地震学における応力インバージョンの新展開

Abstract

The stress inversion is a statistical method to estimate tectonic stress fields from a set of observed data. The methodology of stress inversion, originally developed in geology, is based on the Wallace-Bott hypothesis; the direction of fault slip is parallel to the resolved stress vector on a preexisting fault. In the conventional methods based on this hypothesis, fault slip data have been used as observed data. Since the constitutive property of the source region is unknown, we cannot relate fault slip with stress by means of strain. This is an essential problem in the fault slip data inversion. The conventional methods have another technical but inherent problem that inverted stress patterns depend on the way of area partitioning. The method of CMT data inversion using ABIC resolved both the essential and the technical problems in the fault slip data inversion. In this method, instead of fault slip data, we use CMT data of seismic events, which can be directly related with stress fields without any knowledge of actual processes in the source region. Hierarchic, flexible Bayesian modeling with hyper-parameters and rational model selection with ABIC enable us to resolve the model dependence of inverted stress patterns.