- © 2006 by the Seismological Society of America
The Charleston, South Carolina, region is considered to have the highest seismic hazard on the east coast of the United States and has been chosen as a location for an Urban Strong Motion Array of the Advanced National Seismic System (ANSS). This paper will report on subsurface investigations of the shallow shear wave velocity beneath the first four ANSS strong motion stations in Charleston. Both borehole (seismic cone penetration test or SCPT) and surficial seismic (refraction microtremor or ReMi) methods were used to determine shear wave velocities to a depth of 30 m beneath each site. All sites are National Eearthquake Hazards Reduction Program (NEHRP) Class D (VS30 < 360 m/s), with VS30 ranging from 215 to 344 m/sec. Three sites have low-velocity zones within the shallow sedimentary section. The VS30 derived from the SCPT and ReMi techniques agree within 10% at three sites and within 15% at all four sites.
The 31 August 1886 M 6.9–7.3 Charleston earthquake was the largest historical event to strike a well-populated region of the eastern United States (Johnston 1996; Bakun and Hopper 2004). A repeat of the 1886 earthquake could result in as many as 900 deaths, nearly 800 damaged bridges, and more than $14 billion in building damage (URS et al. 2001). Historical evidence for felt earthquakes in the region before the 1886 event (Stover et al. 1984) and continuing low-level seismicity in the epicentral region (e.g., Madabhushi and Talwani 1993), plus evidence for multiple earthquakes producing liquefaction features on the South Carolina coastal plain (Talwani and Schaeffer 2001) result in this region having the second highest seismic hazard in the continental United States east of the Rocky Mountains (Frankel et al. 2002). The high level of seismic hazard, a significant population, and Charleston's continuing importance as a major seaport led it to be selected as one of 25 Urban Strong Motion Locations of the Advanced National Seismic System (ANSS) (USGS 1999). This urban strong motion network is anticipated to contain as many as 60 strong motion stations divided between free-field sites and seismometers in various types of structures (USGS 1999). The ANSS is anticipated to help provide real-time earthquake information for emergency response, on-scale ground-motion recordings for engineering studies of building and site response, and high-quality data for scientific studies of earthquake processes (USGS 2000).
This paper focuses on studies conducted to characterize the shallow shear wave velocity structure in the vicinity of the first four ANSS strong motion installations of the Charleston Urban Strong Motion Array (CUSMA). Many previous studies have shown that amplification and resonance of seismic ground motion by low shear velocity near surface materials is a major contributor to building damage during earthquakes (e.g., Goto et al. 2005). The paper will briefly describe the four CUSMA sites, data collection, processing, and interpretation of refraction microtremor (ReMi) ground motion, followed by an interpretation of the resulting Rayleigh dispersion curves in terms of shallow shear wave velocity structure. I compared these results with shear wave velocity profiles determined using seismic cone penetration tests (SCPT) collected at the same locations, then interpret both results in terms of the average shear wave velocity in the upper 30 m (VS30).
Site characterization of strong motion stations is necessary to understand the contribution of site response to earthquake ground motion. The surface geology in the Charleston region consists of Quaternary nearshore and shallow marine sedimentary facies together with two major types of human-modified landscape: artificial fill and phosphate mining spoil (Weems et al. 1997). Tertiary sediments underlie these units at shallow (<20 m) depths, with stiff semilithified limestones of the Cooper Group providing a significant velocity contrast (Weems and Lewis 2002; Chapman et al. 2006) with the Quaternary.
STATION SITING AND INSTALLATION
The CUSMA installations are located on the Charleston peninsula and in the suburbs north and west of downtown Charleston (figure 1). These sites were chosen to represent the different site conditions based upon surface geology and expected range of ground motions from an event in the Middleton Place Summerville Seismic Zone northwest of Charleston (Madabhushi and Talwani 1993). Two sites (C1SC and C2SC) are located on the campus of The Citadel, the military college of South Carolina. Site ADSC is located on the grounds of the Addlestone Hebrew Academy and site TRSC is at the Trident Research Center a research campus located near the Charleston International Airport and Air Force Base and operated by the South Carolina Research Authority.
The initial CUSMA sites were all intended to be “free-field” seismometers placed at least one building dimension away from the nearest structure (ANSS Technical Integration Committee 2002). Requirements for power and Internet communication, site security, and the presence of roads and sidewalks between potential sites and structures with required hookups, however, limited many of our choices to quasi free-field sites less than a full building dimension away from a structure. Detailed maps of the four individual sites are given in figure 2.
