the velocity difference in the upper 0 m, that the downhole profiles had suggested (figure 3, thin lines). Model Spectra We compare velocity results from the same sites against each other on the basis of their spectral responses. J. G. Anderson has kindly provided a method of computing the amplitude spectrum at the surface, based on an impulse input to the base of a velocity model. The spectra are entirely linear; and we assume a constant Q=10. Ni et al. (1997) describe this method. At zero frequency, the surface spectral amplitude will have a value of two. Figure 3 (lower) shows the four model surface spectra computed from the four velocity models (upper) from the two borehole sites on the I-580 Extension. We plot spectra only from 3 to 1 Hz, a range over which our shallow velocity data applies and structural engineers have interest as well. The broad spectral peaks within this range result from linear wave resonances within shallow, low-velocity layers in the velocity models. From the I-580 Extension (figure 3) the downhole velocity models (thin lines) have more layers, and the refraction microtremor velocity models (thick lines) have sharper velocity contrasts. The spectra generated from the refraction microtremor models have more prominent peaks of spectral amplification, than the downhole velocities give. Which spectra are correct? The details of the downhole log, subject to point sampling effects in highly heterogeneous rock, control the strength of resonances and their peak frequencies. The details of layer depths and velocities in the refraction microtremor models control the derived resonances as well, but in dispersion-curve modeling velocity trades off against depth (Louie, 001). Since we cannot match spectral peaks between the downhole and refraction microtremor methods, what spectral features can we match? We illustrate what features we can match with spectra computed from velocity models for a soft soil site, the Los Angeles County Fire Station at Newhall, southern California. The Rosrine project of the University of Southern California and Geovision Consultants has collected a shear-velocity log to 107 meters depth with the OYO downhole logging tool. As figure (upper) shows, our refraction microtremor model (thick right-angled line) matches the logged velocities (thick wiggly line) to the 107-m depth of the log. Being a surface survey technique, refraction microtremor cannot match the fine details of the log. If we take the log and the refraction microtremor models only to the 107-m log depth, then the resulting spectra (figure, thick lines, lower plot) match well. Resonance peaks appear at.5 Hz (off the plot) and 3

