← Back to Sessions

Seismotectonics of the Puerto Rico Subduction Zone: Leveraging Machine Learning Analysis of Ocean Bottom Seismometers

Description: 

The Puerto Rico Trench (PRT) is characterized by highly oblique subduction that transitions from predominantly perpendicular convergence in the eastern segment to increasingly oblique convergence toward the western segment. Despite its significance as a tectonic boundary and its history of generating large earthquakes and tsunamis—such as the 1918 Mw 7.3 San Fermín earthquake and the 1943 Mw 7.6 Mona Passage earthquake—the PRT remains insufficiently studied. The absence of a comprehensive seismic catalog and significant uncertainties in the locations of offshore earthquakes and swarms hinder a critical understanding of subduction dynamics and associated hazards. By leveraging automated machine learning (ML) pickers, this study integrates data from six ocean bottom seismometers (OBS) deployed north of Puerto Rico by the USGS to generate a comprehensive, high-resolution earthquake catalog.

Using SeisBench, we applied PhaseNet, a deep neural network-based seismic arrival picker (Zhu and Beroza, 2019), and the phase-associator GaMMA, in conjunction with the Pick-Blue model trained on OBS datasets (Bornstein et al., 2024), to analyze waveforms from both the ocean bottom seismometers (OBS) and the Puerto Rico Seismic Network (PRSN) from mid-2015 to mid-2016. This effort resulted in the detection of 6,335 earthquakes with 64,319 associated seismic phases. The machine learning (ML)-derived catalog for the OBS dataset includes 12,840 P-wave and 6,575 S-wave arrival times, a significant improvement over the USGS catalog (DOI: 10.5066/P13485PX), which documented 9,368 OBS phase arrivals. Our preliminary relocated catalog demonstrates notable enhancements in both the number of detected events and the accuracy of earthquake locations. Future work will focus on processing the PRSN catalog from mid-2016 onward, incorporating differential travel times using cross-correlation techniques, and relocating earthquakes. This approach aims to calibrate phase arrivals and improve relative location uncertainties offshore through spatial clustering of earthquakes during and beyond the OBS deployment period.

Session: ESC-SSA Joint Session:Seismology in the Global Oceans: Advances in Methods and Observations - I

Type: Oral

Date: 4/15/2025

Presentation Time: 05:15 PM (local time)

Presenting Author: Asiye

Student Presenter: No

Invited Presentation: 

Poster Number:

Authors