The two sites on the campus of The Citadel are quasi free-field sites; i.e., closer than one building dimension to a structure (figures 2C and 2D). The advantage of these locations is that we were able to place the stations within 500 m of each other on two very different site conditions common to the Charleston region. Station C1SC is located on a sand facies of the Quaternary Wando Formation (figure 1). During construction of the site it became apparent that the upper 0.5 m consists primarily of soil mixed with construction rubble (broken bricks, etc.), with natural deposits directly below. Station C2SC is located upon artificial fill (af) above a former tidal creek channel (figure 1). This fill was excavated during site construction and consists of very soft sand. As much as one third of the inhabited area on the Charleston peninsula consists of land reclaimed from tidal marshes using artificial fill (figure 1).
Station ADSC is also located upon the Quaternary Wando Formation (figure 1). This site is a true free-field installation because it is more than one building dimension from the nearest structure (figure 2B). Station TRSC (figure 2E is located in an area underlain by spoil from phosphate strip mining (ps), which also underlies part of the Charleston International Airport and Air Force Base (figure 1). A considerable part of the northwestern suburbs of Charleston are built upon similar site conditions. This site is quasi free-field, because the distance of the site from the nearest structure is limited by the extent of existing underground electrical conduit.
DATA AND METHODOLOGY
The shear wave velocity profiles presented in this study were obtained using both a borehole (seismic cone penetration test or SCPT) and a surficial seismic (refraction microtremor or ReMi) technique. SCPT data were collected during summer 2005 and refraction microtremor data were collected at various times between late 2003 and early 2005. In the next section, I compare shear wave velocity structures recovered by the two techniques.
Seismic cone penetration testing is an intrusive method of determining shear wave velocities. An SCPT probe is pushed into the ground to the desired depth, and then a shear wave pulse is generated at the surface and recorded by a seismic transducer in the probe (Robertson et al. 1986). Velocities are determined simply by dividing the differential shear wave travel times recorded at various probe depths by the difference in probe depth, taking into account any corrections for source-receiver geometry. This gives very precise estimates of the velocity of the material near the borehole, together with the capability of making velocity estimates over very short depth intervals.
In contrast, the ReMi method uses ambient ground motion recorded by a standard refraction seismometer system with vertical geophones placed up to 10 m apart in a linear array (Louie 2001). Multiple (usually six) 30-s recordings of ambient microtremor ground motion are analyzed. The resulting ground motion recordings are processed into slowness-frequency (p–f) diagrams (figure 3) and Rayleigh dispersion points are picked from these diagrams. It is assumed that the rapid increase in spectral energy at the largest slowness (slowest velocity) for each frequency represents Rayleigh wave energy traveling parallel to the linear dimension of the array (figure 3). The resulting Rayleigh dispersion curve is then interactively modeled by computing the theoretical dispersion curve of a shear wave structure, with the thickness and shear wave velocity of multiple layers adjusted until a satisfactory fit is obtained to the observed dispersion curve (figure 4). For more details of the method, the reader is referred to Louie (2001).
In this study, several students and I recorded microtremor ground motion using 24 4.5-Hz vertical geophones placed in a linear array at intervals of 3 to 9 m, depending upon the available space (figure 2A). This results in line lengths ranging from 69 to 207 m. The main limitation to the depth resolution of the refraction microtremor method is line length, with the depth resolution being limited to approximately one-half the total line length. Therefore 3 m represents a practical minimum geophone spacing required to resolve shear wave velocities to a depth of 30 m. All geophones were placed into the soil flush with the ground, with the exception of several geophones at station C2SC (see below).
At C1SC and C2SC we were able to place the geophone arrays within ∼10 m of the sensor vaults, with the vault approximately at the center of the array (figures 2C and D). At C2SC we encountered a very limited amount of open space to place the geophones and for this reason several geophones had to be deployed on top of a hard-surfaced parking plot (figure 2D). In this case the ground spikes were removed from the geophones and placed upon the hard surface using a semisolid adhesive (“sticky tack”) to hold them in place. In addition, cultural obstructions required that one geophone be placed 1 m perpendicular to the line and another geophone be placed short, creating two irregular intervals (3.7 m and 6.3 m instead of 5 m) along the line. The seismic data from this site was processed using this detailed geometrical information, which improved the resolution of the resulting p–f diagram. However, the difference in coupling properties between geophones placed in the soil and those on the parking lot still prevented a clear determination of the Rayleigh dispersion at higher frequencies. Therefore, we returned to the site and collected microtremor using only 12 sensors placed on soil (western half of seismic line in figure 2D) and processed those data independently.
