ML21235A106
| ML21235A106 | |
| Person / Time | |
|---|---|
| Issue date: | 08/25/2021 |
| From: | Vladimir Graizer, Dogan Seber, Scott Stovall NRC/RES/DE |
| To: | |
| Graizer V | |
| Shared Package | |
| ML21235A105 | List: |
| References | |
| RIL-2021-06 | |
| Download: ML21235A106 (31) | |
Text
Assessing Site Amplification Variability Using Downhole and Rock-Soil Pairs Site Recordings
Date Published: August 2021
Prepared by:
V. Graizer1 D. Seber2 Scott Stovall1
U.S. Nuclear Regulatory Commission
1 NRC Seismologist 2 NRC Branch Chief
Research Information Le tter Office of Nuclear Regulatory Research Disclaimer
This report was prepared as an account of work sponsored by an agency of the U.S. Government. Neither the U.S. Government nor any agency the reof, nor any employee, makes any warranty, expressed or implied, or assumes any legal liability or responsibility for any third partys use, or the results of such use, of any infor mation, apparatus, product, or process disclosed in this publication, or represents that its u se by such third party complies with applicable law.
This report does not contain or imply legally binding requireme nts. Nor does this report establish or modify any regulatory guidance or positions of the U.S. Nuclear Regulatory Commission. This report is not binding on the Commission.
Page intentionally left blank EXECUTIVE
SUMMARY
For nuclear power plants licensed after January 10, 1997, Title 10 of the Code of Federal Regulations (10 CFR) Part 50, Domestic licensing of production and utiliz ation facilities, and 10 CFR 100.23, Geologic and seismic siting criteria, establis h the seismic design basis.
Appendix S, Earthquake Engineering Criteria for Nuclear Power Plants, to 10 CFR Part 50 defines the safe-shutdown earthquake: Safe-shutdown earthquake ground motion is the vibratory ground motion for which certain structures, systems, and components must be designed to remain functional. The regulation in 10 CFR 100.23 requires that the applicant determine the safe-shutdown earthquake and its uncertainty. A p robabilistic seismic hazard assessment (PSHA) is an acceptable method to capture uncertaint y.
Traditionally, ground motion models and PSHAs are developed bas ed on an ergodic assumption, with a broad range of uncertainties. An ergodic process is a random process in which the distribution of a random variable in space is the sam e as the distribution of that same random variable at a single point when sampled as a function of time (Anderson and Brune, 1999). An ergodic assumption is made when a PSHA treats that spatial uncertainty of ground motions as an uncertainty over time at a single point. T his usually results in overly conservative values in hazard calculations. Recently, the pract ice has trended towards nonergodic PSHAs, when additional information about the site of interest is available. It has been previously shown that, with a sufficient number of earthqu ake recordings, the mean site amplification functions (SAFs) can be well determined, but indi vidual event ratios can be quite variable. From this point of view, it is important to assess an d quantify observed variabilities in SAFs obtained from earthquake recordings.
This report assesses site-specific variability in empirical SAF s calculated using earthquake recordings with a total of 13 datasets, including data from the six California strong motion downhole arrays at Treasure Island, Turkey Flat, San Francisco Bay Bridge, Crockett Carquinez Bridge, Corona Bridge, and Garner Valley and also three soil-rock pair stations. The two-and three-dimensional effects, which include out-of-plane reflection/refraction, focusing, scattering, and conversion of wave types, produce sig nificant variability in empirical SAFs from earthquake data recorded at a single station. This an alysis demonstrates that log-natural standard error ln(f) (sigma) in empirical SAFs calculated using downhole array data can be approximated by a linear function with an average value of 0.221 (1.25 times) in the frequency range of data processing. By using a constant in the frequency range of 0.1-10 hertz, ln(f) can be well approximated and is sli ghtly increasing at higher frequencies for rock-soil pairs. Because of spatial variability, the rock-soil pairs sigm a is higher than that of downhole arrays with an average of 0.272 (1.31 times). Variability in em pirical SAFs helps constrain single-station, nonergodic sigma estimates.
Keywords: Arrays Seismic Motion, Rock and Soil Motions, Site Amplification Variability
INTRODUCTION
For nuclear power plants licensed after January 10, 1997, Title 10 of the Code of Federal Regulations (10 CFR) Part 50, Domestic licensing of production and utiliz ation facilities, and 10 CFR 100.23, Geologic and seismic siting criteria, establis h the seismic design basis.
Appendix S, Earthquake Engineering Criteria for Nuclear Power Plants, to 10 CFR Part 50 defines the safe-shutdown earthquake: Safe-shutdown earthquake ground motion is the vibratory ground motion for which certain structures, systems, and components must be designed to remain functional. The regulation in 10 CFR 100.23 requires that the applicant determine the safe-shutdown earthquake and its uncertainty. A p robabilistic seismic hazard assessment (PSHA) is an acceptable method to capture uncertaint y.
Traditionally, ground motion models and PSHAs are developed bas ed on an ergodic assumption, with a broad range of uncertainties. An ergodic pro cess is a random process in which the distribution of a random variable in space is the sam e as the distribution of that same random variable at a single point when sampled as a function of time (Anderson and Brune, 1999). An ergodic assumption is made when a PSHA treats that spatial uncertainty of ground motions as an uncertainty over time at a single point. T his usually results in overly conservative values in hazard calculations. Recently, the pract ice has trended towards nonergodic PSHAs, when additional information about the site of interest is available. It has been previously shown that, with a sufficient number of earthqu ake recordings, the mean site amplification functions (SAFs) can be well determined, but indi vidual event ratios can be quite variable (e.g., Field et al., 1992; Boore, 2004). From this poi nt of view, it is important to assess and quantify observed variabilities in SAFs obtained from earth quake recordings.
