GROUND TRUTH LOCATIONS USING SYNERGY BETWEEN REMOTE SENSING AND SEISMIC METHODS: SSSC AT IMS STATIONS FOR TIBETAN PLATEAU EARTHQUAKES Gene A. Ichinose 1, Chandan K. Saikia 2*, Donald V. Helmberger 3, and Mark Simons 3 URS Group, Inc. 1,2 and California Institute of Technology 3 Sponsored by National Nuclear Security Administration Office of Nonproliferation Research and Development Office of Defense Nuclear Nonproliferation Contract No. DE-AC52-03NA99505 1,2,3 ABSTRACT The objective of this project is to collect and create earthquake ground truth (GT) locations for calibration of methods and data, which locate, identify, and discriminate explosions and earthquakes. The method involves identifying seismic events which are candidates for Synthetic Aperture Radar Interferometry (InSAR) analysis and then combine and invert the seismic and geodetic data to create high quality source parameters and GT locations for small to moderate sized and shallow earthquakes in North Africa, Middle East, and Asia. In addition, we provide P-wave travel time corrections known as Source Specific Station Corrections (SSSCs) for the International Monitoring System (IMS) based on these GT locations. We selected ten of the largest earthquakes (M > 4) recorded by the Sino-US Program for Array Seismic Studies of the Continental Lithosphere array between 1991 and 1992 previously analyzed by Zhu et al. (2006) using calibration and CAP methodologies. We first estimated the locations and depths using the regional centroid moment tensor (RCMT) inversion method by inverting the long period filtered complete three component displacement records. The moment tensors are very similar to the CAP regional fault parameters (RFPs) including mechanism, depth and scalar moment. There are some slight differences in location for events with the largest azimuthal gap resulting in poor centriod location resolution. We used the CAP and RCMT solutions to generate SSSCs. We recalculated the earthquake origin times and travel-time residuals relative to the global IASPEI91 layered Earth velocity model based on the fixed location and depth. We then estimated the SSSCs at all stations reporting P-wave arrival times and gridded them using two techniques for developing a global correction surface. We used the continuous curvature spline (CCS) and the nearest neighbor gridding algorithms. The nearest neighbor scheme assumes that the correction is zero outside of a distance search radius while the CCS generates a smoothly varying surface. We relocated earthquakes listed in the International Seismological Centre (ISC) catalog from 1980 to 2004 using these SSSCs. In one example, we relocated 130 earthquakes using the direct grid search locator method for the Qaidam Basin and Qilan Shan region of the Qinghai Province. These earthquakes appear to be mislocated to the southwest by an average of 40 km relative to ISC locations. As many others have previously indicated, the ISC catalog depths are also over estimated and most earthquakes in this region are less then 10 km in depth. These preliminary results assume that the calibration earthquakes have high quality GT locations. Since the CAP and RCMT methods still assume velocity models and the station distributions are not optimal, then there may be some bias but including geodetic data from remote sensing data from InSAR analysis should provide independent constraints. The earthquake grid search locator indicates that the depth is poorly resolved for many of the earthquakes even with the SSSCs, therefore high-quality broadband waveform data will continue to be important for identifying shallow regional earthquakes ideal for InSAR analysis. * Chandan K. Saikia has moved on to AFTAC. 417
