IMPROVEMENT OF 3-DIMENSIONAL BASIN STRUCTURE MODEL USING GROUND MOTION RECORDINGS

Transcription

1 13 th World Conference on Earthquake Engineering Vancouver, B.C., Canada August 1-6, 24 Paper No IMPROVEMENT OF 3-DIMENSIONAL BASIN STRUCTURE MODEL USING GROUND MOTION RECORDINGS Hidenori KAWABE 1 and Katsuhiro KAMAE 2 SUMMARY The last few years, some three-dimensional subsurface structure models of the Osaka sedimentary basin have been developed. These models have been constructed by the geophysical and the geological data such as borehole, seismic reflection survey, microtremors and gravity anomaly data. However, these models have partially ambiguity because of the lack of geophysical and geological information. In this study, we try to improve the three-dimensional subsurface structure model of the north side of the Osaka basin through the simulations of observed seismic waveforms. First, we tune the shape of the boundary between bedrock and sedimentary layers based on the forward modeling using direct S-waves. Next, we search the optimum Qs value of sedimentary layers from the investigation of the fittings between the synthetic later phases and the observed ones. Finally, we verify the accuracy of the improved threedimensional subsurface structure model by the simulations of several earthquakes. Resultantly, synthetic S-wave phase and amplitude due to the improved model are good agreement with observed ones. Furthermore, we estimated Qs=Vs/5 as the optimum Qs for the three-dimensional subsurface structure model. INTRODUCTION The improvement in computer capacities has made it possible to carry out simulations of strong ground motions for realistic complex media like a sedimentary basin, such as the Osaka basin, southwest Japan. On the other hand, the last few years, some three-dimensional subsurface structure models of the Osaka sedimentary basin have been developed (Kagawa et al. [1], Miyakoshi et al. [2], Zhao et al. [3], Horikawa et al. [4]). These models have been constructed by the geophysical and the geological data such as borehole, seismic reflection survey, microtremors and gravity anomaly data. However, these models have partially ambiguity because of the lack of geophysical and geological information. As a result, such models can not accurately reproduce the seismic waveforms. In this study, we concentrate on the north side of the Osaka basin, and try to improve the three-dimensional subsurface structure model through the simulations of observed seismic waveforms. First of all, we tune the shape of the boundary between 1 Research Associate, Research Reactor Inst., Kyoto Univ., Osaka, Japan 2 Associate Professor, Research Reactor Inst., Kyoto Univ., Osaka, Japan

2 bedrock and sedimentary layers based on the forward modeling using direct S-waves. Next, we search the optimum Qs value of sedimentary layers from the investigation of the fittings between the synthetic later phases and the observed ones. Finally, we verify the accuracy of the improved three-dimensional subsurface structure model by the simulations of several earthquakes. SURFACE GEOLOGY OF THE NORTHERN PART OF THE OSAKA BASIN The geological structures of the north side of the Osaka basin (north of the Yodo River) are very complicated. As shown in Figure 1, the Senri hills composed of stiff sediments is located off from the surrounding mountains. Moreover, there is a graben structure between the Hokusetsu Mountains and the Senri hills. Several active faults exist in this area. In particular, the Arima-Takatsuki fault zone with total extension of about 5km runs from the east to the west. Also, the Butsunenjiyama fault stretches through the Senri hills from the north to the south, and connects with the Uemachi fault. The total length of these two faults is about 4km F4 F5 F3 F8 F2 F7 F6 F1 F9 F1 F11 5KM Figure 1 A map of faults location in the north of the Osaka basin. F1 F2, F3, F4, F5, F6, F7, F8, F9, F1 and F11 Takatsuku-Tennouzan fault, F2:Nyoidani fault, Satsukugaoka fault, Hanayashiki fault, Rokko fault, Ai fault, Toyokawa fault, Nobatake fault, Bojima fault, Butsunenjiyama fault and Uemachi fault, respectively. The Arima-Takatsuki fault zone consists of F1 F2, F3, F4, F5, F6, F7, F8 and F9.

