Learning the Low Frequency Earthquake Activity on the Central San Andreas Fault
Session: New Insights Into the Preparatory Phase of Earthquakes From Tectonic, Field and Lab Experiments II
Type: Oral
Date: 4/19/2021
Presentation Time: 02:30 PM Pacific
Description:
Advances in machine learning based data processing have proven successful in predicting the timing of slip in controlled laboratory experiments and fault slow-slip in the Cascadia subduction zone. Low frequency earthquakes (LFEs) originating below the central San Andreas Fault are associated with slow-slip within the deeper, more ductile portion of the crust, with some evidence of activity changes before seismogenic events. Efforts to catalog LFEs using template matching techniques produced >1 million detections over 15 years and provide a data set to train models that predict LFE activity. We apply gradient boosted tree models using statistical features that describe seismic waveforms at different bandpass filters to estimate the LFE activity along the San Andreas Fault. Five years of data are used to train thousands of models and obtain the best constrained set of model hyperparameters. Applying the trained model to 7 years of seismic waveforms the model has never seen reproduces the burst-like LFE behavior observed when applying template matching, with the greatest misfit occurs during the largest burst sequences for the 1 hour time intervals. During periods of low LFE activity, the model indicates more activity than template matching. The most informative features to the model are in the higher frequency bands, suggesting the model is utilizing information outside the typical frequency band associated with LFEs. Similarities are found between detecting LFEs and previous efforts to detect tremors, which provides additional evidence tremors are composed of LFEs. The ability to continuously monitor LFE activity provides insight to the occurrence of slow-slip activity, without the need for developing a computational-intensive template-matching catalog, and promotes ongoing efforts to relate slow slip to larger seismogenic events.
Presenting Author: Christopher Johnson
Student Presenter: No
Authors
Christopher Johnson Presenting Author Corresponding Author cwj@lanl.gov Los Alamos National Laboratory |
Paul Johnson paj@lanl.gov Los Alamos National Laboratory |
|
|
|
|
|
|
|
Learning the Low Frequency Earthquake Activity on the Central San Andreas Fault
Category
New Insights Into the Preparatory Phase of Earthquakes From Tectonic, Field and Lab Experiments