Reassessing the North Texas Earthquake Catalog Using Machine Learning
Description:
Machine learning (ML) phase detection and association has altered the landscape for the rapid creation of earthquake catalogs. Many rapid deployments in urban areas use a mix of accelerometers and velocity sensors and have a complicated metadata history, creating scenarios the push the limits of available training datasets. The North Texas Earth Study (NTXES) in the Fort Worth Basin consists of rapidly deployed 1- and 3-component short-period, strong-motion, broadband, and nodal sensors in a large metropolitan area with changing configuration as seismicity evolved from 2008 to present. The NTXES catalog, with over 3000 earthquakes, was originally built using automated detection with manual review (Quinones et al., 2019). While the catalog provided fundamentally important data to understand physical mechanisms of induced seismicity in the basin, original efforts to constrain time variation in b-values, conduct nearest neighbor determinations, etc. were limited by incompleteness at both the high and low magnitude end of the catalog. Preliminary work using EQTransformer (Mousavi et al., 2020) to grow the catalog indicated mixed results for picking accuracy and lowering detection magnitudes below the manual review and poor detection results for accelerometers. Here, I compare PhaseNet (Zhu and Beroza, 2018) arrival time detections with the manual catalog, seeing improved results on velocity sensors. The resulting catalog will be used to calculated time-variation in b-value using the b-positive method (van der Elst, 2021) and assess foreshock-mainshock-aftershock and earthquake swarm behavior using nearest neighbor determination (Zaliapin and Ben Zion, 2020). Preliminary work using Texas earthquake catalogs indicate that statistical approaches, and resulting interpretations, can be sensitive to biases in time and space, especially depth, inherent in both automated and relative relocation catalogs, as will be discussed.
Session: Mechanistic Insights into Fluid-induced Earthquakes from the Laboratory to the Field [Poster]
Type: Poster
Date: 4/15/2025
Presentation Time: 08:00 AM (local time)
Presenting Author: Heather
Student Presenter: No
Invited Presentation:
Poster Number: 93
Authors
Heather DeShon Presenting Author Corresponding Author hdeshon@smu.edu Southern Methodist University |
Yeshey Seldon yseldon@smu.edu Southern Methodist University |
|
|
|
|
|
|
|
Reassessing the North Texas Earthquake Catalog Using Machine Learning
Session
Mechanistic Insights into Fluid-induced Earthquakes from the Laboratory to the Field