← Back to Sessions

WITHDRAWN Performance of a Real-Time Machine Learning Classifier in the Epic Earthquake Early Warning Algorithm

Description: 

WITHDRAWN ShakeAlert is the US West Coast Earthquake Early Warning (EEW) algorithm, now delivering alerts of impending shaking from earthquakes to users throughout California, Oregon, and Washington. The Earthquake Point-source Integrated Code (EPIC) (formerly ElarmS) algorithm is one of two algorithms currently running in the production ShakeAlert EEW System. EPIC is fast, usually the first algorithm to alert, and can create accurate point-source estimates of earthquake magnitudes and locations using a minimum of 1 sec of data from one station, and 0.5 sec of data from three additional stations.

We have implemented the Meier et al., 2019 Machine Learning (ML) classifier into the EPIC algorithm with the goal of reducing the number of false triggers coming into the system. This reduction of spurious triggers could prevent erroneous alerts and also allows EPIC to create alerts using fewer than the currently required 4 stations due to increased confidence that the signals are from earthquakes. This modified version of EPIC allows for the creation of 2-station alerts using triggers that pass the ML classifier. For those 2-station alerts, we have also added a PGA/PGV amplitude threshold for those first two triggers following the methodology of the PLUM algorithm. Here we present the replay and real-time results of the latest version of EPIC with the ML classifier implementation

Session: Earthquake Early Warning Optimization and Efficacy [Poster]

Type: Poster

Date: 4/20/2023

Presentation Time: 08:00 AM (local time)

Presenting Author: Angela I. Lux

Student Presenter: No

Invited Presentation: 

Authors