Real-Time Earthquake Monitoring with Deep Learning: A Case Study of the 2025 ML 6.4 Dapu Earthquake and Its Fault System in Southwestern Taiwan

Abstract

The ML 6.4 Dapu earthquake that struck southwestern Taiwan on 20 January 2025 pro-vides a critical case for understanding the seismogenic mechanisms and fault systems in this tectonically active region. This study applies a deep-learning-based real-time micro-earthquake monitoring system designed for Taiwan to analyze 3893 aftershocks recorded within 15 days of the mainshock. By incorporating a 3D velocity model and the NonLinLoc earthquake location method, we assess the fault systems associated with the earthquake sequence. Focal mechanisms of 10 M 4.8+ events from the Broadband Array in Taiwan for Seismology catalog are also utilized to interpret fault types. Our results suggest that the Dapu earthquake is linked to an upper crustal fault system, distinct from the surface fault structures. The study further evaluates the potential contribution of both thin-skinned and thick-skinned deformation models for the region. The findings emphasize the signifi-cance of real-time seismic monitoring combined with the 3D velocity earthquake location method in enhancing earthquake location accuracy and deepening our understanding of fault systems.