Poster #227, Seismology

Earthquake Phase Association with Graph Neural Networks

Ian McBrearty, & Gregory C. Beroza
Poster Image: 

Poster Presentation

2021 SCEC Annual Meeting, Poster #227, SCEC Contribution #11525 VIEW PDF
In this work we present a new Graph Neural Network (GNN) architecture for earthquake phase association, in which we process streaming pick datasets, determine the number and location of earthquakes in a time window, and associate picks to each active source. The GNN is trained through supervised learning with synthetic pick datasets, for which the ground truth is known, and for which there is high variability and noise in the pick datasets. The network is not trained for a particular configuration of stations, rather it is trained to allow variable: network geometry, numbers of stations, station qualities, and pick rates. By frequently including closely overlapping events in space and time i...n the training data, the GNN learns to untangle overlapping events. As a mathematical function, the GNN maps sets of sets (sets of discrete picks, on each station) to a continuous, smooth, bounded-prediction of source-likelihoods in space-time, similar to the traditional back-projection (BP) mapping; however this method greatly suppresses side-lobes, and large and small earthquakes are mapped to a similar output value (in contrast to BP, where outputs scale with the number of observed picks). The technique has been tested on real data from the NC network of northern California, using PhaseNet-produced picks as input, where we recover at least 95% of previously reported USGS earthquakes > M1 throughout the interval 2000 – 2020. Initial applications suggest a substantial number of new events < M1 are also detectable over this time window, and work is ongoing to build a new catalog for northern CA and quality control the new detections.