OBJECTIVES The objective of this project is to collect and create earthquake GT locations for calibration of methods and data, which locate, identify, and discriminate explosions and earthquakes. Large calibration explosions provide the best GT locations but are rare and limited in geographical distribution. Earthquakes are much more abundant but usually poorly located in space and time because of the sparse spatial distribution in high quality seismic instruments and complex Earth structure. The best approach is to combine the traditional seismic data with the line of sight ground displacements caused by explosions and earthquakes recorded by orbiting satellites based on the Synthetic Aperture Radar Interferometry (InSAR) technique (e.g., Rosen et al., 2000). The overall goal of this ongoing study is to identify seismic events which are candidates for this InSAR analysis and then combine and invert the seismic and geodetic data to create high quality locations and source parameters (e.g., Lohman et al., 2002) for small to moderate sized (4.5 < Magnitude < 6.5) and shallow (< 10 km) earthquakes in North Africa, Middle East, and Asia. In addition, we will provide P-wave travel time corrections in the form of SSSCs for the International Monitoring System (IMS) based on these new GT locations. RESEARCH ACCOMPLISHED In previous proceedings, Saikia et al. (2002, 2003, 2004, 2005) have analyzed and reported the results for the following earthquakes using seismic waveform and geodetic data: 1. 1999/04/30 (Mw 4.9) Southern Iran earthquake 2. 1997/05/05 Southern Iran earthquake 3. 1997/09/18 Southern Iran earthquake 4. 1997/11/08 (Mw 7.6) Manyi, Tibet earthquake 5. 2001/03/05 (Mw 5.7) Tibet earthquake (Manyi aftershock) 6. 1998/01/10 (Mw 5.7) Northeast China earthquake 7. 1999/12/12 Algerian earthquake 8. 1997/03/20 Tunisian earthquake 9. 1998/08/27 (Mw 6.4) Southern Xinjian Province, China earthquake Although the early success of this methodology proved promising, we have later found some difficulty in identifying shallow and small earthquakes for InSAR analysis. Ordering and processing the InSAR data is costly and time consuming and some analyzed earthquakes have not provided usable results primarily due to poor magnitude and depth selection. InSAR data quality including atmospheric noise and topographical uncertainty also limits the data availability to some extent. Success has been made with larger earthquakes but their usefulness for GT is limited due to the uncertainty in hypocenter and origin time relative to centroid solution defined from InSAR. In summary we have concluded that (1) InSAR and seismic data produce similar earthquake source parameters and both are often needed to better constrain a unique GT location, depth, and mechanism, (2) ISC global earthquake locations are often over estimated in depth, (3) the use of SSSC constrained by InSAR defined GT locations improve the locations of smaller localized seismic events including GT events which were left out as validation examples. In this report we continue to explore earthquakes for which we can derive GT locations within the Tibetan Plateau by first examining ISC phase data and high quality PASSCAL and IRIS seismic waveform data. We continue to focus on this region because of the availability of travel time data from local network earthquake catalogs but work will also continue in other regions of Asia including the Korean Peninsula, Northern Africa and the Middle East (including Arabian Peninsula and Caspian Sea regions). Regional Centriod Moment Tensor Inversion We performed a regional centriod moment tensor (RCMT) inversion on regional waveforms recoded by the Sino-US PASSCAL seismic experiment on the Tibetan Plateau (Owens et al., 1993). We selected ten of the largest earthquakes (M > 4) recorded by the array from 1991 to 1992 (Table 1, Figure 1) previously analyzed by Zhu et al. (2006) and Tan et al (2006) using calibration and CAP methodologies (Zhu and Helmberger, 1996b). We inverted the three component long period filtered (100 10 s) displacements (Figure 2) and performed a grid search for the best fitting centriod longitude, latitude, depth, and origin time (e.g., Ritsema and Lay, 1995) while maximizing the variance reduction and percent double couple component (Table 2). We assume a 1D Tibet velocity model (e.g., Zhu 418
