ShakeAlert Earthquake Early Warning Performance for the 2019 Ridgecrest Sequence

Brad T. Aagaard, Sarah E. Minson, Annemarie S. Baltay, Elizabeth S. Cochran, & Tom C. Hanks

Published August 14, 2019, SCEC Contribution #9605, 2019 SCEC Annual Meeting Poster #272

We evaluate the performance of the ShakeAlert earthquake early warning (EEW) system for M5+ earthquakes in the 2019 Ridgecrest earthquake sequence using a cost-savings, ground-motion-based performance metric for public alerting strategies. We find the system could have provided about 40-60 percent of the cost savings of a perfect EEW system. Improving magnitude estimates while accounting for event bias in ground motions relative to median values would significantly increase the performance. The latency involved in transmitting alerts also significantly affects performance.

Our performance metric is the ratio of the cost savings of the ShakeAlert system relative to the cost savings for a perfect EEW system. We consider the perfect EEW system to be one that forecasts the final Southern California Seismic Network ShakeMap ground motions instantaneously at the origin time of the earthquakes. For each alert generated by the ShakeAlert EEW system, we predict the ground motions using the Abrahamson, Silva, and Kamai (2014) ground-motion prediction equation and the lead time before strong shaking arrives with the shear wave. We compute the cost of damage and cost of action using simple cost functions that depend on MMI for both fear avoidance and injury mitigation. We attribute fear avoidance to experiencing less fear when alerted to shaking beforehand, with a reduction starting at MMI III and increasing to MMI V. We attribute injury mitigation to reducing injuries through people doing Drop, Cover, and Hold on when receiving an alert, with mitigation starting at MMI IV and increasing to MMI VIII. The ground-motion-based performance metric accounts for the cost of action and potential savings for timely alerts, late alerts, missed alerts, and false alerts. We incorporate latencies of 0s, 2s, 5s, and 10s to account for the time required to transmit and react to the alerts. To quantify how improvements in the system affect the performance, we examine theoretical cases for issuing alerts 0s, 5s, and 10s after origin time using the catalog earthquake magnitude.

Key Words
earthquake early warning, ground motion, performance

Citation
Aagaard, B. T., Minson, S. E., Baltay, A. S., Cochran, E. S., & Hanks, T. C. (2019, 08). ShakeAlert Earthquake Early Warning Performance for the 2019 Ridgecrest Sequence. Poster Presentation at 2019 SCEC Annual Meeting.

Related Projects & Working Groups
Ridgecrest Earthquakes