抄録
We have simulated long-period ground motions in three 3D subsurface structural models of southwestern part of the Kanto plain for an M4.1 earthquake in the west of the Kanagawa prefecture, Japan with a finite-difference modeling to compare their performances in strong motion prediction. The three 3D models are constructed from the seismic refraction data (KOKETSU, 1995) and mainly from the geological data (SUZUKI, 1999), and the compiled results of microtremor array measurements (YAMANAKA et al., 1999). The major differences of these models are configurations of the sedimentary layers and the basement topographies, especially, the depth to the top the basement of the model from the microtremor array measurements is shallower than the other models.
First, we estimated source parameters by a fitting of synthetic waveforms with observed records at a rock site. The synthetic motions for the models were significantly different at sites in the basin area. The model from the microtremor array measurements was better in reconstructing later phases than the other two models. Because the model from the microtremor array measurements has the low-velocity sedimentary layers around Yokohama, and contains an up-lifted structure of the basement around the Miura peninsula that would be responsible to generation of the later phases. It was suggested that misfit of synthetic motions with observed one are useful information for constructing of the 3D-earth model for the simulation of ground motion.
