Spatio-temporal earthquake clustering: insights and outlooks from Network Analysis

The seismic history of a region is characterized by its earthquake clusters, namely periods when the occurrence rate of earthquakes is higher than usual. Clustering in space and time is an essential key to understanding earthquake source mechanisms (fault geometry, rupture dynamics, status of the stress field, etc.), and several methodologies for cluster analysis have been proposed so far. However the definition of clusters is not univocal. Thus, for the identification of earthquake clusters we consider two recent data-driven declustering algorithms, one based on nearest-neighbor distance and the other on a self-exciting point process. Since different classifications of earthquakes into main and secondary events can be obtained from different methods, we compare their performance by exploiting tools from Network theory. In particular, in order to highlight possible classification similarities/dissimilarities, the earthquake clusters obtained from both algorithms are represented as rooted trees, and their complexity is evaluated and compared through suitable centrality measures.