A Bayesian Inference Method for a Large Magnitude Event in a Spatiotemporal Marked Point Process Representing Seismic Activity

Abstract

A Bayesian method to forecast the occurrence time of a large-scale earthquake utilizing temporal information on earthquakes with smaller magnitudes was proposed in our recent study for a marked point process that simulates seismic activity. In this paper, we show the extension of this Bayesian approach in the spatiotemporal marked point process, aiming to yield a forecasting method for both the occurrence time and location of the next large earthquake. We particularly discuss the contribution of the correlations between the spatial position and the inter-event time interval at different magnitude scales to probabilistic forecasting.