Leveraging a Multi-Task Deep Learning Model to Enhance the California Statewide Earthquake Focal Mechanism Catalog

Junhao Song, Weiqiang Zhu, & Bo Rong

Submitted September 7, 2025, SCEC Contribution #14964, 2025 SCEC Annual Meeting Poster #TBD

Earthquake focal mechanisms, particularly those from smaller magnitude events that occur more frequently, can provide valuable information on subsurface fault geometries and stress fields. They have been better constrained with the development of dense seismic networks, but it also underscores the need for more efficient processing methods for data mining or timely monitoring purposes. A well-developed multi-task deep-learning model (Zhu et al., 2025) specialized in both phase arrival and polarity picking has successfully enhanced the focal mechanism solutions for the 2019 Ridgecrest earthquake sequence. In this work, we extend the application from local scale to statewide scale in California. We applied the model to continuous records from more than 1,200 seismic stations from 2021 to 2023. We associated P- and S-phase arrivals with individual events and derived the absolute locations of well-recorded events using adaptive 1D velocity models. The double-couple focal mechanisms were then computed with first-motion polarities. We successfully retrieved the majority of the earthquakes listed in the routine catalog and extended focal mechanism solutions to lower-magnitude earthquakes—for example, in the Geysers area and the 2022 Ferndale earthquake sequence. Our results are also consistent with existing refined catalogs (e.g., Cheng et al., 2025), which incorporate additional constraints such as relative amplitude ratios. Our deep-learning-based workflow can be applied to both long-term archived seismic datasets and real-time focal mechanism monitoring.

Citation
Song, J., Zhu, W., & Rong, B. (2025, 09). Leveraging a Multi-Task Deep Learning Model to Enhance the California Statewide Earthquake Focal Mechanism Catalog. Poster Presentation at 2025 SCEC Annual Meeting.

Related Projects & Working Groups
Seismology