A Method to Evaluate Characteristics of Simulated Long-Period Ground Motion for Anticipated Nankai Trough Earthquakes

Abstract:

Since recent development of technologies enable us to simulate large-scale and large-number of ground motion, it is possible to not only focus on studying particular events in detail but also looking at many possible events in big picture. In this study, we apply the latter method to generate many possible source models using empirical relationships by taking into account the uncertainty of source parameter. Based on the ground motion simulation using those varieties of source model, impact by the uncertainty of source parameters on simulated ground motion can be reveal.

We have simulated long-period ground motions associated with the Nankai Trough earthquakes (M8~9) in southwest Japan at about 80,000 observation sites using various characterized source models (~400 models) by giving the various dimension of source parameters such as source region, asperity configuration, and hypocenter location (Maeda et al., 2014, AGU fall meeting). Using these results, we have been investigating the relationship between simulated ground motions and source parameters for selected sites based on the seismological knowledge. However, the different approach may be applied to analyze our simulated results. As a preliminary study, we apply the k-nearest neighbor algorithm to the estimated peak amplitudes. Then we can divide the 80,000 sites into several groups that have similar relationship between the source parameters and simulations.

Our future study will lead to several methods to be applied to our results and to investigate the characteristic features of simulated ground motion.