Spatial-temporal seismicity analysis using TSOM and variational density peak clustering

Abstract Seismicity de-clustering is a crucial step in earthquake catalog analysis, essential for understanding earthquake patterns and assessing seismic hazards. Seismicity de-clustering is challenging due to complex geological structures, high spatial-temporal correlation between events, and large amounts of noise. This study proposes an innovative two-stage approach for spatial zone identification and seismicity de-clustering by leveraging the topological self-organizing map (TSOM) method and the variational density peak clustering (VDPC) algorithm. In the first stage, the TSOM method is employed with split-and-merge algorithm to identify spatial interactions in space domain to uncover intricate spatial relationships among earthquake events in two-dimensional space, allowing the effective separation of distinct seismic zones. In the second stage, temporal separation is performed using the VDPC algorithm to distinguish mainshocks from aftershocks further within each seismic zone. This two-stage model enhances the precision of seismicity de-clustering and provides a comprehensive understanding of earthquake catalog dynamics. The results obtained from the proposed model outperform the other benchmark de-clustering algorithms, validated using various statistical tests, including the coefficient of variation, m -Morisita index, cumulative plot, $$\lambda$$ <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"> <mml:mi>λ</mml:mi> </mml:math> -plot and nearest neighbor distance. This innovative methodology enhances our comprehension of seismicity patterns, facilitates more accurate seismic hazard assessment, and ultimately contributes to improved earthquake preparedness in regions susceptible to seismic activity.