Shared Nearest Neighborhood Intensity Based Declustering Model for Analysis of Spatio-Temporal Seismicity

Categorization of seismic events into correlated aftershocks (triggered by the mainshocks) and independent backgrounds (generated due to regular movements of tectonic plates) is essential to carry out reliable hazard analysis in a region of interest. In this paper, a shared nearest neighborhood intensity based declustering (SNN-IBD) model is proposed to categorize seismic events based on their magnitude, event location, and occurrence time. In this approach, events which lie within a spatial-cutoff (epsilon(s)) and temporal-cutoff (epsilon(t)) are considered as core points. Instead of using density (a significant number of events) in a space-time window, intensity (magnitude) indicated by core points is considered here in order to discover aftershock clusters. Effective selection of cutoffs (epsilon(s), epsilon(t), intensity/magnitude threshold S-M) and classification accuracy in spatio-temporal domain are validated using statistical parameters: Coefficient of Variation (COV) and m-Morisita index. The regional earthquake catalogs of the Philippines (19732012) and Iran (1966-2015) are analyzed using the proposed model. From the simulation studies, it is observed that background seismicity follows a homogeneous Poisson process in the time domain. In the spatial domain, background seismicity reflects multifractal behavior similar to true events of the catalog. The superior performance of the proposed method is demonstrated over tetra-stage clustering model and benchmark declustering algorithms.