This report assesses site-specific variability in empirical SAF s observed using recorded earthquake data from the six California strong motion downhole arrays at Treasure Island, Turkey Flat, San Francisco Bay Bridge, Crockett Carquinez Bridg e, Corona Bridge, and Garner Valley and also three rock-soil pairs of stations. These downhole arrays have at least one sensor in rock site condition. It also considers 13 alterna tive scenarios, giving the most attention to the data from the Treasure Island and Turkey Flat arrays because these sites have been studied extensively and have detailed geological and geote chnical information. Presented uncertainties in SAFs are caused by randomness of earthquake lo cations and magnitudes and can be considered to be aleatory.
ANALYSIS OF THE TREASURE ISLAND DOWNHOLE ARRAY RECORDS
Treasure Island (TI) is an artificial island in San Francisco B ay in California. Constructed in the 1930s for the 1939 Golden Gate International Exposition (Figure 1, left panel), the island was created by fine-to-medium-grained sand dragged from San Francis co Bay and used as a fill material. The fill material was deposited hydraulically by a cl amshell dredge. Below the fill and native sand is a layer of bay mud composed of silty clay with regions of sand and silt. The older bay sediments of Pleistocene age are generally stiff to sandy, silty, or peaty clays that extend down to Franciscan bedrock (Figure 1, right panel) (de Alba et al., 1994).
This report analyzes high-quality, low-amplitude earthquake dat a recorded at TI and Yerba Buena Island (YBI) near San Francisco, CA. Most publications de scribe the YBI geology as rock (e.g., Darragh and Shakal, 1991). TI is a manmade island situat ed between San Francisco and Oakland, CA, and attached to the natural YBI by a short causewa y (Figure 1, left panel). Many places on TI experienced liquefaction during the moment magnitu de (M) 6.9 Loma Prieta earthquake of 1989 (Ferritto, 1992). Following this earthquake, the California Geological Survey (CGS), with support from the National Science Foundation, insta lled the TI array in 1992 to study the response of a soft soil over rock geologic structure to earthquake motion (Darragh et al., 1993; de Alba et al., 1994; Graizer, 2014). The TI down hole array had sensors located in the bedrock (104-and 122-meter (m) depths), alluvium (31-and 44-m depths), and artificial fill (7-and 16-m depths) and at the surface (shown with triangles i n Figure 1, right panel). The downhole instrument at 122-m depth was added in 1996. In Septem ber 2003, the original digital 12-and 16-bit instrumentation was replaced with modern 19-bit instruments (Graizer and Shakal, 2004).
Figure 1 (right panel) shows the velocity and geology profiles at the TI array. Weathered Franciscan shale and sandstone are encountered at 88 m beneath the site, with more competent sandstone found at a depth of about 98 m (Darragh and Idriss, 1997). The U.S. Geological Survey performed original downhole S-wave velocity measurements in the 104-m deep hole (Gibbs et al., 1992). More recent S-wave velocity averaging was performed based on the P-S suspension logging measurements conducted in the deepest 122-m borehole drilled in 1996 (Graizer and Shakal, 2004). S-wave velocities vary from ~134 meters per second (m/s) in the gray fine sand layer to ~2,523 m/s in the deepest Franciscan bedrock (Figure 1, right panel).
The authors downloaded 26 processed earthquake records from the Center for Engineering Strong Motion Data (CESMD) https://strongmotioncenter.org/) recorded by the CGS Station Treasure IslandGeotechnical Array. All recordings are low-ampl itude ground motions with maximum peak ground acceleration (PGA) at the surface of 0.038 g (Table 1). In contrast to previous studies of TI recordings (Graizer, 2014), all except t he local magnitude ML 5.0 1999 Bolinas earthquake were recorded by the array after September 2 003 with the newly installed modern 19-bit instruments that replaced the original 12-and 16 -bit instrumentation (Graizer and Shakal, 2004). Most of the older records were not used in order to avoid problems with low signal-to-noise ratios especially affecting bedrock downhole re cordings because of the lower resolution of older equipment. First, the authors studied ampli fication of ground motions from the deepest sensor in Franciscan bedrock at the depth of 122 m to t he surface (Figure 1, right panel) by comparing 5-percent damped response spectral accelera tions of earthquake recordings. At the second stage, the analysis examined amplific ation from bedrock recordings at a depth of 104 m to the surface.
Figures 2 and 3 demonstrate examples of recorded accelerations and calculated displacements at different depths representing the two different types of rec ordings: (1) relatively simple signals dominated by S-waves (Figure 2) and (2) signals dominated by surface waves (Fi gure 3).
Figure 4 demonstrates response spectral ratios of motions at th e surface relative to the bedrock at 122-m depth. As is typical in earthquake engineering practic e, individual earthquake SAFs at each frequency are calculated as a ratio of geometric means of the two horizontal components oriented at 90 and 360 degrees:
,, )
= (,(,,, ) (1)
The main peak in the response spectral site amplification is ob served at ~0.8 hertz (Hz), and the second peak is at ~1.75 Hz (Figure 4). As shown by Haskell (196 0), for a vertically incident SH-wave on a plane layer having a shear wave velocity VS and a thickness h, mechanical resonances occur at frequencies fn (quarter wavelength approximation):
fn=(2n+1)xVS /4h (2)
The lowest frequency (first mode) can be associated with the al luvium-bedrock interface at the 88-m depth characterized by the S-wave velocity increase from 386 to 1,230 m/s (Figure 1, right panel). Using average S-wave velocity of VS ~267 m/s in the upper layer of thickness h = 88 m results in the resonance frequency of 0.76 Hz, which is close t o the average empirical value of 0.8 Hz. The second peak in SAF at 1.75 Hz can be associated wit h the bay mud-alluvium interface at the depth of 28.8 m, characterized by a significan t S-wave velocity increase from 176 to 317 m/s.