At ADSC we collected microtremor data in the sports field of the Addlestone Hebrew Academy, with the sensor vault lying south of the east end of the seismic line (figure 2B). At TRSC we initially collected data along an edge of a parking area away from buildings, in anticipation of placing the ANSS sensor vault at either the northwest or northeast corner of the parking area. However, existing underground electrical conduit did not reach those locations, resulting in the sensor vault being placed closer to the building. Therefore we also collected microtremor data along a shorter line in the median between the parking area and a road on the Trident Research Center property, with the sensor vault lying just west of this line (figure 2E).
P-wave refraction data was collected using a sledgehammer source at each site except C1SC. (J. Louie, personal communication 2004) pointed out that modeling the ReMi-derived Rayleigh dispersion curve often gives shear wave velocities that are too high when the P/S velocity ratio is >> 1.73. In this case inclusion of a P-wave velocity structure gives more accurate results. The first arrivals from the P refraction data were interpreted in terms of a simple two-layer model (e.g., Lillie 1999, 75–76), with the top layer interpreted as material above the water table (300 to 570 m/s depending upon the site) and the lower layer as material below the water table (1,250 m/s and higher). For site C1SC, I used the P-wave velocity model of Odum et al. (2003), who collected P- and S-wave refraction data along a line nearly co-located with our microtremor line. Brown et al. (2005) analyzed a larger set of refraction and microtremor data in the Charleston region. In accordance with the suggestion of J. Louie (personal communication, 2004), Brown et al. noted that including a P-velocity model constraint leads to generally lower shear wave velocities. Only the results using the P-velocity model constraint are presented here.
Figure 3 shows p–f diagrams derived from microtremor data collected at sites C1SC and TRSC, together with the Rayleigh dispersion picks. For sites C1SC and C2SC, the slowness of the Rayleigh dispersion decreases continuously as frequency decreases (figure 3A), indicating continuously increasing shear wave velocity with depth. At sites TRSC and ADSC, there is an apparent increase in slowness with decreasing frequency along part of dispersion curve (between 10 and 15 Hz for TRSC, figure 3B), indicative of a low-velocity zone at depth. Note that this is apparent both in the data collected across the parking area (8-m geophone spacing, shown in figure 3B) from the sensor vault and the data collected in the median (3-m geophone spacing) adjacent to the vault (figures 4B and 4C), indicating a low-velocity layer extends across the site in figure 2(E).
Figure 4 shows examples of model fits to the Rayleigh dispersion picks. My preferred models are those fit through the center of the dispersion picks. Once I had a satisfactory fit through the center of the dispersion picks (solid lines, figure 4), I fixed the geometry of the layers and increased/decreased the layer velocities to estimate upper and lower bounds for the velocity (dashed lines, figure 4).
This procedure was the most straightforward when the velocity monotonically increased with period (figure 4A). It was generally impossible to fit all the dispersion picks when a low-velocity layer was present. In this case I adopted a two-step procedure, first estimating the velocities (and their upper/lower bounds) and thicknesses of layers above the low-velocity zone (figure 4B), fixing these values, and then estimating the velocities and thicknesses of the deeper layers (figure 4C). Note that I generally was unable to create a velocity model with a predicted Rayleigh dispersion that ran through all the picks when a low-velocity zone was present.
Figure 5 shows my results for the shear wave velocity structure in the upper 30 m for each site, using both the SCPT and ReMi techniques. At ADSC the two techniques agree quite closely in the upper 25 m (figure 5A), with both detecting a sharp increase in shear wave velocity near 7 m depth. Unfortunately, the SCPT borehole stopped before encountering a low-velocity zone starting at a depth of approximately 27 m, so I am unable to confirm its presence. From the SCPT data I estimate a VS30 of 320 m/s (NEHRP Class D), assuming the velocity at 23 m continues to 30 m. The VS30 from the ReMi shear wave models range from 327 to 359 m/s, slightly higher than the SCPT results.
