Automated Detection, Classification, and Migration Analysis of Earthquake Sequences: Applications to California and Japanese Subduction Zones

Machine learning has dramatically expanded earthquake catalogs, but efficiently extracting meaningful patterns from these datasets remains challenging. We present an automated workflow integrating clustering detection, sequence classification, and migration analysis with minimal manual intervention.

Our framework combines two clustering algorithms to identify spatiotemporal earthquake groupings, classifies sequences based on characteristic features, and detects migration patterns.

We apply this workflow to California catalogs (Southern California relocated catalog, Northern California catalog, and QTM template-matching catalog) and Japanese subduction zone catalog based on the S-net seafloor observatory network. Results demonstrate robust identification of diverse sequence types across different tectonic settings and spatial scales. Migration analysis reveals widespread fluid-driven characteristics in California earthquake swarms and potential fluid activity in the forearc region of the Japanese subduction zone.

This automated approach provides consistent, reproducible results while uncovering patterns potentially missed in manual analysis. The workflow enables rapid characterization of seismic sequences, which can improved seismic hazard assessment in tectonically active regions.