A Modified Self-organizing Map with Mean-Shift Clustering for Seismicity Analysis of Earthquake Catalogs

The process of seismic de-clustering activity is highly important for comprehending the characteristics and trends of earthquakes, which can help in the development of earthquake prediction models, hazard assessment, and understanding of the earth’s layer for future seismic activity. De-clustering seismicity is a complex task due to the high correlation of space-time components. Self-organizing map (SOM) has been extensively used for clustering and visualization of seismic data. However, traditional SOM suffers from limitations such as topological preservation and the inability to handle data with varying densities due to a fixed learning rate. In this paper, we propose a modified SOM algorithm that integrates mean-shift clustering to overcome these limitations. The proposed algorithm not only preserves the topological properties of the data but also enables the identification of clusters with varying densities. The effectiveness of the proposed algorithm is evaluated on two earthquake catalogs from California and Japan, and the results demonstrate that the proposed algorithm outperforms benchmark de-clustering techniques in identifying seismic clusters accurately. The proposed algorithm has the potential to contribute to seismic hazard assessment and early warning systems.