Deep Learning Enhanced Earthquake Catalog for Northern California

Ian W. McBrearty, Weiqiang Zhu, Bo Rong, & Gregory C. Beroza

Published September 8, 2024, SCEC Contribution #14056, 2024 SCEC Annual Meeting Poster #069 (PDF)

Developing enhanced earthquake catalogs is essential for nearly all aspects of seismology, including understanding foreshock and aftershock behavior, imaging the Earth structure, and identifying active faults. Here we present progress on a deep learning enhanced catalog for northern California from 2023, where we use the PhaseNet picker and the GENIE phase association algorithm to develop a catalog of ~4.2x more earthquakes than the routine NCEDC catalog. The set of seismic stations we use to develop this catalog is highly variable, with hundreds of stations throughout the bay area, and far fewer stations throughout the central valley and Sierras, as well as dense local networks at Parkfield and Geysers. Despite this heterogeneity, the use of GENIE association appears effective at associating picks reliably across the full spatial domain, spanning from south of Parkfield and Ridgecrest, to north of the Mendocino Triple junction, and from the west coast into western Nevada. Our results are further verified by confirming that ~95% of all reported NCEDC earthquakes are re-detected, and the spatial locations of new events improves the detected seismicity rate primarily at the expected fault locations.

Key Words
Earthquake Detection

Citation
McBrearty, I. W., Zhu, W., Rong, B., & Beroza, G. C. (2024, 09). Deep Learning Enhanced Earthquake Catalog for Northern California. Poster Presentation at 2024 SCEC Annual Meeting.

Related Projects & Working Groups
Seismology