Statistical analysis on background seismicity of Southern California region: application of nearest neighbour declustering and network analysis

Abstract

We analyse the background seismicity, including mainshocks and isolated events, from a distinct clustered component using the nearest-neighbour declustering method. After declustering the seismic catalog, two components were identified: background and clustered. The clustered component includes isolated networks, and for mainshock selection within each network, we applied outdegree and closeness centrality measures from network theory. This approach differs from the conventional method, which selects mainshocks from individual clusters network based on the highest magnitude. The background events dataset was obtained using the nearest-neighbour method and network analysis. This methodology was applied to the Southern California region, encompassing four significant events with magnitudes greater than 7, over the period 1981–2021. The primary objective is to assess the relationship between background seismicity and the Poisson process, as well as to identify the magnitude threshold at which it aligns with the Poisson model. To accomplish this, the background dataset was divided into specified magnitude ranges from 3 to 4.2, with intervals of 0.2. Temporal statistical tests, including the conditional chi-square test, Brown-Zhao test, and Kolmogorov–Smirnov test, were performed, while the Luen and Stark statistical test was applied for space–time analysis. For nearly all magnitude cut-offs, the temporal statistical tests reject the null hypothesis. The exception is at a magnitude of 3.4, where the temporal test is satisfied; however, the space–time statistical test still rejects the null hypothesis. However, the background dataset for the study region does not conform to the Poisson process in either the temporal or space–time tests across all magnitude thresholds. This inconsistency may be attributed to a limited number of data points at certain magnitude cutoffs, the declustering method used, or the potential need for an alternative conditional model for analysing background events.