SAFs shown in Figure 4 (upper panel) demonstrate significant va riations in amplitudes of the first peak at ~0.8 Hz from a factor of ~4 to ~13 for individual events. These SAFs can be split into the two different groups depending on the amplitudes of th e first ~0.8-Hz peak: the lower amplification group (LAG) events (maximum peak averages ~4.9) ( Figure 4, middle panel) and the higher amplification group (HAG) events (maximum peak avera ges ~9.0) (Figure 4, lower panel). The first group (LAG) includes 14 events, and the secon d one (HAG) includes 12 events (Table 1). Events that produce lower amplitude spectral ratios are mostly close-by earthquakes with dominant direct S-waves and relatively low-amplitude surface waves (Figure 2).
Earthquakes producing relatively higher amplitude response spec tral ratios are more distant earthquakes with larger amplitude surface waves compared to S-wave amplitudes (Figure 3).
Figure 5 demonstrates mean +/-1 standard deviation response SAFs for all 26 surface/122-m depth records and also for the LAG and HAG. As expected, the gr oup that includes all events demonstrates higher log-natural sigma ln(f) than that of the low SAF events. In all cases, average sigmas are almost flat with only slight change at frequ encies higher than 30 Hz.
At the second stage, the authors calculated SAFs for 11 surface /104-m depth available records.
Unfortunately, after 2007, the three channels in the 104-m deep downhole died and did not record earthquakes. SAFs shown in Figure 6 demonstrate first an d second peaks at the same frequencies of 0.8 and 1.8 Hz and also significant variations i n amplitudes of the first peak in individual events. This group of recordings demonstrates practi cally flat log-natural sigma ln(f),
very similar to that of 122-m downhole up to the frequency of 3 0 Hz. In the frequency range of data processing of 0.3-40 Hz, average sigma is slightly higher for the surface/104-m case (0.202 versus 0.189).
TREASURE ISLAND AND YERBA BUENA ISLANDS GROUND MOTIONS
In the next series of tests, the analysis compares YBI surface recordings at two nearby stations with the response of the TI array surface data (downhole horizo ntal Channels 1 and 3 of CGS Station 58642). The stations are (1) YBI CGS Station 58163 and (2) YBI CYB USGS-NCSN station (Table 2).
TI is connected by a small isthmus to YBI with the distance bet ween those sites of 2.20 and 2.25 kilometers (km) (Figure 7). Most publications describe the geology of YBI as rock (Franciscan formation with a mix of sandstone, limestone metamo rphic, and other rocks)
(Darragh and Shakal, 1991; Darragh and Idriss, 1997; Baise et a l., 2003), while the CESMD Web site (https://strongmotioncenter.org/) estimates VS30 at 660 m/s and considers it to be Class C. Liquefaction occurred at TI during the 1989 M 6.9 Loma Prieta earthquake (Ferritto, 1992; de Alba et al., 1994).
The authors downloaded YBI data from the CESMD for seven earthq uakes that were also recorded by the TI array surface channels. Figure 8 (upper pane l) demonstrates horizontal component response spectral ratios between the TI surface and Y BI. The data were split into two groups: five records from earthquakes with M4.5 and all seven events including the M 6.9 Loma Prieta and the South Napa M 6.0 earthquakes. Because of liquefaction at TI, the Loma Prieta record is affected by the nonlinearity with a shift of peak SAF toward low frequencies. As expected, ln(f) is lower for the first group of smaller events not affected by nonlinearity. The two main peaks in site amplifications are at the frequencies of ~0.8 and 1.8 Hz, demonstrating similarity to those of the TI downhole ar ray shown in Figures 4 through 6.
A number of previous studies (e.g., Darragh and Shakal, 1991) h ave used rock ground motions at YBI as a reference site to estimate the site response at TI applying a one-dimensional equivalent linear approach. Baise et al. (2003) demonstrated th at a two-dimensional basin structure is needed to analyze TI site response. A comparison o f SAFs from the YBI and TI bedrock 122-m depth recordings in Figure 8 (lower panel) demons trates an average amplification of 3.5 times at 10-20 Hz at the YBI relative to T I bedrock, while at frequencies higher than 30 Hz, the difference is about 2.35 (close to the t heoretical one of ~2.0) between the media motion relative to the outcrop. These comparisons show th at the YBI site is not an ideal rock site input for the TI surface recordings.
TURKEY FLAT ARRAY RECORDINGS
The CGS established the Turkey Flat Site Effects International Test Area in 1987, in a shallow valley at Turkey Flat, located 8 km southeast of the town of Pa rkfield and about 5 km east of the San Andreas Fault in central California. The array was intended to provide data with which to investigate the accuracy and consistency of current methods for estimating the effects of site conditions on ground surface motions (Real and Tucker, 1988). F igure 9 shows the location of the Turkey Flat strong motion array and epicenters of the eight earthquakes used in the prediction exercise provided by the CGS and two additional even ts (9 and 10), shown in Table 3, downloaded from the CESMD at https://strongmotioncenter.org/.
The Turkey Flat array is located in a northwest-trending valley within the central California Coastal Range. The valley is filled with a relatively thin laye r of stiff alluvial sediments with basement rock outcrops at the south and north ends of the valle y (Figure 10, left panel). The valley is about 6.5 km long and 1.6 km wide and is aligned with the southwest plunging Parkfield syncline (Real and Tucker, 1988). Turkey Flat was chosen as a t est area to begin with a geologically simple site where a moderate event producing stron g motion is expected before moving to more complicated sites.
The Turkey Flat array includes four recording sites: Rock South or Turkey Flat #1 (TF#1) (R1 in Figure 10), Valley Center or Turkey Flat #2 (TF#2) (V1), Valley North or Turkey Flat #3 (TF#3)
(V2), and Rock North or Turkey Flat #4 (TF#4) (R2). Surface ins truments were installed at each of these sites, and downhole instruments were installed at the Rock South (R1) and Valley Center (V1) sites. Downhole instrument D1 was located at a dept h of approximately 24 m at the Rock South site, and downhole instruments D2 and D3 were locate d at depths of 10 m and 23.5 m, respectively, at the Valley Center. Valley Center instr ument D3 (23.5-m depth) was located about 1 m below the soil/rock boundary. Each instrument location included a three-component force-balance accelerometer and a velocity transducer with 12-bit solid-state digital recording. Unfortunately, the downhole E-W oriented sensor at T F#1 did not work. For reference, the distance from V1 to V2 is about 510 m, from V1 t o R1 about 850 m, and from V2 to R2 550 m. The distance between R1 and R2 is about 1,600 m (F igure 10, left panel).
