Three-Dimensional Fault Reconstruction from Earthquake Catalogues Based on the Density Peak Clustering Algorithm

Summary Earthquake catalogues contain information concerning active faults. Extracting this fault-related information contributes to seismic hazard assessment. We employ the Density Peak Clustering (DPC) algorithm with adaptive parameter selection and Principal Component Analysis (PCA) to analyze earthquake catalogues The DPC algorithm performs clustering analysis on the earthquake catalogues. Specifically, the cutoff distance parameter of the DPC algorithm is automatically determined by minimizing the entropy of local density information, and the number of cluster centers is automatically obtained by setting a threshold. The PCA method is then applied to identify planes from the classified clusters, yielding the corresponding planes and parameters; the identified planes are considered fault planes and fault segments. This fully automated workflow was applied to the Cahuilla region in the United States for validation. The geometric parameters of the constructed fault planes agreed well with those derived from focal mechanism inversion, validating the effectiveness of the proposed approach. Furthermore, this workflow is inspired by the HSF -ULA procedure. Therefore, a comparative analysis was conducted between the proposed Fault Construction with Adaptive Parameter Selection (FC-APS) method and the HSF-ULA method using the L’Aquila earthquake catalogues, highlighting the advantages and limitations of each approach, and discussed potential future research directions for the method.