4 7 Hz, and the spectral amplifications are the same. At Newhall, the deep soil might be more laterally homogeneous than the hard volcanic rock of the I-580 Extension. The 00-m-long refraction microtremor array was also centered just 5 m from the borehole. The fine details of the log that refraction microtremor could not match do not add spectral peaks within the 3-1 Hz range of interest. As discussed in Louie (001), the refraction microtremor dispersions suggested a velocity increase below the 107 m logged depth of the Rosrine borehole. Figure also shows a model putting the velocity increase at the shallowest depth possible, 1 m (thin red line). The corresponding spectrum shows additional peaks of stronger resonance. Such a high velocity at this shallow depth is not geologically reasonable; so we employed basin depths from the SCEC Phase 3 Report (Magistrale et al., 000) to construct a more reasonable model that has shear velocity reaching 500 m/s at 133 m depth. This deepbasin model (figure, green lines) includes the refraction microtremor velocities in the upper 107 m. The model produces an entirely different spectral response, having its principal resonance peaks below 3 Hz. Figure 5 shows spectra computed from velocity models for the hard-rock sites shown in figure. At Keenwild, a hard-rock site in the San Jacinto Mts. of Southern California, the spectral values (figure 5, upper left) all converge towards two below Hz. This convergence suggests that neither the 30-m borehole nor the two refraction analyses are providing velocity data deep enough to constrain spectral response below Hz. Small details from borehole log and shallow low velocities from the refraction microtremor model add disparate resonance peaks above 10 Hz. The velocity gradient in the refraction SeisOpt model (figure ) gives an 8 Hz resonance peak. The upper right plot of figure 5 shows spectra from the Pinon Flat Observatory (PFO), also a hard -rock site in the San Jacinto Mts. As with Keenwild, little data constrain the response below Hz. Small details from the borehole add resonances above 10 Hz; while an ill-constrained layer depth from refraction microtremor has placed a resonance peak at 9 Hz. The refraction SeisOpt model has a broad, muted resonance at 10 Hz. As figure 1 shows, PFO is a highly heterogeneous site. The ability of the refraction SeisOpt models to accurately assess the lateral heterogeneity, and then average through them horizontally, may be producing the most accurate spectral responses. Given any of the direct downhole or refraction microtremor assessments of shear velocity, more accurate shear velocities could be estimated from the refraction SeisOpt P-wave velocity results. With a lower overall shear velocity than was assumed from the Poisson solid, the resonance peak computed at PFO for the refraction SeisOpt model (figure 5, upper right) would have shown a large amplification. On the Yucca Mountain crest, southern Nevada (figure 5, lower left), the refraction microtremor assessment agrees with the downhole log in predicting no significant amplification in the 3-1 Hz band. This prediction results from the models having almost no velocity gradient near the surface (figure ). This site, consisting of evenly layered volcanic tuffs, is laterally homogeneous enough that the borehole was not subject to point-sampling problems. Results of Comparisons Figure 5 (lower right plot) compares spectra from all the sites, mostly using the downhole velocities. The Generic Rock model of Boore and Joyner (1997) is an average of logs, yet its spectral response is most similar to the spectra from the B-C Boundary model of Frankel et al. (199), and Newhall results at that deep soil site. The Keenwild and PFO spectra are very different, and represent the response of truly hard rock. Those spectra are most affected by thin, low-velocity soils in the upper 10 or 0 meters, putting resonances above 10 Hz. We propose from figure 5 that it is important to use refraction microtremor together with refraction to accurately assess the heterogeneity of hard-rock sites. Downhole velocities at such highly heterogeneous sites are not representative of the entire site area. The ability of refraction SeisOpt to directly find lateral heterogeneity is more useful in constructing accurate spectra than is the higher shearvelocity precision of borehole measurements. Velocity models estimated with refraction methods produce similar average velocities versus models from borehole logging (Louie, 001). As the spectra from more or less heterogeneous sites show, spectral resonance peaks may only match over broad ranges of frequency and amplification. Both borehole and refraction methods can certainly separate those sites without significant amplification, such as the Yucca Mountain crest, from those having near-surface velocity gradients that amplify through resonance.