In 1987-1988, multiple investigation teams, both domestic and a broad, carried out a comprehensive program of site characterization. The teams condu cted a broad range of field and laboratory geophysical and geotechnical tests. Eight boreho les were drilled through valley sediments into the underlying basement rocks, in which in situ testing was performed and rock and sediment samples were acquired for laboratory analysis. Shear-wave velocity (Figure 10, right panel) was measured in boreholes at the two vertical arra ys using downhole, crosshole, and suspension logging methods performed by numerous groups, in cluding LeRoy Crandall and Associates, Hardin Lawson Associates, QEST Consultants, OYO Cor poration, Kajima Corporation, the California Division of Mines and Geology (rena med California Geological Survey in 2006), and Woodward-Clyde Consultants (Real et al., 2 008; Haddadi et al., 2008; Real and Tucker, 1988).
All records used for this analysis except for the M 6.0 Parkfield 2004 earthquake are low-amplitude ground motions indicative of mostly linear site a mplifications (Table 3). The Parkfield earthquake was recorded at an epicentral distance of less than 10 km with PGA up to 0.3 g at the TF#2 (Valley Center) site. Usually, amplitudes hig her than 0.2-0.3 g are considered to be the level where nonlinearity effects start. According to the results compiled by Kramer (2009), the site responded essentially linearly in the 2004 Parkfield e vent.
One of the purposes of establishing the Turkey Flat array and a n international test area was to perform a blind test experiment similar to that done in typical construction projects. In the first phase of the blind test experiment, participants were provided with all available subsurface data and the recorded R1 rock motions and asked to predict the respo nse of the Valley Center V1 soil motion 850 m apart. In the second phase, which did not beg in until all first-phase predictions had been received, participants were provided with the D3 motio ns and asked to predict the D2 and V1 motions. A number of papers describe the results of the experiment (e.g., Kwok et al.,
2008; Kramer, 2009).
All the strong motion earthquake records except for Parkfield m ain shock were processed in the frequency range of 0.3-40 Hz. The M 6.0 Parkfield record was processed in the frequency range of 0.125-40 Hz.
This report considers empirical SAFs. Site amplifications were calculated for the following four cases:
(1) Valley Center downhole TF#2 bedrock (23.5-m depth) to surfa ce (Figure 11)
(2) Rock South (TF#1) to Valley Center (TF#2) surface motions ( Figure 12)
(3) Rock North (TF#4) to Valley North (TF#3) (Figure 13)
(4) Combined rock (Rock South + Rock North) to combined soil (V alley Center + Valley North) (Figure 14)
Empirical SAFs demonstrate significant variabilities for different events (Figures 11 through 14),
with the average log-natural sigma varying in the range from ln=0.262 to ln=0.340 with the highest variability observed in the Rock South-Valley Center ou tcrops (Figure 12), and the lowest for Rock North-Valley North outcrops (Table 4 and Figure 13).
The transfer function V1 (TF#2 surface) - R1 (TF#1 surface) has the highest natural variability of ln=0.340 between analyzed pairs. Natural variability in SAFs betw een TF#2 downhole bedrock (23.5 m deep) and surface motion is relatively high ( ln=0.288), considering the relatively simple Turkey Flat geologic structure. Tsai et al. ( 2017) suggested that the two-dimensional effect influenced the response at sites located near the edge of the basin and makes SAFs dependent on wave propagation paths.
ADDITIONAL DOWNHOLE DATA AND
SUMMARY
RESULTS
This study also included two sets of 16 earthquakes recorded at the Garner Valley downhole array (GVDA) at the surface and at the depths of 50 m ( VS=580 m/s) and 150 m (VS=3,000 m/s).
GVDA records were downloaded from http://nees.ucsb.edu/data-portal. Appendix A (Figure A-1) shows the GVDA sensor locations and P-and S-wave velocity profiles. GVDA ln demonstrates the same behavior as others outlined above: it is almost flat f rom 0.3 to 40 Hz (Figure 15).
Additionally, the authors analyzed earthquake recordings from t he three CGS instrumented strong-motion downhole arrays: San Francisco Bay Bridge with th e deepest sensor at 39.9 m and corresponding VS ~2,000 m/s (10 earthquakes), Crockett Carquinez Bridge #2 with the deepest sensor at 125 m and VS >1,000 m/s (8 earthquakes), and Corona-I15/Hwy 91 (11 earthquakes) with the deepest sensor at 41.8 m and VS ~2,000-3,000 m/s. Data were downloaded from the CESMD at https://strongmotioncenter.org/. Appendix A (Figures A-2, A-3, and A-4) shows schematics of downhole locations and P-and S-wave velocity profiles.
Figure 15 and Table 4 show a compilation of all data from downh ole arrays and an average sigma. In almost all cases, log-natural sigma can be approximat ed by linear dependence with a low slope. Average logarithmic sigma is practically flat in the frequency range 0.1 to 100 Hz.
All this can be considered as aleatory variability and also as the lowest end of nonergodic site-specific sigma. In average log-natural sigma, ln(f) from downhole array data varies:
ln(f)= 0.0000005xf+0.2351 (3)
In the data processing frequency range of 0.3-40 Hz, standard e rror ln(f) in empirical site amplification has an average value of 0.221 (1.247 times the me an value of SAF).
