Simulation-Based Hazard Assessment for Long-Period Ground Motions of the Nankai Trough Megathrust Earthquake

Wednesday, 17 December 2014
Takahiro Maeda1, Nobuyuki Morikawa1, Asako Iwaki1, Shin Aoi1 and Hiroyuki Fujiwara2, (1)NIED National Research Institute for Earth Science and Disaster Prevention, Tsukuba, Japan, (2)Natl Res Inst Earth Science, Tsukuba Ibaraki, Japan
We evaluate a long-period ground motion hazard for the Nankai Trough earthquakes (M8~9) in west Japan. Past large earthquakes in the Nankai Trough that have occurred in an interval of 100~200 years showed various occurrence patterns and caused serious damages due to strong ground motion and tsunami. However, such large interplate earthquake potentially causes damages due to long-period ground motion even at long-distance basins. For evaluating the long-period ground motion of large earthquakes, it is important to take into account the uncertainty of source model and the effect of 3-D underground structure. In this study, we evaluate the long-period ground motion by the finite difference method (FDM) using "characterized source models" and the 3-D underground structure model. We construct various characterized source models (369 scenarios). Although most of parameters of the model are determined based on the "recipe" for predicting strong ground motion, we assume various possible source parameters including rupture area, asperity configuration, and hypocenter location. To perform the large-scale simulation for many source models, we apply a 3-D FDM scheme using discontinuous grids and utilize the GPGPU for our simulation. We use the system called GMS (Ground Motion Simulator) for the FD simulation. The grid spacing for the shallow region is 200 m and 100 m in horizontal and vertical, respectively. The grid spacing for the deep region is three times coarser. The total number of grid points is about 3.2 billion, which is about the eighth in the case of using uniform grids. We use GMS adapted for multi GPU simulation on the supercomputer TSUBAME operated by Tokyo Institute of Technology. Simulated peak ground velocity (PGV) and velocity response spectra (Sv) are strongly affected by the hypocenter location and show a large variation up to 10-fold at each site even in a group that have the same source area. We evaluate hazard curves and maps for PGV and Sv using the simulation results by assuming the occurrence probability of each scenario. The hazard maps for lower conditional exceedance probability clearly show that the long-period ground motions are amplified particularly at sedimentary basins. Our future study will consider the uncertainty of the underground structure model and the source modeling not following the recipe.