Deep Learning based approach to Integrate MyShake’s Trigger Data with ShakeAlert for Faster and Robust Earthquake Early Warning Alerts

Abstract:

Earthquake early warning can potentially provide seconds to tens of seconds of warning to the target area before the strongest shaking hit the region. In the race with seismic waves, every second counts in terms of reducing the earthquake hazards. There have been many developments in infrastructures, algorithms and other aspects to make the system as fast as it can. Starting from 2019, ShakeAlert, the earthquake early warning system for the west coast of the United States, has been sending public warning in the state of California in 2019 via various ways. This presentation aims for analyzing and training a convolutional neural network model to integrate the dense but low-quality MyShake smartphone triggers into the current ShakeAlert system with the goal of enabling faster detection in some cases to increase the warning time and robust alerts. The initial results of this development builds on progress made during the deployment and operation of the MyShake smartphone seismic network. Data from MyShake will provide more triggers during the earthquake especially in densely populated areas. The goal is to reduce the earthquake detection time for the ShakeAlert system, therefore increasing the warning time.