Validation of 3-D basin structure models for long-period ground motion simulation in the Osaka basin, western Japan

We studied the applicability of two types of existing three-dimensional (3-D) basin velocity structure models of the Osaka basin, western Japan for long-period ground motion simulations. We synthesized long-period (3–20 s) ground motions in the Osaka basin during a M6.5 earthquake that occurred near the hypothetical Tonankai earthquake source area, approximately 200 km from Osaka. The simulations were performed using a 3-D finite-difference method with nonuniform staggered grids using the two basin velocity structure models. To study the ground motion characteristics inside the basin, we evaluated the wave field inside the basin using the transfer functions derived from the synthetics at the basin and a reference rock site outside the basin. The synthetic waveforms at the basin site were obtained by a convolution of the calculated transfer function and the observed waveform at the reference rock site. First, we estimated the appropriate Q values for the sediment layers. Assuming that the Q value depends on the S wave velocity V _S and period T, it was set to Q = (1/3V _S)(T ₀/T) where V _S is in m/s and the reference period T ₀ is 3.0 s. Second, we compared the synthetics and the observations using waveforms and pseudovelocity response spectra, together with a comparison of the velocity structures of the two basin models. We also introduced a goodness-of-fit factor to the pseudovelocity response spectra as an objective index. The synthetics of both the models reproduced the observations reasonably well at most of the stations in the central part the basin. At some stations, however, especially where the bedrock depth varies sharply, there were noticeable discrepancies in the simulation results of the models, and the synthetics did not accurately reproduce the observation. Our results indicate that the superiority of one model over the other cannot be determined and that an improvement in the basin velocity structure models based on simulation studies is required, especially along the basin edges. We also conclude that our transfer function method can be used to examine the applicability of the basin velocity structure models for long-period ground motion simulations.