Simultaneous Inverse Analysis Method of Fault Slip and Asthenosphere Viscosity Using Large Scale Finite Element Simulation of Postseismic Deformation
Abstract:Inverse analysis of the coseismic/postseismic slip using postseismic deformation observation data is an important topic in geodetic inversion. Inverse analysis method may be improved by using numerical simulation (e.g. finite element (FE) method) of viscoelastic deformation, the model of which is of high-fidelity to the available high-resolution crustal data. The authors had been developing a large-scale simulation method using such FE high-fidelity models (HFM), assuming use of K computer, the current fastest supercomputer in Japan. In this study, we developed an inverse analysis method incorporating HFM, in which the asthenosphere viscosity and fault slip are estimated simultaneously, since the value of viscosity in the simulation is not trivial.
We carried out numerical experiments using synthetic crustal deformation data. Based on Ichimura et al. (2013), we constructed an HFM in the domain of 2048x1536x850 km, which includes the Tohoku region in northeast Japan. We used the data set of JTOPO30 (2003), Koketsu et al. (2008) and CAMP standard model (Hashimoto et al. 2004) for the model geometry. The HFM is currently in 2km resolution, resulting in 0.5 billion degrees-of-freedom. The figure shows the overview of HFM. Synthetic crustal deformation data of three years after an earthquake in the location of GEONET, GPS/A observation points, and S-net were used. Inverse analysis was formulated as minimization of L2 norm of the difference between the FE simulation results and the observation data with respect to viscosity and fault slip, combining quasi-Newton algorithm with adjoint method. Coseismic slip was expressed by superposition of 53 subfaults, with four viscoelastic layers. We carried out 90 forward simulations, and the 57 parameters converged to the true values. Due to the fast computation method, it took only five hours using 2048 nodes (1/40 of entire resource) of K computer. In the future, we would like to also consider estimation of after slip and apply the method to the actual data.