3 METHODOLOGY FOR IMPROVEMENT OF 3-DIMENSIONAL BASIN STRUCTURE MODEL The algorithm for improving 3-dimensional basin structure model is shown in Figure 2. The algorithm consists of three steps. The first step is to construct the initial model for three-dimensional sedimentary basin structure using geophysical and geological information. Here, the sediment and bedrock are divided into three layers with three kinds of physical parameters, respectively. The existence of the boundary between bedrock and sediment gives much effect to propagation of seismic waves due to the high contrast of physical parameters. The second step is to reproduce the direct body waves (phase and amplitude) appeared in the seismic recordings. Resultantly, we determine the optimum boundary shapes between layers. The last step is to estimate the optimum attenuation factors from the comparison of the amplitudes between the later phases appeared the synthetics and observed ones. St ep 1 (a) Initial model of 3-D basin structure St ep 2 (b) Synthesize strong ground motions (c) Evaluation by comparing observed waveforms with synthetic ones for the body wave Change boundary shapes Optimum boundary shapes model St ep 3 (d) Synthesize strong ground motions (e) Evaluation by comparing observed waveforms with synthetic ones for the later phase Change Qs value Optimum model of 3-D basin structure Figure 2. The algorithm for improving 3-dimensional basin structure mode SIMULATION METHOD In order to simulate observed waveforms, we use three-dimensional finite difference scheme, second-order accuracy in time, fourth-order accuracy in space, with nonuniform spacing staggered-grid formulation (Pitarka, et al. [5]). We set an absorbing region outside the finite computational region, and apply nonreflecting boundary condition of Cerjan et al. [6] to the region. In Cerjan et al. [6], Gaussian functions given by W = exp(-α(j -j) 2 ), (j=1,2,,j ). (1)

4 In this study, α=.5 and J =6 are employed. In addition, we use the A1 absorbing boundary condition of Clayton and Engquist [7], as applied to the velocity components of the wave field. Anelastic attenuation was described by Graves [8]. He used a spatially variable Q operator having a linear dependence on frequency. Such Q operator is consistent with the constant Q model at a specified reference frequency without depending on the frequency. In this study, the reference frequency was set at 1 Hz. IMPROVEMENT OF THE NORTHERN PART OF THE OSAKA BASIN STRUCTURE MODEL Here, we construct an initial three-dimensional subsurface structure model for the northern part of the Osaka sedimentary basin. After that, we improve the initial model from the waveform simulations of three events (Table 1). A map of the epicenters (JMA) and observation stations are shown in Figure 3. The source parameters for event 1 have been estimated by Kawabe et al. [9]. The source parameters for other events are due to the F-net data of the National Institute for Earth Science and Disaster Research, Japan. Main parameters for the simulations by the finite difference method are listed in Table 2. First of all, we constructed a velocity structure model referring to Miyakoshi et al. [2] and Sekiguchi et al. [1]. The model has three sedimentary layers and three bedrock layers. Physical parameters of each layer are shown in Table 3. The boundary shape between bedrock (Layer4) and the deepest sedimentary layer (Layer3) of the initial model has been determined by the geophysical and geological data such as borehole, seismic reflection survey, microtremors. The boundary depths between Layer1 and Layer2, Layer2 and Layer3 have been determined from the following equations by Miyakoshi et al. [2]. d 12 =.193 d 34 (2) d 23 =.473 d 34 (3) where d 12, d 23 and d 34 indicate the depth of the boundary between Layer1 and Layer2, Layer2 and Layer3, Layer3 and Layer4, respectively. Next, the initial model is improved by the forward modeling using direct S-waves for three events shown in Table 1. The synthetic waveforms due to the initial model and the improved model are compared with the observed ones in Figure 4. You can see that synthetic S-wave phase and amplitude due to the improved model are good agreement with observed ones at every site. The depth of bedrock (layer 4) in the improved model is shown in Figure 5. Finally, Qs values of each layer were determined by reproduction of the later phases in the recordings. Here, we assumed that the anelastic attenuation factor for a reference frequency (Graves [9]) of S-wave (Qs) is proportional to S-wave velocity (Vs). We simulated the ground motions from three events (Table 1) using six cases of Qs (Qs=Vs/2, Vs/1, Vs/5, Vs/2.5, Vs/2, infinity). Comparisons of the synthetic waveforms with the observed ones are shown in Figure 6. This figure shows that the amplitude of the later phases synthesized using Qs=Vs/5 is good agreement with the observed one at 5 sites. Resultantly, we can recommend to use Qs=Vs/5 as the optimum Qs in waveform simulations. Table 1 Location parameters of the events used for simulations Origin Time (JST) Lon. Lat. Depth (km) M JMA Event1 1995/1/17, 7: Event2 21/8/25, 22: Event3 2/1/31, 1:

5 135 3' 136 ' 136 3' km 35 3' 5 Event 2 35 ' Event 1 TYN The target region of the improvement of the basin structure model 34 3' AMA Osaka Bay FKS MRG YAE Event 3 Figure 3. A map showing the location of Sources, observation stations and target region of the improvement of the basin structure model. Table 2. Model parameters for finite difference shimulations Event 1 Event 2 Event 3 Spatial discretization (km) Inside basin Outside basin Model size (Total grid points) Temporal discretization Time step Min. S-wave velocity (km/s) Max. P-wave velocity (km/s) Sedimentay layers Bedrock layers Table 3. Physical parameters of sedimentary layers and bedrock layers depth Vp Vs Density Qs (for Step 2) Layer km 1.6 km/s.4 km/s 1.7 tf/m3 7 Layer km 1.8 km/s.55 km/s 1.8 tf/m3 11 Layer km 2.5 km/s 1. km/s 2.1 tf/m3 2 Layer km 5.4 km/s 3.2 km/s 2.6 tf/m3 46 Layer5 5.1 km 5.9 km/s 3.46 km/s 2.7 tf/m3 49 Layer6 18. km 6.6 km/s 3.87 km/s 2.8 tf/m3 55

6 AMA NS Syn.1 Max:.659cm/s Syn.2 Max:.438cm/s Obs. Max:.481cm/s FKS NS Syn.1 Max:.261cm/s Syn.2 Max:.229cm/s Obs. Max:.321cm/s EW Syn.1 Max:.153cm/s Syn.2 Max:.91cm/s Obs. Max:.513cm/s EW Syn.1 Max:.783cm/s Syn.2 Max:.587cm/s Obs. Max:.678cm/s UD Syn.1 Max:.175cm/s Syn.2 Max:.143cm/s Obs. Max:.184cm/s UD Syn.1 Max:.17cm/s Syn.2 Max:.168cm/s Obs. Max:.265cm/s MRG NS Syn.1 Max:.488cm/s Syn.2 Max:.31cm/s Obs. Max:.327cm/s YAE NS Syn.1 Max:.48cm/s Syn.2 Max:.36cm/s Obs. Max:.613cm/s EW Syn.1 Max:.776cm/s Syn.2 Max:.526cm/s Obs. Max:.588cm/s EW Syn.1 Max:.744cm/s Syn.2 Max:.53cm/s Obs. Max:.527cm/s UD Syn.1 Max:.198cm/s Syn.2 Max:.135cm/s Obs. Max:.134cm/s UD Syn.1 Max:.269cm/s Syn.2 Max:.157cm/s Obs. Max:.235cm/s Figure 4. Comparison of the observed waveforms with synthetic ones. Syn.1 and Syn.2 indicate initial model simulation result and optimum boundary model simulation result respectively. (Bandpass filter:.1-.7hz)

7 TYN NS Syn.1 Max:.853cm/s Syn.2 Max:.525cm/s Obs. Max:.573cm/s EW Syn.1 Max:.38cm/s Syn.2 Max:.196cm/s Obs. Max:.297cm/s UD Syn.1 Max:.98cm/s Syn.2 Max:.56cm/s Obs. Max:.71cm/s Figure 4. Continued ' 135 3' ' 135 4' ' 135 3' ' 135 4' 34 5' 34 45' 34 4' TYN AMA FKS MRG YAE (a) The bedrock depth of initial model 34 5' 34 45' 34 4' AMA FKS TYN MRG YAE (b) The bedrock depth of optimum boundary model (m) Figure 5. The depth of bedrock (layer 4) of initial model and optimum boundary model.