et al. 2006) that works well in generating realistic f-κ synthetic seismograms (e.g., Zhu et al., 1996a; Rodgers and Schwartz, 1998; Saikia et al. 2002) for this region. The model has a 61 km thick crust with a P-wave velocity of 6.2 km/s and a 4 km thick shallow low velocity layer of 4.7 km/s. The upper mantle is 8.2 km/s. The RCMT results are shown in Figure 1, indicating that the moment tensors are very similar to the CAP regional fault parameters (RFP) compared with Zhu et al. (2006), including mechanism, depth, and moment. There are some slight differences in location, primarily for earthquakes located outside of the array with the largest azimuthal gap in station coverage. A centriod solution is necessary because some of the CAP locations were inadequate and led to some RCMT with artificially high Compensated Linear Vector Dipole (CLVD) components. In some cases we included seismic waveform data recorded by the China Digital Seismic Network (CDSN) and IRIS stations that improve the azimuthal coverage and the moment tensors but may introduce artifacts and biases in the centriod location due to the velocity model complexity outside of the plateau. Because of this complexity, we instead used the CAP solutions to generate SSSCs. The RCMT centroid locations led to very different and complex SSSCs relative to the much smoother ones based on the CAP RFPs. Source-Specific Station Corrections at IMS Seismic Stations We recalculated the earthquake origin times and travel-time residuals relative to the global IASPEI91 layered Earth velocity model based on the fixed location and depth estimated from the CAP methodology (Zhu et al., 2006). We will order and analyze InSAR data for these events in the future. We estimated the SSSCs at seismic stations reporting P-wave arrival times and gridded them using two techniques for developing a global correction surface for this region. We used was the continuous curvature spline (CCS) gridding algorithm (Smith and Wessel, 1990) shown in Figure 3 and the nearest neighbor gridding algorithm (e.g., Wessel and Smith, 1998) shown in Figure 4. The Nearest Neighbor scheme assumes that the correction is zero where no data is present (outside of a 15 distance search radius) while the CCS generates a much smoother interpolated surface. Northern Tibet Earthquake Relocations Using SSSCs We selected ISC catalog earthquakes between 1980 and 2004 within 200 km radius of each of the calibration events for relocations with the SSSCs. We use a direct grid search scheme to relocate the earthquakes and select the locations with the lowest root-mean-square (RMS) error. The relocations are fast because of the use of pre-computed lookup tables in LOCSAT format using the TauP program (Crotwell et al., 1999). The tables are read in and interpolated using a cubic spline interpolation method. We also iterate over hypocenter depths for the precomputed travel times in the tables. Origin times are updated for each grid point to minimize any biases in the average residual before the RMS error calculation. Figure 5 shows examples of the relocations. The RMS error for the optimal depth and origin time is plotted for one of the calibration earthquakes and then used to relocate other nearby earthquakes. As a validation, the calibration earthquake shown in Figure 5 is relocated to within 5.7 km of the CAP solution using the SSSCs but is mislocated by 48.7 km to the southwest near the ISC location. These accuracy estimates are consistent with cluster analysis (Engdahl and Bergman, 2001) and we expect the relocated earthquakes to be promoted to at least GT5 or GT10 status. Figure 5 also shows that the minimum RMS error do not vary significantly (<10%) over a radius of < 5 to 10 km around the best solution (Figure 6) so the data can only resolve the location to within 5 to 10 km particularly when teleseismic phase data are included in the relocation. The addition of local phase arrival times greatly improves the resolution. Figure 7 shows the locations for 130 ISC catalog earthquakes in the Qaidam Basin and Qilan Shan region of the Qinghai Province between 1980 and 2004 within a 200 km radius of the 1991/09/02 (Julian