This variability in site amplification is due to variations in site-to-source azimuths, wave propagation paths and wave types, magnitudes of earthquakes, an d nonlinearity effects modifying amplitudes of ground motions at the site. All data co nsidered in this paper except for the moment magnitude MW 6.9 Loma Prieta 1989 and the MW 6.0 Parkfield 2004 earthquakes are low-amplitude recordings with no nonlinearity effects.
For soil-rock pairs, the average sigma ln(f) can be approximated as follows:
ln(f)= 0.0017xf+0.2525 (4)
As expected, because of spatial variability the soil-rock pair sigma is higher than that of the downhole arrays with an average of 0.272 (1.31 times the mean v alue of SAF) in the 0.3-40 Hz frequency range (Figure 16). The increase of sigma at high freq uencies can be explained by (1) increased randomness at frequencies 10 Hz, (2) variation o f incident wave angles, (3) variations in media resulting from different paths, and (4) increase of instrumental noise at higher frequencies.
The results of this study agree with Boores (2004) observation that variability in ground motions is large, making it difficult to accurately predict site-and e arthquake-specific response, and with its recommendation to concentrate on predicting mean amplificat ions for many events.
CONCLUSIONS
This analysis assessed site-specific variability in empirical S AFs calculated using earthquake recordings of a total of 13 datasets, including data from the s ix California strong motion downhole arrays at Treasure Island, Turkey Flat, San Francisco Bay Bridge, Crockett Carquinez Bridge, Corona Bridge, and Garner Valley and also three soil-rock pairs of stations.
In the frequency range of 0.1-100 Hz, log-natural standard erro r ln(f) in empirical downhole site amplification can be approximated by a linear function with an average value of 0.221 (1.25 times the mean value of SAF). For frequencies 10 Hz ln(f), function is practically flat and starts increasing at higher frequencies for soil-rock pairs. Th e lowest sigma corresponds to the downhole amplification associated with the vertically propagati ng S-waves, while the highest sigma is associated with a rock-soil TF#1 and TF#2 pair where r ock station is located near the edge of the two-dimensional basin. As expected, because of spat ial variability in soil, the rock pairs sigmas on average are higher (0.272) than those of downho le arrays (average of 0.221)
(Figure 16).
The two-and three-dimensional effects, which include lateral r efraction, focusing, scattering, and conversion of wave types, produce significant variability i n empirical SAFs from earthquake data recorded at a single station and help constrain single-sta tion, nonergodic sigma estimates.
The results also give insights into the accuracy that can be ac hieved in site response predictions. Standard error estimates based on downhole array d ata represent aleatory variability and can also be considered as the lowest end of non ergodic site-specific sigma.
DATA AND RESOURCES
GVDA data and information were downloaded from http://nees.ucsb.edu/data-portal. All other earthquake records and downhole information were downloaded fro m the CESMD at https://strongmotioncenter.org/.
ACKNOWLEDGMENTS
We appreciate help from Hamid Haddadi and David Branum in provi ding additional strong motion data and supporting information about California Geologi cal Survey stations. We thank J.P. Stewart for sharing Turkey Flat geotechnical profiles data.
REFERENCES
Anderson, J., and J. Brune (1999). Probabilistic seismic hazar d assessment without the ergodic assumption, Seism. Res. Lett. 70, 19-28.
Baise L.G, S.D. Glaser, and D. Dreger (2003). Site response at Treasure and Yerba Buena Islands, California, Journ. Geotechnical and Geoenvironmental Engineering. 129 (5):415-426.
Boore, D.M. (2004). Can site response be predicted? Journ. Earthquake Engineering. 8, Special Issue 1. 1-41.
Darragh, R.B., and I.M. Idriss (1997). A Tale of Two Sites: Gi lroy #2 and Treasure Island Site Response Using an Equivalent Linear Technique, Earthquake Engi neering Research Institute, Oakland, CA.
Darragh, R.B., and A.F. Shakal (1991). The site response of t wo rock and soil station pairs to strong and weak ground motion, Bull. Seismol. Soc. Am., 81, 1885-1899. Darragh, R., M.
Huang and A. Shakal (1993). "Processed CSMIP strong-motion data from the Treasure Island geotechnical array from the Gilroy area earthquake of January 1 6, 1993." Calif. Div. Mines and Geology, OSMS 93-09, 37 p.
de Alba, P., J. Benoit, D.G. Pass, J.L. Carter, T.L. Youd, and A.F. Shakal (1994). Deep instrumentation array at Treasure Island Naval Station, U.S. Geol. Surv. Prof. Pap. 1551-A, 155-168.
Ferritto, J.M. (1992). Ground motion amplification and seismic liquefaction: A study of Treasure Island and the Loma Prieta earthquake, Naval Civil Engineering Laboratory (NCEL) Technical Note 92-21874, Port Hueneme, CA.
Field, E.H., K.H. Jacob, and S.E. Hough (1992). Earthquake sit e response estimation: A weak-motion case study, Bull. Seism. Soc. Am. 82, 2283-2307.
Gibbs, J.F., T.E. Fumal, D.M. Boore, and W.B. Joyner (1992). S eismic velocities and geological logs from borehole measurements at seven strong-motion stations that recorded the Loma Prieta earthquake, U.S. Geological Survey Open-File Repor t 92-287, Menlo Park, CA.
Graizer, V., and A.F. Shakal (2004). Analysis of CSMIP strong-motion geotechnical array recordings, in P. de Alba, R.L. Nigbor, J.H. Steidl, and J.C. Stepp, editors, In Proceedings:
International Workshop for Site Selection, Installation, and Operation of Geotechnical Strong-Motion Arrays, Richmond, CA, 2004. Consortium of Organizations for Strong Motion Observation Systems (COSMOS).
Graizer, V. (2014). Comment on Comparison of time series and random-vibration theory site-response methods by Albert R. Kottke and Ellen M. Rathje, Bull. Seismol. Soc. Am., 104, 540-546.
