[S-SS10] 地震発生の物理・断層のレオロジー

[SSS10-P06] Dynamic Rupture Inversion of Finite-fault with Curved Grid Finite-Difference Method: Synthetic Test and Applications

*Yuchen Zhang¹、Zhenguo Zhang¹、Zhongqiu He¹ (1. Department of Earth and Space Sciences, Southern University of Science and Technology, Shenzhen 518055, China)

キーワード：Earthquake Source Dynamic Inversion, Curved Grid Finite-Difference Method, Finite Fault Model, Parallel MCMC Algorithm

Dynamic source inversion can provide essential physical parameters of the fault, which enhances the resolution of seismic wave simulations and aids in risk analysis for engineering seismology, earthquake disaster loss assessment, and earthquake forecasting. However, such a task remains challenging due to its high computational demands, and when the fault geometry in the inversion differs significantly from the actual situation, the uncertain model prediction errors will exacerbate this burden and impact. We propose a novel Bayesian dynamic inversion method to infer stress and frictional parameters on faults. In the inversion, we utilize the GPU-accelerated curved grid finite difference method (CG-FDM) and an improved parallel-MCMC method to derive the model's posterior probability density function. We conduct synthetic inversion tests using the TPV8 and TPV10 models from the SCEC/USGS Spontaneous Rupture Code Verification Project. The results demonstrate that our inversion method converges to stable solutions with acceptable margins of error for both vertical strike-slip faults and moderately inclined normal faults. Subsequently, we applied this method to real earthquake events by constructing a non-planar fault model and incorporating the b-value distribution as a Bayesian prior constraint. Using this approach, we inverted the spatial distribution of fault friction coefficients and initial shear stress for the 2017 Mw 6.5 Jiuzhaigou earthquake, ultimately developing a three-dimensional dynamic rupture model.