8 AMA.7 Syn. NS Max:.433cm/s Obs. NS Max:.4828cm/s.3 Syn. EW Max:.85cm/s Obs. EW Max:.1822cm/s.4 Syn. UD Max:.1415cm/s Obs. UD Max:.1627cm/s Qs = Vs/1 Syn. NS Max:.4337cm/s Obs. NS Max:.4828cm/s.3 Qs = Vs/1 Syn. EW Max:.86cm/s Obs. EW Max:.1822cm/s.4 Qs = Vs/1 Syn. UD Max:.1421cm/s Obs. UD Max:.1627cm/s Syn. NS Max:.4814cm/s Obs. NS Max:.4828cm/s.3 Syn. EW Max:.94cm/s Obs. EW Max:.1822cm/s.4 Syn. UD Max:.1485cm/s Obs. UD Max:.1627cm/s Syn. NS Max:.563cm/s Obs. NS Max:.4828cm/s.3.5 Syn. EW Max:.98cm/s Obs. EW Max:.1822cm/s.4.5 Syn. UD Max:.1739cm/s Obs. UD Max:.1627cm/s Syn. NS Max:.512cm/s Obs. NS Max:.4828cm/s.3 Syn. EW Max:.992cm/s Obs. EW Max:.1822cm/s.4 Syn. UD Max:.1873cm/s Obs. UD Max:.1627cm/s Qs : infinity Syn. NS Max:.5314cm/s Obs. NS Max:.4828cm/s.3 Qs : infinity Syn. EW Max:.1332cm/s Obs. EW Max:.1822cm/s.4 Qs : infinity Syn. UD Max:.2551cm/s Obs. UD Max:.1627cm/s FKS.5 Syn. NS Max:.1811cm/s Obs. NS Max:.3226cm/s.9 Syn. EW Max:.5434cm/s Obs. EW Max:.6789cm/s.8 Syn. UD Max:.924cm/s Obs. UD Max:.2649cm/s Qs = Vs/1 Syn. NS Max:.1952cm/s Obs. NS Max:.3226cm/s.9 Qs = Vs/1 Syn. EW Max:.5858cm/s Obs. EW Max:.6789cm/s.8 Qs = Vs/1 Syn. UD Max:.275cm/s Obs. UD Max:.2649cm/s Syn. NS Max:.218cm/s Obs. NS Max:.3226cm/s.9 Syn. EW Max:.6471cm/s Obs. EW Max:.6789cm/s.8 Syn. UD Max:.3548cm/s Obs. UD Max:.2649cm/s Syn. NS Max:.2574cm/s Obs. NS Max:.3226cm/s.9.5 Syn. EW Max:.679cm/s Obs. EW Max:.6789cm/s.8.5 Syn. UD Max:.4632cm/s Obs. UD Max:.2649cm/s Syn. NS Max:.2669cm/s Obs. NS Max:.3226cm/s.9 Syn. EW Max:.6839cm/s Obs. EW Max:.6789cm/s.8 Syn. UD Max:.4873cm/s Obs. UD Max:.2649cm/s Qs : infinity Syn. NS Max:.3126cm/s Obs. NS Max:.3226cm/s.9 Qs : infinity Syn. EW Max:.711cm/s Obs. EW Max:.6789cm/s.8 Qs : infinity Syn. UD Max:.637cm/s Obs. UD Max:.2649cm/s Figure 6. Comparison of the observed waveforms with synthetic ones of six cases of Qs. (Bandpass Filter:.1-.7Hz)

