An Algorithm for Earthquakes Clustering based on Maximum Likelihood

In this paper we propose a clustering technique set up to separate andfind out the two main components of seismicity: the background seismicity and thetriggered one. We suppose that a seismic catalogue is the realization of a non homogeneousspace-time Poisson clustered process, with a different parametrizationfor the intensity function of the Poisson-type component and of the clustered (triggered)component. The method here proposed assigns each earthquake to the clusterof earthquakes, or to the set of independent events, according to the increment to thelikelihood function, computed using the conditional intensity function estimated bymaximum likelihood methods and iteratively changing the assignment of the events;after a change of partition, MLE of parameters are estimated again and the processis iterated until there is no more improvement in the likelihood.