Day 245) earthquake shown in Figure 5. The ISC locations are shown relative to the grid search relocations using the SSSC that is estimated by using the continuous spline interpolation gridding scheme (Figure 3). Figure 8 is similar to Figure 7 but the SSSC is estimated using the nearest-neighbor gridding scheme (Figure 4). The results are very similar because the phase arrival times are recorded from the same global regions of at least the past 24 years; therefore, it is not significant to have travel time information from all regions for constructing SSSC surfaces. CONCLUSIONS AND RECOMMENDATIONS Earthquakes in the ISC catalog appear to be mislocated to the southwest by an average of 40 km in the Qaidam Basin and Qilan Shan region of the Qinghai Province. As many others have indicated, the ISC catalog depths are often over estimated and most earthquakes in this region are less then 10 km in depth. These preliminary results 419
assume that the calibration earthquakes have high quality GT locations. Since the CAP and RCMT methods still assume velocity models, there may be some inherent bias in the GT locations and SSSCs. InSAR should provide independent constraints for GT and can be combined with this seismic analysis. We confirm that all ten of the calibration earthquakes we examined in Table 1 are shallow and the ones with Mw > 4.8 should provide good InSAR results. We also have additional local network phase arrival data from Chinese earthquake catalogs for 1996 to 2000 that can be used to better validate the relocations, improve the SSSCs, or constrain GT locations. We can also provide SSSCs relative to other global 1D velocity models, regionalized models including Regionalized Upper-Mantle models and also with 3D crustal and upper-mantle CUB shear wave velocity model (e.g., Shapiro and Ritzwoller, 2002). REFERENCES Crotwell, P. H., T. J. Owens, and J. Ritsema (1999). The TauP Toolkit: Flexible seismic travel-time and ray-path utilities, Seismol. Res. Lett. 70: 154 160. Engdahl, E. R. and E. A. Bergman (2001). Validation and generation of reference events by cluster analysis, in Proceedings of the 23rd Seismic Research Review: Worldwide Monitoring of Nuclear Explosions, LA-UR-01-4454, Vol. 1, pp. 205 214. Lohman, R., M. Simons, and B. Savage (2002). Location and mechanism of the Little Skull Mountain earthquake as constrained by satellite radar interferometry and seismic waveform modeling, J. Geophys. Res. 107 (B6): 10.1029/2001JB000627. Owens, T. J., G. E. Randall, F. T. Wu, and R. S. Zeng(1993). PASSCAL instrument performance during the Tibetan Plateau passive seismic experiment, BSSA 83: 1957 1970. Rodgers, A. J. and S. Y. Schwartz (1998). Lithospheric structure of the Qiangtang Terrane, northern Tibetan Plateau, from complete regional waveform modeling: Evidence for partial melt, J. Geophys. Res. 103 (B4): 7137 7152. Ritsema, J. and T. Lay (1995). Long period regional wave moment tensor inversion for earthquakes in the western United States, J. Geophys. Res. 100: 9853 9864. Rosen, P. A., S. Hensley, I. Joughin, F. K. Li, S. N. Madsen, E. Rodriguez, and R. M. Goldstein (2000). Synthetic aperture radar Intererometry, in Proceedings of the IEEE, Vol. 88, pp. 333 382. Saikia, C. K., H. K. Thio, D. V. Helmberger, G. Ichinose, and C. Ji (2005). Ground truth locations using synergy between remote sensing and seismic methods Application to Chinese and North African earthquakes, in Proceedings of the 27th Seismic Research Review: Ground-Based Nuclear Explosion Monitoring Technologies, LA-UR-05-6407, Vol. 1, pp. 443 453. Saikia, C. K., J. Chen, G. Ichinose, D. V. Helmberger, A. O. Konca, and M. Simons (2004). Ground truth locations using synergy between remote sensing and seismic methods Application to Chinese earthquakes, in Proceedings of the 26th Seismic Research Review: Trends in Nuclear Explosion Monitoring, LA-UR-04-5801, Vol. 1, pp. 328 337. Saikia, C. K., R. Lohman, G. Ichinose, and M. Simons (2003). Ground truth locations using a synergy between INSAR and seismic methods, in Proceedings of the 25th Seismic Research Review Nuclear Explosion Monitoring: Building the Knowledge Base, LA-UR-03-6029, Vol. 1, pp. 324 333. Saikia, C. K., R. Lohman, G. Ichinose, D. V. Helmberger, M. Simons, and P. Rosen (2002). Ground truth locations A synergy of seismic and synthetic aperture radar interferometric methods, in Proceedings of the 24th Seismic Research Review Nuclear Explosion Monitoring: Innovation and Integration, LA-UR-02-5048, Vol. 1, pp. 420 429. Shapiro, N. M. and M. H. Ritzwoller (2002). Monte-Carlo inversion for a global shear-velocity model of the crust and upper-mantle, Geophys. J. Int. 151: 88 105. 420
