An end-to-end earthquake monitoring method for joint earthquake detection and association using deep learning
Abstract
High-accuracy picking and association of seismic phases across a network are central problems for earthquake monitoring systems. Current event detection methods typically treat these two tasks as distinct sub-problems: first, a phase picking algorithm identifies candidate phase arrivals at each station; then, an association algorithm combines these candidate arrival times with a velocity model to determine possible event hypocenters. A disadvantage of this two-stage approach is its reliance on accurate phase picking at each station -- a difficult task for low signal-to-noise ratio arrivals from low-magnitude events. Additionally, the association stage typically ignores potentially informative waveform similarity features across stations. In this work, we propose an end-to-end event detection approach that considers the picking and association stages jointly. Our method is based on a deep neural network architecture that incorporates known physical constraints via a pre-defined velocity model for the geographical region of interest. Given an input seismogram from each station, the neural network extracts sequences of features from the raw time series. At any candidate hypocenter, the feature sequences are then shifted using the travel times given by the velocity model. Finally, the shifted features are aggregated to predict whether an event occurred at a given space-time coordinate. The parameters of the neural network can be jointly optimized using a dataset of seismogram records and ground-truth events. By combining phase picking and association, we can both reduce false positive picks and also improve the sensitivity of earthquake monitoring systems to events that are too weak to be detected by any single station in the network.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2019
- Bibcode:
- 2019AGUFM.S43D0681Z
- Keywords:
-
- 0555 Neural networks;
- fuzzy logic;
- machine learning;
- COMPUTATIONAL GEOPHYSICS;
- 1910 Data assimilation;
- integration and fusion;
- INFORMATICS;
- 1914 Data mining;
- INFORMATICS;
- 1942 Machine learning;
- INFORMATICS