Haddadi, H., A.F. Shakal, and C.R. Real (2008). The Turkey Fla t Blind Prediction Experiment for the September 28, 2004 Parkfield Earthquake: Comparison wit h Other Turkey Flat Recordings, Geotechnical Earthquake Engineering and Soil Dynamics IV Congress 2008,
pp. 1-10.
Haskell, N. (1960). Crustal reflection of plane SH waves, J. Geophys. Res. 65, 4147-4450.
Kramer, S.L. (2009). Analysis of Turkey Flat ground motion prediction experimentLe ssons learned and implication for practice, Proc. SMIP09 Seminar on Utilization of Strong-Motion Data, Oakland, CA, pp.1-22.
Kwok, A.O.L., J.P. Stewart, and Y.M.A. Hashash (2008). Nonline ar ground-response analysis of Turkey flat shallow stiff-soil site to strong ground motion, Bull. Seismol. Soc. Am. 98, 331-343.
Real, C.R. and B.E. Tucker (1988). Ground Motion Site Effects Test Area Near Parkfield, California, Proceedings Ninth World Conference on Earthquake Engineering, Tokyo, Japan VIII, pp. 187-191.
Real, C.R., A.F. Shakal, and B.E. Tucker (2006). Turkey Flat, U.S.A. site effects test area:
Anatomy of a blind ground-motion prediction test, Proceedings, 3rd International Symposium Effects of Surface Geology on Seismic Motion, 29 August-1 September 2006, Grenoble, France, Paper KN 3, pp. 1-19.
Real, C., A. Shakal, and B. Tucker (2008). The Turkey Flat Bli nd Prediction Experiment for the September 28, 2004 Earthquake: General Overview and Models Test ed, Geotechnical Earthquake Engineering and Soil Dynamics IV Congress 2008, pp. 1-10.
Tsai, C.-C., W.-S. Chang, and J.-S. Chiou (2017). Enhancing Pr ediction of Ground Response at the Turkey Flat Geotechnical Array, Bull. Seismol. Soc. Am. 107, 2043-2054.
U.S. Code of Federal Regulations, Reactor Site Criteria, Part 100, Chapter I, Title 10, Energy.
U.S. Code of Federal Regulations, Domestic Licensing of Production and Utilization Facilities Part 50, Chapter I, Title 10, Energy.
Figure 1. View of TI and YBI (left panel) and P-and S-wave velocities and soil profile at TI array (right panel). Triangles show locations of seismic instru ments.
Figure 2. Shear waves dominated record of the M L 3.4 earthquake with the epicenter at Berkeley, CA, at a distance of 12.5 km (Table 1): acceleration (left panel) and displacement (right panel).
Figure 3. Surface waves dominated record of the M W 4.2 earthquake with the epicenter at Piedmont, CA, at a distance of 16.5 km (Table 1): acceleration (left panel) and displacement (right panel).
Figure 4. Response SAFs from bedrock at 122-m depth to the surf ace at the TI array:
records from all 26 events (upper panel), records from 14 LAG e vents (middle panel), and records from 12 HAG events (lower panel).
Rat_1 Treasure Island Spectral Amplification Functions Surf/122 m Rat_2 Rat_3 Rat_1 Rat_2 10 Rat_3 Rat_4 Rat_5 Rat_6 Rat_7 Rat_8 Rat_9 Rat_10 Rat_11 1 Rat_12 Rat_13 Rat_14 Rat_15 y = -0.0005x + 0.2338 High Rat_16 Rat_17 Rat_18 Rat_19 Rat_20 Rat_21 0.1 Rat_22 Rat_23 y = -0.0009x + 0.2218 All y = -4E-05x + 0.1415 Low Rat_24 Rat_25 Rat_26 Mean_all STD_All Mean+1std Mean-1std 0.01 Mean_low STD_Low 0.1 1 10 100Frequency, HzMeanL+1std MeanL-1std
Figure 5. Response spectral SAFs and natural logarithmic standa rd deviations of SAFs for the TI downhole array surface/122-m depth.
Treasure Island Spectral Amplification Functions Surf/104 m Rat_1 10 Rat_2
Rat_3
Rat_4
Rat_5
Rat_6 1 Rat_7
Rat_8 y = -0.0002x + 0.2242 Surf/104 m Rat_9
Rat_10
Rat_11 0.1 y = -0.0009x + 0.2218 Surf/122 m 104_Mean_SAF
104_Mean-1std
104_Mean+1std
104_STD
122_STD
Linear (104_STD) 0.01 0.1 1 10 100Linear (122_STD)
Frequency, Hz
Figure 6. Response SAFs and natural logarithmic standard deviat ions of SAFs for the TI downhole array surface/104-m depth.
Figure 7. Map of TI and YBI stations. 58117 is the old TI Fire Station, and 58163 is the YBI Station. Those two stations recorded the 1989 M 6.9 Loma Pr ieta earthquake.
Treasure - Yerba Island SAF Berkeley_2011_NCCYB 10 Berkeley_2011_CSMIP
ElCerrito_2012_CSMIP Berkeley_2018_CSMIP Berkeley_2018_NCCYB Mean_low Mean+1std 1 Mean-1std
LomaPrieta_1989 y = 0.0025x + 0.2735 all Napa_2014_NCCYB Log_std_low Mean_all Mean+1std 0.1 y = 0.0014x + 0.1792 low Mean-1std
Log_std_all Linear (Log_std_low)
Linear (Log_std_all)
0.01 0.1 1 10 100Frequency, Hz
10 Ratio Yerba Buena to TI 122 m depth
1 y = 0.0037x + 0.1577
Berkeley_2011_NCCYB 0.1 Berkeley_2011_CSMIP ElCerrito_2012_CSMIP Napa_2014_NCCYB Geomean Log_Std Mean+1sigma Mean-1sigma Linear (Log_Std) 0.01 0.1 1 10 100Frequency, Hz
Figure 8. Response spectral SAFs and natural logarithmic standa rd deviations of SAFs for the TI surface to YBI (upper panel), and comparison of SAFs from the YBI to TI bedrock 122-m depth recordings (lower panel).
