Abstract
First we utilized Stochastic Green's functions obtained from strong motion data of K-NET, KiK-net, and the JMA Shindo-kei network in Japan to predict strong motions for a large subduction-zone earthquake. Then considering earthquake type and site characteristics, earthquake spectrum characteristics and propagation path effects were extracted. Then we used a conventional Green's function method to sum up statistical Green's functions for a moderate size earthquake and synthesize them to predict strong motions due to the expected Nankai earthquake. The resultant strong motions show similar PGA values of empirical relations and the calculated seismic intensities show similar values as observed in the previous two events.
Finally, we input these strong motions to dynamic nonlinear structural models. Our nonlinear structural models are unique because they are multiple models with different strengths with different existence ratios. From the theoretical calculations we can determine damage ratios for different types of building. For quantitative prediction we determined model parameters based on the damage statistics in Kobe after the Hyogo-ken Nanbu (Kobe) earthquake of 1995. We found that heavy damage to structures can only be found in the near-source region. We have very heavy damage ratios to low-rise steel structures, while for low-rise RC structures we have relatively small damage ratios. However, the damages in high-rise buildings were limited to those at near fault locations, and high-rise buildings in Osaka Plain did not have serious damages or plastic deformations.
