Improved seismic velocity structure in southwestern Japan using pronounced sP phase

In southwestern Japan the Philippine Sea plate (PHSP) subducts along the Nankai trough and this subduction causes the megathrust earthquakes in the Nankai seismic zone as well as large intraslab and inland earthquakes in the vicinity. The dip angle of the PHSP varies significantly along strike. In this region the sP phase is widely observed and its amplitude sometimes becomes larger than that of the direct S wave. This suggests that the phase could control the peak ground velocities and be a significant factor in evaluating regional seismic hazards. We previously showed that the arrivals of this strong phase are explained by incorporating the structure with shallow bedrock depths in the velocity model, and presented consistent 2D seismic velocity models. The 2D models derived for profiles between earthquake hypocenters and observation stations represent the configuration of the subducting PHSP and the depth variation of the Conrad and Moho discontinuities (Hayashida et al., 2010). Here, we present a 3D seismic velocity model that accounts for observed waveforms at a number of local stations where the pronounced sP phase was recorded. We present the results of finite difference modeling of the observed seismograms for an intraslab earthquake (2001/3/26 Mw5.1, h = 46 km) using the e3d code (e3d; Larsen and Schultz, 1995). First we calculated synthetics using a simple layered structure (Asano et al., 1986) to fit the arrival times and amplitudes of sP phases, and then matched the P- and S-wave arrivals by tuning the slab configurations and the Conrad and Moho depths. At an early stage of the 3D modeling we referred to the travel time tomography model (Nakajima and Hasegawa, 2007) and receiver function images (e.g. Shiomi et al., 2006; Ueno et al., 2008). The results show that the dip angle of the PHSP beneath the region varies significantly along the trench strike, and the agreement between data and synthetics varies among the models. At stations located to the north and west of the epicenter, the synthetics calculated with the previous layered model generally agree with the data in the frequency range between 0.1 and 0.5 Hz. On the other hand, at stations located to the east and south of the epicenter, the agreement was improved by including the slab configuration in the structure, since P and S waves propagate within the slab in these directions. In addition, at stations located on the thicker subsurface soil layer (~400 m), the reproduction of sP phase amplitudes and coda durations were improved by incorporating the subsurface structure model by the National Research Institute for Earth Science and Disaster Prevention (2010). The constructed velocity model with detailed features associated of the slab and crust configuration provides a better assessment of strong ground motions in this region.