Hypocentre determination with prior information for clustering earthquakes

Abstract

We have developed and tested a conventional Bayesian algorithm for estimating the locations of clustering earthquakes. The main purpose of this algorithm is to provide a tool for selecting the most probable fault plane among several possible candidates, based on lineaments of aftershock hypocentres. An additional benefit is that the proposed procedure enables us to use any current location algorithm and to incorporate various prior distributions. Our algorithm requires only a set of the normal equation and the final location for each earthquake as observation data. We assume that refinement of the hypocentre is limited within a range where the normal equation varies no significantly. The refinement can be solved as a usual penalized least-square problem. Hyperparameters of the prior distribution can be optimized with the Akaike Bayesian Information Criterion. For example, we assume that aftershocks may originate near the main shock fault, on which a simple prior distribution is introduced. This procedure is applied to both a set of simulated earthquakes and a set of aftershocks occurring in an offshore area. The proposed method can reasonably revise hypocentres for the simulation data. We consider two different prior distributions of aftershocks, which conform to two nodal planes of the main shock. The optimum solution suggests one fault plane as more probable than the other with a high level of significance. © 2010 The Author Journal compilation © 2010 RAS.