[S-SS10] Strong Ground Motion and Earthquake Disaster

[SSS10-15] Investigation for relationships between strong ground motions and dynamic parameters through dynamic source inversion

*Takeshi Miyamoto¹, Kimiyuki Asano², Tomotaka Iwata² (1.Graduate School of Science, Kyoto University, 2.Disaster Prevention Research Institute, Kyoto University)

Keywords:dynamic source inversion, strong ground motion, dynamic parameter

Source models have been estimated for large earthquakes by, for example, kinematic source inversion technique using observed waveform data. Relationships between source process and characteristics of observed waveforms have also been discussed. The 2016 Kumamoto earthquake sequence includes two large events with JMA Intensity 7, M_JMA6.5 on April 14, 2016, and M_JMA 7.3 on April 16, 2016 (hereafter, we refer to the former event as the foreshock and the latter as the mainshock). Many studies have been conducted particularly for the mainshock. Tsuda (2021) and Kaneko and Goto (2022) attempted to construct dynamic source models that reproduced the waveforms recorded nearby the source fault through dynamic rupture simulations. In this study, a dynamic source inversion method is developed for estimating a dynamic source model which reproduces waveforms observed not only nearby the source fault but also far away from it (e.g., epicentral distance is larger than the fault length) by referring to a similar method proposed by Gallovič et al. (2019). This dynamic source inversion method mainly consists of two parts: generating waveforms and updating dynamic source models. Synthetic waveforms are obtained by convolving slip time functions derived from dynamic rupture simulation with Green's functions as in Gallovič et al. (2019). This method for generating synthetic waveforms reduces computational cost and allows us to introduce complex velocity structures between the source fault and stations. A new dynamic source model, which is perturbated randomly from the former model by using the Markov Chain Monte Carlo (MCMC) method, is proposed and is accepted or rejected depending on the result of the Metropolis test. An open-source code adopting the finite difference method (fd3d_TSN; Premus et al., 2020) was used for dynamic rupture simulations. Green's functions were calculated by using the discrete wavenumber method (Bouchon, 1981) and the reflection and transmission matrix method (Kennett and Kerry, 1979). Synthetic waveforms derived from convolutions of slip time functions and Green's functions and resultant time series of velocity obtained by dynamic rupture simulation were compared at five surface stations for validation. Waveforms derived by both methods were almost the same at all stations and Variance Reduction was 99%, which showed that our method could correctly give waveform data. A validation test of our method by using a simple dynamic source model and actual distribution of strong motion stations in Kumamoto Prefecture is now being conducted. Asano and Iwata (2016) assumed a nearly vertical plane as the source fault for the foreshock based on the aftershock distribution immediately after the occurrence of the foreshock and the moment tensor solution. Thus, one vertical fault plane with right lateral slip was assumed for this validation test. The position of the nucleation point was set with reference to Mitsuoka et al. (2020), which re-located the hypocenters by using the double-difference tomography method. Velocity model beneath each station used for calculation of Green's function was constructed by combining two models: the 3-D velocity model of Kumamoto Prefecture (Asano et al., 2019) and the Japan Integrated Velocity Structure Model (JIVSM; Koketsu et al., 2012). After this test, our dynamic source inversion will be applied to the strong motion data observed during the foreshock.