9 Syn. NS Max:.2379cm/s Obs. NS Max:.3263cm/s Qs = Vs/ Syn. NS Max:.2879cm/s Obs. NS Max:.3263cm/s Syn. NS Max:.3525cm/s Obs. NS Max:.3263cm/s Syn. NS Max:.3892cm/s Obs. NS Max:.3263cm/s Syn. NS Max:.3958cm/s Obs. NS Max:.3263cm/s Qs : infinity -.7 Syn. NS Max:.4288cm/s Obs. NS Max:.3263cm/s Qs = Vs/ Syn. NS Max:.1789cm/s Obs. NS Max:.2739cm/s Syn. NS Max:.2663cm/s Obs. NS Max:.2739cm/s Syn. NS Max:.4151cm/s Obs. NS Max:.2739cm/s Syn. NS Max:.5175cm/s Obs. NS Max:.2739cm/s Qs : infinity -.7 Syn. NS Max:.539cm/s Obs. NS Max:.2739cm/s Syn. NS Max:.6436cm/s Obs. NS Max:.2739cm/s MRG Syn. EW Max:.542cm/s Obs. EW Max:.5898cm/s Qs = Vs/ Syn. EW Max:.5233cm/s Obs. EW Max:.5898cm/s Syn. EW Max:.5777cm/s Obs. EW Max:.5898cm/s Syn. EW Max:.654cm/s Obs. EW Max:.5898cm/s Syn. EW Max:.689cm/s Obs. EW Max:.5898cm/s Qs : infinity -.8 Syn. EW Max:.6323cm/s Obs. EW Max:.5898cm/s YAE Qs = Vs/ Syn. EW Max:.4623cm/s Obs. EW Max:.5163cm/s Syn. EW Max:.4924cm/s Obs. EW Max:.5163cm/s Syn. EW Max:.5543cm/s Obs. EW Max:.5163cm/s Syn. EW Max:.5943cm/s Obs. EW Max:.5163cm/s Qs : infinity -.9 Syn. EW Max:.6297cm/s Obs. EW Max:.5163cm/s Syn. EW Max:.852cm/s Obs. EW Max:.5163cm/s Figure 6. Continued Syn. UD Max:.1199cm/s Obs. UD Max:.1335cm/s Qs = Vs/ Syn. UD Max:.136cm/s Obs. UD Max:.1335cm/s Syn. UD Max:.1617cm/s Obs. UD Max:.1335cm/s Syn. UD Max:.2217cm/s Obs. UD Max:.1335cm/s Syn. UD Max:.2361cm/s Obs. UD Max:.1335cm/s Qs : infinity -.4 Syn. UD Max:.373cm/s Obs. UD Max:.1335cm/s Syn. UD Max:.1536cm/s Obs. UD Max:.1842cm/s Qs = Vs/ Syn. UD Max:.157cm/s Obs. UD Max:.1842cm/s Syn. UD Max:.2431cm/s Obs. UD Max:.1842cm/s Syn. UD Max:.3368cm/s Obs. UD Max:.1842cm/s Syn. UD Max:.3582cm/s Obs. UD Max:.1842cm/s Qs : infinity -.5 Syn. UD Max:.4662cm/s Obs. UD Max:.1842cm/s

