Further applications of the stochastic stress release model to historical earthquake data

Abstract

The stochastic stress release model provides a stochastic version of the elastic rebound model which may be used to fit, simulate and forecast occurrence of large earthquakes within a seismic region. This paper reviews its applications to historical earthquakes from China, Iran and Japan, thereby extending and supporting earlier work on the Chinese historical data. The model generally provides an improvement in fit over Poisson models with either constant rate or linear trend. A crucial factor is the identification of suitable subregions which can act as largely independent seismic units. In northern China, divisions into two and four such units are considered based on recognized geophysical subdivisions. In Iran, the best results are obtained from a subdivision into three such units. The situation in Japan seems more complex: a preliminary model with two units gives a significant improvement of fit in one out of the two cases. In all regions treating the subregions separately gives marked improvements over treating the whole region as a single unit. However, the quality of fit within subregions differs sharply from case to case, pointing to substantial differences in the seismic regimes as reflected in the historical data.