Smith, W. H. F. and P. Wessel (1990). Gridding with continuous curvature splines in tension, Geophysics, 55: 293 305. Wessel, P. and W. H. F. Smith (1998). New, improved version of Generic Mapping Tools released, EOS Trans. AGU 79: 579. Zhu, L. and D. V. Helmberger (1996a). Intermediate depth earthquakes beneath the India-Tibet collision zone, Geophys. Res. Lett. 23: 435 438. Zhu, L. and D. V. Helmberger (1996b). Advancement in source estimation techniques using broadband regional seismograms, Bull. Seismol. Soc. Am. 86: 1634 1641. Zhu, L., Y. Tan, D. V. Helmberger, and C. K. Saikia (2006). Calibration of the Tibetan Plateau using regional seismic waveforms, Pure Appl. Geophys. (submitted). Table 1. Tibet Plateau calibration earthquakes Julian Day Date (y/m/d) Origin Time sec Latitude ( N) Longitude ( E) Z (km) M1 Type 1 M2 Type 2 Agency Regional Location 222 1991/08/10 20:21 52.7 33.874 92.191 15 4.5 mb 4.8 Ms ISC Qinghai Province 52.5 33.970 92.180 10 4.7 Mw - - CAP 54.0 33.900 92.200 8 4.8 Mw - - RMTI 245 1991/09/02 11:05 50.5 37.408 95.384 10 5.3 mb 4.8 Ms ISC Qinghai Province 47.0 37.770 95.570 12 4.7 Mw - - CAP 52.0 37.440 95.540 14 4.9 Mw - - RMTI 263 1991/09/20 11:16 11.7 36.177 100.058 13 5.4 mb 5.1 Ms ISC Qinghai Province 10.8 36.240 100.160 12 4.9 Mw - - CAP 10.0 36.300 100.100 13 5.0 Mw - - RMTI 325 1991/11/21 13:37 43.8 33.736 90.237 49 4.0 mb - Ms ISC Qinghai Province 38.6 34.020 90.140 5 4.5 Mw - - CAP 38.0 33.950 90.100 5 4.6 Mw - - RMTI 348 1991/12/14 08:20 23.7 33.925 88.842 31 5.1 mb 4.6 Ms ISC Xizang 19.0 34.040 88.830 5 4.8 Mw - - CAP 22.0 33.900 88.950 4 4.9 Mw - - RMTI 357a 1991/12/23 01:58 25.0 33.898 88.887 32 5.2 mb 4.6 Ms ISC Xizang 20.5 34.050 88.840 5 4.8 Mw - - CAP 28.0 33.800 89.100 4 4.9 Mw - - RMTI 357b 1991/12/23 02:14 56.0 33.897 88.965 48 4.8 mb - Ms ISC Xizang 50.7 34.010 88.860 5 4.5 Mw - - CAP 52.0 33.800 88.900 4 4.6 Mw - - RMTI 34 1992/02/03 15:44 22.6 34.452 93.260 10 4.9 mb 4.5 Ms ISC Qinghai Province 22.7 34.600 93.230 5 4.5 Mw - - CAP 24.0 34.500 93.300 4 4.5 Mw - - RMTI 37 1992/02/06 03:35 15.0 29.608 95.639 15 5.5 mb 4.9 Ms ISC India-China 15.0 29.710 95.660 5 5.0 Mw - - CAP Border Region 15.0 29.700 95.900 4 5.0 Mw - - RMTI 104 1992/04/13 03:47 50.7 31.948 88.306 35 4.4 mb 4.4 Ms ISC Xizang 45.3 31.750 88.220 5 4.5 Mw - - CAP 48.0 31.700 88.400 4 4.6 Mw - - RMTI 421
Table 2. Tibetan Plateau moment tensors and focal mechanisms Julian Day TYPE Nodal Plane 1 (Strike /Dip /Rake ) Nodal Plane 2 (Strike /Dip /Rake ) Mo (dyne*cm) Mw OT (sec) Z (km) DC (%) VRED (%) 222 RMTI 249/70/4 158/86/160 2.03 10 23 4.8 54 8 97 91.6 CAP 252/83/2 1.40 10 23 4.7 53 10 245 RMTI 103/50/70 313/44/112 2.68 10 23 4.9 52 14 93 62.5 CAP 102/61/79 1.40 10 23 4.7 47 12 263 RMTI 109/59/62 334/41/127 3.64 10 23 5 10 13 96 90.2 CAP 99/76/60 2.79 10 23 4.9 11 12 325 RMTI 71/81/-11 162/80/-170 1.08 10 23 4.6 38 5 100 88.2 CAP 251/60/-16 7.00 10 23 4.5 39 5 348 RMTI 216/45/-66 4/50/-112 2.51 10 23 4.9 22 4 93 86.5 CAP 212/30/-72 1.97 10 23 4.8 19 5 357a RMTI 196/40/-90 16/50/-90 2.50 10 23 4.9 28 4 97 86.5 CAP 210/31/-70 1.97 10 23 4.8 20 5 357b RMTI 168/43/-95 356/47/-85 8.77 10 23 4.6 52 4 94 87.4 CAP 358/60/-87 7.00 10 22 4.5 51 5 34 RMTI 293/35/66 141/58/106 6.17 10 22 4.5 24 4 88 80 CAP 107/19/53 7.00 10 22 4.5 23 5 37 RMTI 3/56/64 224/42/123 4.32 10 23 5 15 4 87 81.5 CAP 0/61/60 3.94 10 23 5 15 5 104 RMTI 147/56/-173 53/84/-35 9.42 10 22 4.6 48 4 100 78.5 CAP 227/49/-23 7.00 10 22 4.5 45 5 Table 3. SSSC at IMS stations (N nearest