10 Syn. NS Max:.522cm/s Obs. NS Max:.5743cm/s Qs = Vs/ Syn. NS Max:.5213cm/s Obs. NS Max:.5743cm/s Syn. NS Max:.5468cm/s Obs. NS Max:.5743cm/s Syn. NS Max:.567cm/s Obs. NS Max:.5743cm/s Syn. NS Max:.5624cm/s Obs. NS Max:.5743cm/s Qs : infinity -.7 Syn. NS Max:.574cm/s Obs. NS Max:.5743cm/s TYN Syn. EW Max:.1661cm/s Obs. EW Max:.2971cm/s Qs = Vs/ Syn. EW Max:.1879cm/s Obs. EW Max:.2971cm/s Syn. EW Max:.2129cm/s Obs. EW Max:.2971cm/s Syn. EW Max:.2262cm/s Obs. EW Max:.2971cm/s Syn. EW Max:.2285cm/s Obs. EW Max:.2971cm/s Qs : infinity -.4 Syn. EW Max:.24cm/s Obs. EW Max:.2971cm/s Figure 6. Continued Syn. UD Max:.576cm/s Obs. UD Max:.715cm/s Qs = Vs/ Syn. UD Max:.559cm/s Obs. UD Max:.715cm/s Syn. UD Max:.571cm/s Obs. UD Max:.715cm/s Syn. UD Max:.575cm/s Obs. UD Max:.715cm/s Qs : infinity -.1 Syn. UD Max:.575cm/s Obs. UD Max:.715cm/s Syn. UD Max:.579cm/s Obs. UD Max:.715cm/s CONCLUSION We improved the three-dimensional sedimentary basin structure model for the northern area of the Osaka basin by the forward modeling for direct S-waves as well as the later phases. Furthermore, we estimated the optimum anelastic attenuation factor from the fittings in the amplitudes of the later phases. Resultantly, we estimated Qs=Vs/5 as the optimum Qs. However, the fittings of the waveforms are not necessarily good except for amplitudes. To increase the fittings, we need to do more improvements for the shape of the basin edge and the physical parameters for each layer. ACKNOWLEDGMENTS This work is partially supported by the Special Project for Earthquake Disaster Mitigation in Urban Areas by the Ministry of Education, Culture, Sports, Science and Technology, Japan. We used strong motion records observed by the Committee of Earthquake Observation and Research in Kansai Area (CEORKA), and F-net and KiK-net data of the National Institute for Earth Science and Disaster Research (NIED), Japan. The authors would like to thank these organizations. REFERENCES 1. Kagawa, T., S. Sawada, Y. Iwasaki, and A. Nanjo (1993). Modeling of deep sedimentary structure of the Osaka Basin, Proc. 22nd JSCE Eqrthq. Eng. Symp., (in Japanese with English abstract).

11 2. Miyakoshi, K., T. Kagawa, B. ZHAO, T. Tokubayashi and S. Sawada (1999). Modeling of deep sedimentary structure of the Osaka Basin (3), Proc. 25th JSCE Eqrthq. Eng. Symp., (in Japanese with English abstract). 3. Zhao, B., M. Tsurugi and T. Kagawa (24). Strong motion simulation for large subduction earthquake, WCEE (in press). 4. Horikawa, H., K. Mizuno, K. Satake, H. Sekiguchi, Y. Kase, Y. Sugiyama, H. Yokota, M. Suehiro and A. Pitarka (22). Three-dimensional subsurface structure model beneath the Osaka Plain, Annual Report on Active Fault and Paleoearthquake Researches, Geological Survey of Japan/AIST, No. 2 (in Japanese with English abstract). 5. Pitarka, A., K. Irikura, T. Iwata, and H. Sekiguchi (1998). Three-dimensional simulation of the nearfault ground motion for the 1995 Hyogo-ken Nanbu (Kobe), Japan, earthquake, Bull. Seism. Soc. Am., 88, Cerjan, C., D. Kosloff, R. Kosloff, and M. Reshef (1985). A nonreflecting boundary condition for discrete acoustic and elastic wave equations, Geophysics 5, Clayton, R. and B. Engquist (1977). Absorbing boundary conditions for acoustic and elastic wave equations, Bull. Seism. Soc. Am. 67, Graves, R. W. (1996). Simulating seismic wave propagation in 3D elastic media using staggeredgrid finite differences, Bull. Seism. Soc. Am. 86, Kawabe, H., M. Horike and K. Kusakabe (21). Seismic Source Inversion Method Using Genetic Algorithm, Memories of Graduate School of Science and Technology, Kobe University, No.19-A. 1. Sekiguchi, H, K. Irikura and T. Iwata (2). Fault Geometry at the Rupture Termination of the 1995 Hyogo-ken Nanbu Earthquake, Bull. Seism. Soc. Am. 9,