Machine-Learning catalog for the California-Baja California, Mexico, border region

Erik E. Ramírez-Ramos, Joann M. Stock, Zachary E. Ross, & Antonio Vidal-Villegas

Published September 11, 2022, SCEC Contribution #12045, 2022 SCEC Annual Meeting Poster #011 (PDF)

The region of the southernmost California-Northern Baja California, Mexico, presents significant seismicity, generated mainly by the San Andreas-Gulf of California fault system. This system has caused earthquakes of magnitudes Mw 6.5, 1979 Imperial Valley earthquake and the Mw 7.2, El Mayor-Cucapah, 4 April 2010. This last earthquake ruptures the unmapped Indiviso fault, a sediment-covered fault. To have a detailed catalog of fault structures, we analyzed the waveforms of the permanent and temporary broadband seismic stations between 2012 and 2020. For northern Baja California, the seismic stations belong to the Northwest Mexico Seismic Network (RESNOM) and stations in southernmost California to the Southern California Seismic Network (SCSN). The yearly-based analysis consisted of applying the deep learning techniques of the PhaseLink system to identify P- and S-arrivals and their association to the combined database. The location was done with Non-Linear Location and relocated with HypoDD. Due to large crustal velocity and thickness variations, the relocation process was separated into two areas: the Mexicali Valley and the Peninsular Rages of Baja California (PRBC), using the appropriate velocity model for each region. Our catalog generally has 2-3 times more events than those reported by RESNOM. For the Mexicali Valley region, profiles of seismic alignments show more constrained seismicity (between ~5 to ~10 km depth) when comparing them with the SCSN and RESNOM catalogs. This is especially notable in the Indiviso, Dixieland, and Laguna Salada-Indiviso faults. Three seismicity profiles for the PRBC region also show better constraints than the RESNOM catalog (the SCEC catalog is unavailable for this region). Having a precise-hypocenter complete catalog and the analyzed seismicity profiles for the California's border region will allow us to improve and expand the active fault geometries to be included in the Community Fault Model.

Key Words
Machine-learning, US-Mexico seismicity,

Citation
Ramírez-Ramos, E. E., Stock, J. M., Ross, Z. E., & Vidal-Villegas, A. (2022, 09). Machine-Learning catalog for the California-Baja California, Mexico, border region. Poster Presentation at 2022 SCEC Annual Meeting.

Related Projects & Working Groups
Seismology