Earthquake Signal Characterization Using Deep-Residual Convolutional-Recurrent Networks
Abstract
Advances in deep learning have led to a surge of interest in machine learning for the constituent tasks of seismic monitoring. Seismic waveforms are strongly non-stationary, and the waveforms that seismic sources produce are complex, with spatial and temporal behaviors that reflect complexity in the sources that generate seismic waves and in the medium that those waves propagate through. New approaches can exploit these complex data more fully than conventional approaches. Under this contract we developed and demonstrated multiple new and innovative signal characterization methods based on deep learning of local and regional seismic data to improve seismic event detection.
Document Details
- Document Type
- Technical Report
- Publication Date
- Nov 26, 2023
- Accession Number
- AD1223262
Entities
People
- Gregory C. Beroza
Organizations
- Stanford University