Figure 9. Map of earthquakes and Turkey Flat strong motion stat ions (modified from Haddadi et al., 2008).
Figure 10. Oblique aerial view of Turkey Flat strong-motion arr ay (Real et al., 2008) (left panel) and S-wave velocity profiles at mid-valley site (V1-D3 array) from Real et al., 2006 (right panel).
10 Turkey Flat #2 Surface/Downhole 23.5 m SAF
Ratio_1 1 Ratio_2 Ratio_3 Ratio_4 Ratio_5 Ratio_6 Ratio_7 Ratio_8 Ratio_9 0.1 y = 0.0034x + 0.2489 Ratio_10 Mean_Ratio Mean+1std Mean-1std Log_Stdev Linear (Log_Stdev)
0.01 0.1 1 10 100Frequency, Hz
Figure 11. Response spectral SAFs and natural logarithmic stand ard deviations of SAFs for the Turkey Flat downhole array TF#2.
10 Turkey Flat TF#2 Surface (soil) to TF#1 Surface (rock)
Ratio_Ev_1 Ratio_Ev_2 Ratio_Ev_3 1 Ratio_Ev_4 Ratio_Ev_5 Ratio_Ev_6 Ratio_Ev_7 Ratio_Ev_8 Ratio_Ev_9 Ratio_Ev_10 y = 0.0015x + 0.3272 Log_N_TF1_TF2 0.1 Mean Mean+1sd Mean-1sd Linear (Log_N_TF1_TF2)
0.01 0.1 1 10 100Frequency, Hz
Figure 12. Response spectral SAFs and natural logarithmic stand ard deviations of SAFs for the Turkey Flat downhole array TF#2 surface to TF#1 surface.
10 Turkey Flat TF#3 (soil) to TF#4 (rock) SAF
1
Ratio_1 Ratio_2 Ratio_3 Ratio_4 Ratio_6 0.1 y = 0.0024x + 0.2287 Ratio_7 Ratio_8 Mean Log_N_TF3_TF4 Mean+1std Mean-1std Linear (Log_N_TF3_TF4)
0.01 0.1 1 10 100Frequency, Hz
Figure 13. Response spectral SAFs and natural logarithmic stand ard deviations of SAFs for TF#3 (soil) to TF#4 (rock).
Turkey Flat Average Soil to Average Rock SAFs
10
Ratio_1 Ratio_2 Ratio_3 1 Ratio_4 Ratio_5 Ratio_6 Ratio_7 Ratio_8 Ratio_9 y = 0.0016x + 0.275 Ratio_10 Mean 0.1 Mean+1std Mean-1std Log-N_all Linear (Log-N_all)
0.01 0.1 1 10 100Frequency, Hz
Figure 14. Response spectral SAFs and natural logarithmic stand ard deviations of SAFs for the Turkey Flat average soil/rock.
1 TI_122m Downhole Arrays Sigma TI_104m
Corona
TF#2
SanFrancisco
Carquinez_2 0.1 GV_50 y = -7E-05x + 0.2351 GV_150
Av Sigma
Av+1std
Av-1std
Linear (Av 0.01 Sigma) 0.1 1 10 100Frequency, Hz
Figure 15. Frequency dependence of sigma ln of the SAFs.
1 Soil - Rock Sigma y = 0.0017x + 0.2525 soil-rock
Yerba_TI
TF1_TF2 0.1 y = -7E-05x + 0.2351 downholes TF3_TF4
TF_Soil_Rock
Soil/Rock
Sigma Downholes
Linear (Soil/Rock)
Linear (Sigma Downholes) 0.01 0.1 1 10 100Frequency, Hz
Figure 16. Soil-rock station pairs sigma compared to downhole a rrays sigma.
APPENDIX A
ADDITIONAL DOWNHOLE ARRAYS SCHEMATIC SENSOR LOCATIONS AND P-AND S-WAVES VELOCITY PROFILES
Figure A-1. Garner Valley downhole array (GVDA) and P-and S-wave velocity profiles.
Downloaded from http://nees.ucsb.edu/facilities/GVDA.
Figure A-2. San Francisco Bay Bridge Geotech Array (58961). Sen sors: Surface,14.3, 39.9 meters (m). Downloaded from the Center for Engineering Str ong Motion Data (CESMD) at https://strongmotioncenter.org/.
Figure A-3. Crockett Carquinez Bridge Geotech Array #2 (68259). Sensors: Surface, 61, 125 m. Downloaded from the CESMD at https://strongmotioncenter.org/.
Figure A-4. Corona-I15/Hwy 91 Geotech Array (13186). Sensors: S urface, 7.9, 21.6, 41.8 m. Downloaded from the CESMD at https://strongmotioncenter.org/.