neighbor algorithm/s continuous spline interpolation) Station Longitude Latitude SSSC(N) SSSC(S) Station Longitude Latitude SSSC(N) SSSC(S) PS01-70.6-40.7 0-4.7 PS27 10.8 60.8 0.23 0.67 PS02 134.3-19.9-5.41-5.49 PS28 25.5 69.5 1.5 1.59 PS03 133.9-23.7-4.62-4.66 PS29 73.3 33.7-3.03-3.16 PS04 141.6-31.9-4.9-4.83 PS30-57.3-26.3 0-3.81 PS05 62.9-67.6 0-7.44 PS31 127.9 37.5 0.74 0.12 PS06-68.1-16.3 0-3.06 PS32 42.9 43.7-0.24-0.13 PS07-48 -15.6 0-3.15 PS33 84.8 53.9 0.63 0.44 PS08-95.9 50.2 0.68 1.21 PS34 88 69 0.91 1.08 PS09-114.6 62.5 1.24 1.37 PS35 112.6 59.6 1.72 2.05 PS10-66.8 54.8 0.89 0.89 PS36 157.8 53.1 1.46 1.03 PS11 18.4 5.2 0-2.03 PS37 132 44.2 1.14 0.85 PS12 119.7 49.3 0.63 1.72 PS39 25.6-28.6-3.41-4.34 PS13 103.8 36.1 0.25 0.3 PS40-4 39.7 1.31 1.3 PS14-74.3 4.9 0-1.6 PS41 99 18.8-4.75-4.59 PS15-4.9 6.7-1.53-1.59 PS42 8.7 35.6 1.07 0.38 PS16 33 26 0.49-0.51 PS43 32.8 39.9-0.52 0 PS17 28.1 61.4 0.89 1.23 PS44 58.1 37.9-1.18-1.04 PS18-149.6-17.6 0-2.48 PS45 29.1 50.4 0.27 0.53 PS19 13.7 48.9 1.19 1.19 PS46-103.7 29.3 0.65 0.59 PS21 51.4 35.8 1.08-0.29 PS47-118.2 38.4 2.6 1.74 PS22 138.2 36.5-1.88-1.59 PS48-109.6 42.8 1.69 1.79 PS23 82 46.8-1.53-0.89 PS49-146.9 64.8 2.67 2.82 PS24 37.2-1.1 0-2.9 PS50 161.9-77.5 0-7.4 PS25 106.8 48 6.28 3.75 422
Figure 1. Regional centriod moment tensor solutions (left side) and CAP (right side) and calibration results (Zhu et al., 2006) estimated using broadband regional waveform data from PASSCAL deployment of seismic stations. Figure 2. Regional centriod moment tensor solutions of event 357a (Figure 1, Tables 1 and 2) estimated using broadband regional waveform data from PASSCAL deployment of high quality seismic array. The centriod (location, depth, and origin time) was also estimated using direct grid search scheme. We used the 1D Tibet model layered Earth model for calculating the f-k Green s functions (e.g., Zhu et al., 2006). 423
Figure 3. SSSC surface relative to IASPEI91 layered Earth model. We used a continuous curvature spline gridding algorithm to construct a global correction surface. The dots shown are a group of average station residuals from earthquakes in northern Tibet, Qaidam basin and Qilan Shan region. Figure 4. SSSC surface relative to IASPEI91 layered Earth model. We used a nearest neighbor gridding algorithm to construct a global surface. The dots shown are a group of average station residuals from earthquakes in northern Tibet, the Qaidam basin and the Qilan Shan region. The major difference for this gridding scheme, compared with the nearest neighbor method in Figure 3, is that the correction is not constrained to zero in areas without data and the curvature spline method generates a smoother correction surface. 424
Figure 5. Direct grid search relocations of the 1991/09/02 (Julian Day 245 in Table 1; Figure 1) GT calibration earthquakes with and without SSSC. The root-mean-square (RMS) residuals are contoured at different depths based on the IASPEI91 layered Earth velocity model. The individual grid points (1600) are also shown to indicate where the RMS residual surface is constrained. The best fit location shown as the circle is less than 5 km from the CAP solution with the SSSC but mislocated by 40 km to the southwest without the SSSC. Figure 6. Direct grid search relocation results for two selected individual earthquakes with SSSC not from the GT calibration subset. The RMS residuals for the events are contoured at different depths based on the IASPEI91 layered Earth velocity model. The individual grid points (1600) are also shown to indicate where the RMS residual surface is constrained. The RMS error scale is different in the right panel. 425
Figure 7. Direct grid search earthquake relocations for ISC catalog events between 1980 and 2004 using IASPEI91 layered Earth model and SSSC derived by the curvature spline fitting method shown in Figure 3b. Figure 8. Direct grid search earthquake relocations for ISC catalog events between 1980 and 2004 using IASPEI91 layered Earth model and SSSC derived by the nearest neighbor fitting method shown in Figure 3a. 426