Statistical analysis of temporal variations of seismicity level in some Italian regions

Rotondi and Pagliano considered the problem of identifying periods with different seismicity levels as a multiple-changepoint problem: after having estimated the complete part of the historical catalog related to a seismic source zone, inference was made about the number and the location of k greater than or equal to 1 changepoints of the occurrence rate in a sequence of n observations from a generalized Poisson process through a stepwise procedure. In this work, we extend such a method estimating the number of changepoints, their location, and other model parameters simultaneously. A fundamental role is played by new computational methods based on the simulation of Markov chains. In particular, we have applied a new version of the reversible jump Metropolis-Hastings algorithm to the data of some Italian seismic sources and we have identified periods with alternately higher and lower seismic activity levels.