Occurrence of earthquakes in South India - a Bayesian approach

Abstract

In order to demonstrate the performance of Bayesian approach with respect to earthquake occurrence data, the probabilities of non occurrence of earthquakes with magnitude ranges 4-5 and 5-6 during 1996 to 2008 are determined for South India region. To determine these probability values, the earthquake occurrence data of 196 years from 1800-1996 in Tamil Nadu, Andhra Pradesh, Karnataka and Kerala regions is collected from different sources. Modeling of occurrence of earthquakes is done by using simple Poisson process, Bayesian approach by assuming a diffused prior, and a conjugate distribution (Gamma distribution) as prior distribution for mean recurrence rate. The evaluated probability values are compared with the data on occurrence of earthquakes, from 1996-2008. From these comparisons, it is inferred that Gamma prior is a suitable prior in making rational estimation of probability of occurrence of earthquakes. Based on this observation, predictions are made with respect to probabilities of occurrences of earthquake for next twelve years (i.e., 2008-2020) and conclusions are drawn by comparing the probability values of the regions considered regarding their relative risks.