A Deep Learning Pipeline for Earthquake Detection

Date: 4/26/2019

Time: 08:30 AM

Room: Elliott Bay

Deep learning has led to significant recent advances in earthquake detection, phase picking, and association. Here, we discuss an end-to-end detection pipeline suitable for producing seismicity catalogs from raw continuous data. First, continuous data is processed with a deep convolutional network designed for generalized detection of body waves. Next, the resulting phase detections are examined by a recurrent neural network that is trained to link together phases that originate from the same earthquake. The final clusters of phase picks can then be used with a formal hypocenter inversion. We demonstrate the performance of the complete pipeline on datasets from southern California and Japan, and compare the results with that of template matching.

Presenting Author: Zachary E. Ross

Authors