[S-SS11] 強震動・地震災害

[SSS11-P08] Subsurface Velocity Model Estimation Based on Microtremor Array Measurements in Eastern Taiwan

*CHOU YUTING¹、松島 信一²、長嶋 史明²、Rau Ruey-Juin³、Wu Cheng-Feng³、小穴 温子⁴、宮腰 研⁵ (1.京都大学大学院工学研究科、2.京都大学防災研究所、3.國立成功大學地球科學系、4.清水建設技術研究所、5.大崎総合研究所)

キーワード：20220918台東地震、常時微動、位相速度、水平上下スペクトル比、速度構造、地震基盤

In 2022, a local magnitude 6.8 earthquake occurred in southeastern Taiwan (20220918 Taitung earthquake). During this earthquake, velocity pulses and permanent displacements were observed by the stations of Taiwan Strong Motion Instrumental Program (TSMIP) and Central Weather Administration Seismographic Network (CWASN) surrounding the source area. Furthermore, surface ruptures were observed in the source area. For predicting long-period ground motion including the effect of permanent displacement of the surface rupture, a fault model that considers slip occurring in layers shallower than the seismic bedrock is necessary. In order to understand the relationship between the strong motion caused by the 20220918 Taitung earthquake, the source model, the surface rupture and the subsurface velocity model, three microtremor arrays were conducted around the strong motion stations G020, F042, and F073 (EYUL) from TSMIP. These stations were chosen because velocity pulses were observed during the earthquake, and surface ruptures occurred near these stations. First, microtremors were recorded for 5 to 21 hours with array geometries ranging from 20 to 3548 m in station pairs. Each array contains 21 to 22 stations. Then, high-resolution frequency-wavenumber method was applied to the microtremor recordings. As a result, Rayleigh wave phase velocities of approximately 0.4 to 3 km/s were obtained in a wide frequency range (0.45-9 Hz for G020 array, 1.7-10 Hz for F042 array, and 0.4-6 Hz for EYUL array). Next, microtremor horizontal-to-vertical spectral ratios (MHVR) were calculated and averaged within near-central stations for each microtremor array. Also, earthquake horizontal-to-vertical spectral ratios (EHVR) of the three strong motion stations were calculated. In order to ensure there is energy in low frequency while avoiding the nonlinear effect, records for events with local magnitude larger than 5 and peak ground acceleration lower than 75 cm/s/s were chosen. For each record, 40.96-s-time-window waveforms were trimmed from horizontal and vertical components from 2 s before S-arrival, and their spectral ratios were calculated. Then, the results of all chosen records were averaged. After that, inversions of the phase velocity and MHVR were performed for each microtremor array. Due to the lack of a priori information of the local structure, we performed a two-step inversion. First, inversion of the phase velocity was performed for each array. Then, joint inversion of the phase velocity and MHVR was performed. The best model (with minimum misfit) from inversion of the phase velocity for each array was taken as the initial model for the joint inversion. Finally, the velocity models obtained from joint inversion were tuned, considering the theoretical and observed EHVR. As a result, three velocity models down to depths of a few kilometers were obtained. Based on the results of these models, the depths of the seismic bedrock (Vs=3 km/s) for G020, F042, and EYUL are estimated to be 2402 m, 2140 m, and 2305 m, respectively, which are shallower than those in the previous study (Huang et al., 2014).