Unveiling hidden seismicity: An AI-derived catalog of 380k earthquakes for Sichuan Province, China
2026-09-09
Long-term, high-resolution seismic monitoring is a prerequisite for characterizing seismic structures, delineating fault geometry, and mitigating earthquake risks. Located on the eastern margin of the Qinghai-Xizang Plateau, Sichuan Province is among the most seismically active regions in China, with particularly intense activity concentrated in the shale gas extraction zones of the southern Sichuan Basin. However, conventional manual earthquake catalogs are often incomplete with respect to microseismic events, limiting detailed seismological research. In this study, we constructed and released a high-resolution earthquake catalog for the Sichuan Province ranging from January 2019 to December 2020. Using continuous waveform data from 261 seismic stations, we adopted a deep learning-based workflow that involved phase picking, followed by event association and absolute location constrained by a high-resolution regional 3-D velocity model. The resulting catalog comprises 380,886 seismic events, approximately 5 times the number reported in the standard manual catalog for the same period. Completeness magnitude improved substantially, decreasing from M L 1.0 in the manual catalog to M L 0.1 in our catalog. This dataset provides a high-quality basis for subsequent studies in seismic activity analysis, seismic tomography, and hazard assessment.