At C1SC the SCPT data reveal a thin, shallow (∼14 m) low-velocity zone and possibly a deeper one near 30 m (figure 5B). In this case the ReMi technique appears to average through the shallow low-velocity layer and the underlying higher velocity layer, placing a velocity increase near 12 m depth. The S-wave refraction model of Odum et al. (2003), also shown in figure 5(B), clearly detects the next sharp velocity increase around 17–18 m depth. All three techniques give very similar VS30 values, 245 m/s for the SCPT data, 255 m/s for the S-wave refraction of Odum et al. (2003), and 248–260 m/s for the ReMi results, all NEHRP Class D.
At C2SC the SCPT and ReMi velocity models are generally similar, although the ReMi model generally underestimates velocities below 15 m compared with the SCPT results (figure 5C). The SCPT results show velocities greater than 300 m/s at 10 m, but the ReMi-derived velocities do not reach this value until 18 m. The SCPT VS30 is 249 m/s, while the ReMi VS30 ranges from 201 to 229 m/s. This site is also NEHRP Class D.
Site TRSC produces some of the most satisfactory results, in that both the SCPT and ReMi techniques detect a sharp velocity increase at 4 to 5 m and suggest a low-velocity layer starting at 16 m depth (figure 5D). The SCPT VS30 is 325 m/s, while the ReMi VS30 ranges from 278 to 312 m/s. This site is NEHRP Class D like the other CUSMA sites.
It is important to note that the shear wave velocity models I derive using the ReMi technique are not unique. Models are constructed by forward modeling of the Rayleigh dispersion and those constructed by different individuals (or the same individual at different times) can vary somewhat in layer thicknesses and velocity. This was especially the case when a low-velocity zone was present, since I found there can be a tradeoff between the thickness and velocity of the various layers. Figure 6 shows an example of this for site ADSC, which compares the present results with that of earlier work by Jaume (2005). That study estimated a ReMi-derived VS30 of 314 m/s, compared with a VS30 of 327–359 m/s in this work.
For all CUSMA sites the mean VS30 determined using the ReMi method is within 15% of the SCPT VS30, with the difference being less than 10% at three of the four sites (table 1). This is well within the 20% accuracy assumed by Louie (2001). Interestingly, the sites with the longest seismic line lengths (and therefore best depth resolution) give ReMi VS30 estimates all within 10% of the SCPT VS30. The site with the largest discrepancy (C2SC) both has the shortest line lengths and was a site where data collection was most difficult.
All of the CUSMA strong motion sites are NEHRP Class D (VS30 from 360 to 180 m/s), as determined by both the SCPT and ReMi methods. However, there is some evidence that the observed variations in the near-surface shear wave velocities are sufficient to create different site response. Stephens and Jaume (2003) found that an M 4.2 earthquake 42 km south of Charleston produced 50–60% greater peak accelerations at C2SC compared with C1SC. Examining the shear wave velocities at the two sites (figure 5), we find the primary difference between them lies in the thickness and velocity of the surficial layer (C1SC: 200 m/s to 12–18 m depth; C2SC: 100 m/s to 5–7 m depth). Chapman et al. (2006) computed spectral acceleration response ratios for various input ground motions at 52 sites in the Charleston region where SCPT velocities were available. They found that the thickness and velocity of the Quaternary section exerted the most influence on the results, with the greatest variation (up to a factor of three) occurring between 1 and 10 Hz. Observations from the 1886 earthquake show that a higher percentage of wood frame buildings located on artificial fill suffered damage compared with those not on fill (Robinson and Talwani 1983). Taken together, this strongly suggests that enough variation in site response exists within the Charleston region to strongly influence earthquake damage.▪
Undergraduate students S. Cooper, J. Sattler, J. Stephens, and T. Tinker assisted in the seismic data collection and interpretation. Processing and analysis of ambient seismic noise data was conducted using SeisOpt ReMi™ 3.0. T. Cleary, M. Cox, and M. Schuitema of S&ME, Inc. conducted the SCPT data collection and interpretation. Several figures were made using the GMT system (Wessel and Smith 1991). We would like to thank officials at The Citadel, Addlestone Hebrew Academy, and Trident Research Center for allowing us to place ANSS strong motion sites at their facilities and conduct the field data collection for this study. J. Louie and S. Pullammanappalli advised on the use of the ReMi technique. N. Levine and an anonymous reviewer suggested several important improvements for this manuscript. This research was supported by the U.S. Geological Survey, award number 04HQAG0009. The views and conclusions contained in this document are those of the author and should not be interpreted as necessarily representing the official policies, either express or implied, of the U.S. government.
Department of Geology and Environmental Geosciences, College of Charleston