Table 1. Earthquakes recorded at the Treasure Island geotechnic al array at the surface and at 122-m depth
Depth, Epicenter Surface Earthquake Name Date Time Magnitude km Dist., km PGA S-waves dominated lower amplification group Bolinas 1999-08-17 18:06:18 PDT 5.0 ML 6.9 29.0 0.018 Piedmont 2005-05-08 3:35:55 PDT 3.3 MW 5.0 13.1 0.003 Piedmont 2005-09-24 4:25:16 PDT 3.3 MW 5.4 13.5 0.002 Berkeley 2006-12-22 22:49:57 PST 3.6 ML 9.2 11.9 0.029 Berkeley 2006-12-23 09:21:15 PST 3.4 ML 9.3 11.6 0.015 Berkeley 2007-02-23 15:46:15 PST 3.4 ML 10.9 12.5 0.019 Alamo 2008-09-05 21:00:15 PDT 4.0 MW 16.2 33.6 0.025 Berkeley 2009-05-13 15:34:05 PDT 3.1 MW 10.8 13.2 0.004 San Francisco 2010-06-28 07:47:04 PDT 3.3 MW 7.7 18.6 0.003 Morgan Hill 2011-01-07 16:10:16 PST 4.1 ML 7.1 86.5 0.005 Berkeley 2011-10-20 14:41:04 PDT 4.0 MW 8.0 11.7 0.023 Berkeley 2011-10-20 20:16:05 PDT 3.8 MW 9.6 11.8 0.038 Berkeley 2011-10-27 05:36:44 PDT 3.6 ML 9.7 12.1 0.013 El Cerrito 2012-03-05 05:33:19 PST 4.0 ML 9.2 13.1 0.019 Surface waves dominated higher amplification group 11:15:56 AM San Simeon 2003-12-22 PST 6.5 MW 4.7 261.0 0.004 Orinda 2006-03-01 11:34:52 PST 3.4 MW 8.3 14.9 0.006 Glen Ellen 2006-08-02 20:08:12 PDT 4.4 ML 9.1 52.1 0.014 Lafayette 2007-03-01 20:40:00 PST 4.2 ML 16.6 25.7 0.014 Berkeley 2011-07-16 03:31:26 PDT 3.3 MW 6.0 11.0 0.004 Piedmont 2007-07-20 04:42:22 PDT 4.2 MW 5.8 16.5 0.018 San Leandro 2011-08-23 23:36:54 PDT 3.6 ML 8.1 21.8 0.007 Alum Rock 2007-10-30 20:04:54 PDT 5.4 MW 9.2 68.5 0.012 Moraga 2008-06-06 02:02:53 PDT 3.5 MW 7.6 26.3 0.003 El Cerrito 2009-06-06 15:30:56 PDT 3.3 MW 5.6 12.0 0.003 South Napa 2014-08-24 03:20:44 PDT 6.0 MW 11.3 44.1 0.017 Pleasant Hill 2019-10-14 22:33:42 PDT 4.5 MW 14.0 30.5 0.011
Table 2. Earthquakes recorded at Yerba Buena and surface channe ls of Treasure Island
Epicenter Surface Epicenter Earthquake Depth, Dist., km PGA Dist., km PGA Name Date Time Magnitude km Treasure Island Yerba Buena Loma Prieta 1989-10-17 17:04:00 PDT 6.9 MW 18.0 97.7 0.160 95.4 0.060 Berkeley 2011-10-20 14:41:04 PDT 4.0 MW 8.0 11.7 0.023 11.7 0.021 Berkeley 2011-10-20 20:16:05 PDT 3.8 MW 9.6 11.8 0.038 11.8 0.039 El Cerrito 2012-03-05 05:33:19 PST 4.0 ML 9.2 13.1 0.019 14.5 0.015 South Napa 2014-08-24 03:20:44 PDT 6.0 MW 11.3 44.1 0.017 45.9 0.005 Berkeley 2018-01-04 02:39:37 PDT 4.4 MW 12.3 10.8 0.064 10.7 0.042 Pleasant Hill 2019-10-14 22:33:42 PDT 4.5 MW 14.0 30.5 0.011 30.5 0.005
Table 3. Events producing moderate to strong motions at Turkey Flat array (updated after Haddadi et al., 2008)
Event Event Epicenter Distance from Epicenter to: PGA @ Surface No. Name Date Time Mag Lat Lon TF#1 TF#2 TF#3 TF#4 TF#1 TF#2 TF#3 TF#4 1 Parkfield 4/3/1993 21:21:24 PST 4.2 35.942 120.493 14.1 14.5 14.3 13.9 0.026 0.033 0.081 0.047 2 San Simeon 12/22/2003 11:15:56 PST 6.5 35.710 121.100 69.6 70.4 70.6 70.6 0.035 0.036 0.031 0.023 Parkfield 3 (Mainshock) 9/28/2004 10:15:24 PDT 6.0 35.810 120.370 7.6 8.2 8.6 9.2 0.245 0.300 0.260 0.110 4 Aftershock 9/28/2004 10:19:24 PDT 4.2 35.844 120.402 5.5 6.3 6.6 7.0 0.052 0.170 0.072 0.034 5 Aftershock 9/28/2004 10:24:15 PDT 4.7 35.810 120.350 7.6 8.0 8.4 9.1 0.046 0.074 0.053 0.013 6 Aftershock 9/28/2004 10:33:56 PDT 3.7 35.815 120.363 7.0 7.5 8.0 8.6 0.016 0.026 0.026 0.006 7 Aftershock 9/28/2004 12:31:27 PDT 4.0 35.840 120.390 5.1 5.9 6.3 6.7 0.012 0.049 0.024 0.008 8 Aftershock 9/29/2004 10:10:04 PDT 5.0 35.954 120.502 15.5 15.9 15.7 15.2 0.016 0.042 0.037 0.030 9 Parkfield 5/22/2007 04:34:12 PDT 4.0 35.860 120.414 5.4 6.3 NR* NR 0.035 0.054 NR NR 10 Cholame 12/17/2019 10:29:21 PST 4.3 35.806 120.356 7.9 8.5 NR NR 0.051 0.047 NR NR
Table 4. Average logarithmic standard deviation
Average in range of 0.3-40 Hz Site Log_Nat_Sigma ln Ratio Treasure Island 122 m 0.189 1.208 Treasure Island 104 m 0.202 1.224 Corona Bridge 0.282 1.326 TF#2 0.287 1.332 San Francisco Bay Bridge 0.211 1.235 Crockett Carquinez_2 0.216 1.241 Garner Valley 50 m 0.163 1.177 Garner Valley 150 m 0.218 1.243 Av Sigma 0.221 1.247 Yerba_TI 0.203 1.225 TF1_TF2 0.341 1.407 TF3_TF4 0.263 1.301 TF_Soil_Rock 0.282 1.326 Av Sigma 0.272 1.313