5 Adding geological information to models, as at Newhall, suggests velocities from only the upper 100 meters are not adequate for estimating complete spectra. Acknowledgements Refraction microtremor data collection was supported by the University of Nevada, Reno, Foundation, and by the Southern California Earthquake Center (SCEC). We thank Ted Beeston and Larry Johnson of Black Eagle Consultants for access to the I-580 extension borehole data, and Frank Vernon of U.C. San Diego for access to the Pinon Flat Observatory (PFO) site and its borehole data. The Los Angeles County Flood Control District and the Newhall Fire Station provided access to the Newhall site. Glenn Biasi and Rasool Anooshehpoor of the Nevada Seismological Lab collected refraction data on the Yucca Mountain crest, and provided borehole log data. Shane Smith of the UNR Dept. of Geological Sciences assisted in the collection of microtremor data on the I-580 extension. John G. Anderson of the Nevada Seismological Lab provided his STK5 spectral modeling code. Additional information is available at References Cited Bonilla, L. F., J. H. Steidl, G. T. Lindley, A. G. Tumarkin, and R. J. Archuleta, 1997, Site amplification in the San Fernando Valley, California: Variability of site-effect estimation using the S-wave, coda, and H/V methods: Bull. Seismol. Soc. Amer., 87, Boore, D. M., and W. B. Joyner, 1997, Site amplifications for generic rock sites: Bull. Seismol. Soc. Amer., 87, Bowman, S. D., H. E. Beeston, M. Ashour, and G. Norris, 001, Use of down-hole measurements of shear-wave velocities in analysis and design of drilled shaft foundations in weak rock and cemented soils: in Luke, Jacobeson, and Werle, Eds., Proceedings, 3 th Ann. Symp. on Eng. Geol. and Geotech. Eng., Las Vegas, Mar Field, E. H., and Jacob, K. H., 1995, A comparison and test of various site-response estimation techniques, including three that are not reference-site dependent, Bull. Seismol. Soc. Amer., 85, Frankel, A. D., C. S. Mueller, T. P. Barnhard, D. M. Perkins, E. V. Leyendecker, N. Dickman, S. L. Hanson, and M. G. Hopper, 199, National seismic-hazard maps: documentation June 199, U. S. Geological Survey, Reston, 110 pp. Horike, M., 1985, Inversion of phase velocity of long-period microtremors to the S-wave-velocity structure down to the basement in urbanized areas, J. Phys. Earth, 33, Liu, H. P., Boore, D. M., Joyner, W. B., Oppenheimer, D. H., Warrick, R. E., Zhang, W., Hamilton, J. C., and Brown, L. T, 000, Comparison of phase velocities from array measurements of Rayleigh waves associated with microtremor and results calculated from borehole shear-wave velocity profiles: Bull. Seismol. Soc. Amer., 90, -78. Louie, J. N., 001, Faster, better: shear-wave velocity to 100 meters depth from refraction microtremor arrays: Bull. Seismol. Soc. Amer., 91, Magistrale, H., S. Day, R. W. Clayton, and R. Graves, 000, The SCEC Southern California reference three-dimensional velocity model version : Bull. Seismol. Soc. Amer., 90, S5-S7. Ni, S.-D., R. Siddharthan, and J. G. Anderson, 1997, Characteristics of nonlinear response of deep saturated soil deposits: Bull. Seismol. Soc. Amer., 87, Pullammanappallil, S. K., and J. N. Louie, 199, A generalized simulated-annealing optimization for inversion of first-arrival times: Bull. Seismol. Soc. Amer., 8, Satoh, T., H. Kawase, and S. Matsushima, 001a, Estimation of S-wave velocity structures in and around the Sendai Basin, Japan, using array records of microtremors: Bull. Seismol. Soc. Amer., 91, Satoh, T., H. Kawase, and S. Matsushima, 001b, Differences between site characteristics obtained from microtremors, S-waves, P-waves, and codas: Bull. Seismol. Soc. Amer., 91, Seekins, L. C., L. Wennerberg, L. Margheriti, and H.-P. Liu, 199, Site amplification at five locations in San Francisco, California: A comparison of S waves, codas, and microtremors: Bull. Seismol. Soc. Amer., 8,

6 Steidl, J. H., A. G. Tumarkin, and R. J. Archuleta, 199, What is a reference site?: Bull. Seismol. Soc. Amer., 8,

7 PFOB: Refr. Microtremor p-f Analysis PFOA: Refr. Microtremor p-f Analysis Slowness, sec/meter Dispersion Curve Slowness, sec/meter Dispersion Curve Spectral Ratio over Average Power at All Slownesses Spectral Ratio over Average Power at All Slownesses Elevation, m Line PFOB: SeisOpt Refraction P-Velocity Model Distance, m Elevation, m Line PFOA: SeisOpt Refraction P-Velocity Model Distance, m 000 Velocity, m/s Figure 1: Example velocity analyses by the refraction microtremor and SeisOpt methods, at the Pinon Flat Observatory, southern California. Dark areas slanting down to the right in the refraction microtremor p-f plots show the dispersion of Rayleigh-wave phase velocities to lower values with increasing frequency (Louie, 001). Modeling dispersions with a 1-d shear-velocity profile produces the refraction-microtremor velocity data shown in this paper. The SeisOpt results are -d P-velocity sections produced from first-arrival time picks on hammer refraction records.

