Modeling of Wave Propagation in the Osaka Sedimentary Basin during the 2013 Awaji Island Earthquake (Mw5.8)

The three-dimensional velocity structure model for the Osaka sedimentary basin, southwest Japan is developed and improved based on many kinds of geophysical explorations for decades (e.g., Kagawa et al., 1993; Horikawa et al., 2003; Iwata et al., 2008). Recently, our project (Sekiguchi et al., 2013) developed a new three-dimensional velocity model for strong motion prediction of the Uemachi fault earthquake in the Osaka basin considering both geophysical and geological information by adding newly obtained exploration data such as reflection surveys, microtremor surveys, and receiver function analysis (hereafter we call UMC2013 model) . On April 13, 2013, an inland earthquake of Mw5.8 occurred in Awaji Island, which is close to the southwestern boundary of the aftershock area of the 1995 Kobe earthquake. The strong ground motions are densely observed at more than 100 stations in the basin. The ground motion lasted longer than four minutes in the Osaka urban area where its bedrock depth is about 1-2 km. This long-duration ground motions are mainly due to the surface waves excited in this sedimentary basin whereas the magnitude of this earthquake is moderate and the rupture duration is expected to be less than 5 s. In this study, we modeled long-period (more than 2s) ground motions during this earthquake to check the performance of the present UMC2013 model and to obtain a better constraint on the attenuation factor of sedimentary part of the basin. The seismic wave propagation in the region including the source and the Osaka basin is modeled by the finite difference method using the staggered grid solving the elasto-dynamic equations. The domain of 90km×85km×25.5km is modeled and discretized with a grid spacing of 50 m. Since the minimum S-wave velocity of the UMC2013 model is about 250 m/s, this calculation is valid up to the period of about 1 s. The effect of attenuation is included in the form of Q(f)=Q0(T₀/T) proposed by Graves (1996). A PML is implemented in the side and bottom of the domain. It is parallelized by the MPI and OpenMP. We computed for 120000 steps with a time step of 0.0025 s, which equals to 300s from the origin time. The source is represented by a point source having the focal mechanism determined by the F-net, NIED and the duration of its source time function is set to 3.1s referring to the waveform fitting at the rock stations outside the basin. In the previous studies for the Osaka basin (Horikawa et al., 2003; Kawabe and Kamae, 2006; Iwaki and Iwata, 2011), Q₀ at a reference period T₀ is given by a function of S-wave velocity; Q₀=αVs. We fixed T₀ at 5s, and tested Q₀ value changing α from 0.1 to 1.0. Comparing the envelope of synthetic velocity waveforms with that of observed waveforms in the basin, α of 0.3 fits to the observation well, whereas the difference from 0.2 to 0.5 is not significant. The simulation explains the characteristics of observed seismic waves propagating inside the basin in terms of duration and amplitude at most stations. The response velocity spectra and dominant period would be demonstrated to see the areal performance of present velocity model (UMC2013), and we will discuss how later phases generates and propagates based on the simulated wave field. Acknowledgements: We used strong motion data from K-NET/KiK-net, JMA, Osaka Prefectural Government, BRI, and CEORKA.