← Back to Sessions

Shakealert Earthquake Early Warning: Testing New Vs30 With Population Weighted Values, New Epic With Bayesian Priors, and Finder Triggering With CE Stations

Description: 

The ShakeAlert West Coast Earthquake Early warning system has been sending public alerts since October 17, 2019. To assess continuing algorithm development and proposed improvements, ShakeAlert has a System Testing and Performance (STP) platform. The overarching principle is that proposed changes should improve either the warning time and/or accuracy location and magnitude estimates while minimizing additional complexity to the system and alerting. Performance analysis must balance these competing goals. To properly assess the value of proposed software changes, STP replays a suite of waveforms (recordings of moderate to large earthquakes plus a few synthetic earthquakes) through 4-8 instances of the algorithm stack and assesses with metrics from Cochran et al. (2017). In addition, proposed software is run on a set of test systems which are configured like the production system to process real-time data for a minimum period of 2 weeks.

In two major ShakeAlert algorithm stack submissions a number of new features are tested. In the v.3.0.2 submission, the EqInfo2GM algorithm uses Vs30 values with population-weighting within 1km of a ShakeAlert grid cell. This module generates ground motion estimates from source parameters. The change is being tested vs baseline with an MMI tolerance criteria that displays the degree of over- and under-alerting in ground motion space. In the v.3.1.0 submission, the major changes are to EPIC and FinDer, the core algorithms for determining source parameters from seismic data. This version of EPIC uses a Bayesian prior to determine source locations of events, which should significantly improve out-of-network earthquake detections where the data could be limited. The new version of FinDer now incorporates CE network stations for creating a new trigger. (CE stations are strong ground motion stations operated by the California Geological Survey.) This should speed up FinDer’s first alert in some cases and potentially allow FinDer to trigger for more earthquakes in sparse networks.

Session: Performance and Progress of Earthquake Early Warning Systems Around the World - I

Type: Oral

Date: 4/16/2025

Presentation Time: 08:15 AM (local time)

Presenting Author: Angie

Student Presenter: No

Invited Presentation: 

Poster Number:

Authors