8 10 Depth, m B-C Boundary Generic Rock Keenwild Refraction Optim Keenwild Log Keenwild Refraction Microtremor PFO Log PFO Refraction Optim PFO Refraction Microtremor YM Crest N7 Log YM Crest Refraction Microtremor Velocity, m/s Figure : Velocity data resulting from application of three methods at three hard-rock sites in southern California and Nevada. The Keenwild and Pinon Flat (PFO) sites have P-wave refraction SeisOpt results, shown here after horizontal averaging of the -d velocity models and estimation of shear velocity from P velocity assuming a Poisson s ratio of 0.5. These sites as well as the Yucca Mountain, Nevada, crest site have refraction microtremor and downhole logging results. The Generic Rock summary is from Boore and Joyner (1997); the B-C Boundary model is from Frankel et al. (199).

9 10 Depth, m 30 Galena Cr North Refraction Microtremor Galena Cr North Log Galena Cr South Refraction Microtremor Galena Cr South Log Velocity, m/s 000 Surface Amplitude 8 Galena Cr South Log South Refr. Microtr. North Refr. Microtr. Galena Cr North Log Figure 3: Velocity data (upper) and model seismic spectra (lower) at the two abutments of a 300-m concrete arch bridge span under construction for the Interstate Highway 580 extension south of Reno, Nevada. Velocity profiles were estimated from refraction microtremor, and by B. Redpath with downhole seismic surveys for Black Eagle Consultants and the Nevada Dept. of Transportation. The refraction microtremor arrays extended 00 m from each borehole. The lower plot shows seismic amplitude spectra at the surface, through each profile with an impulse incident from below.

10 0 0 Rosrine vs. Refraction Microtremor at Newhall F.S. 0 Depth, m Logged Shear Velocity Logged P Velocity P-Wave Refraction Shallow refr. microtr. Unconstrained refr. microtr. Refr. microtr. plus Phase 3 1 m Unconstrained refr. microtremor 33 m 3 m Model Incorporating Basin Depth from SCEC Phase 3 Report 133 m Velocity, m/s Surface Amplitude Newhall Log Refr. microtr. + Phase 3 Refraction microtremor Shallow refr. microtremor Figure : Velocity profiles (above) and surface seismic spectra (below) for refraction microtremor and OYO logger models of the Newhall Fire Station deep soil site, southern California. Refraction microtremor matched the Rosrine log to the 107-m log depth. Poor microtremor velocity resolution at - Hz frequencies left that model unconstrained below 10 m depth. Removing the unconstrained velocities for a shallow model produces a spectrum very similar to that derived from the log. Creating a geologically realistic model by including deeper velocities in the basin, from the Southern California Earthquake Center (SCEC) Phase 3 Report (Magistrale et al., 000) produces an entirely different spectrum.

11 Surface Amplitude Keenwild Refraction Microtremor Keenwild Refract Optim Keenwild Log Surface Amplitude PFO Log PFO Refraction Microtremor PFO Refract Optim Surface Amplitude YM Crest N7 Log YM Crest Refraction Microtremor Surface Amplitude B-C Boundary Generic Rock Newhall Geological Refr. Microtremor Newhall Log YM Crest N7 Log Keenwild Log PFO Log Figure 5: Surface amplitude spectra for all sites. The hard-rock sites Keenwild, Pinon Flat, and the Yucca Mountain crest show little constraint on spectra below 5 Hz. Refraction microtremor and SeisOpt results agree well with logs that any amplification peaks from shallow resonances will occur above 8 Hz. (All spectra were computed assuming Q=10.) The comparison of all borehole-log spectra, lower right, shows that the Generic Rock model of Boore and Joyner (1997) produces a spectrum more resembling the Newhall and B-C Boundary (Frankel et al., 199) spectra for deep soil sites, with amplification below 5 Hz, than it resembles the true rock spectra such as from Pinon Flat.