Successful ShakeAlert Performance for the Napa Quake

Richard M Allen, University of California Berkeley, Department of Earth and Planetary Science, Berkeley, United States, Douglas D Given, USGS, Pasadena, CA, United States, Thomas H Heaton, California Institute of Technology, Division of Engineering and Applied Science, Pasadena, CA, United States and Prof. John Emilio Vidale, University of Washington, Seattle, United States
Abstract:
ShakeAlert, the demonstration earthquake early warning system, developed by the USGS, UC Berkeley, Caltech, ETH, and the University of Washington, functioned as expected for the August 24, 2014, M6.0 Napa earthquake. The first ShakeAlert was generated by the ElarmS algorithm 5.1 sec after the origin time of the earthquake, and 3.3 sec after the P-wave arrived at the closest station 6.5 km from the epicenter. This initial alert, based on P-wave triggers from four stations, estimated the magnitude to be 5.7. The warning was received at the UC Berkeley Seismological Laboratory 5 seconds before the S-wave and about 10 sec prior to the onset of the strongest shaking. ShakeAlert beta-testers across the San Francisco Bay Area simultaneously received the alert, including the San Francisco 911 center with 8 sec warning, and the BART train system. BART has implemented an automated train-stopping system that was activated (although no trains were running at 3:20 am). With the available network geometry and communications, the blind zone of the first alert had a radius of 16 km. The four stations that contributed to the first alert all encapsulate data into 1-second packets, but the latency in transmitting data to the processing center ranged from 0.27 to 2.62 seconds. If all the stations were to deliver data in 0.27 seconds, then the alert would have been available 2.3 sec sooner and the blind zone would be reduced to about 8 km. This would also mean that the city of Napa would have received about 1 second of warning. The magnitude estimate and event location were accurate from the initial alert onwards. The magnitude estimate did first increase to 5.8 and then dip to 5.4 2.6 sec after the initial alert, stayed at that level for 2 sec, and then returned to 5.7. The final magnitude estimate was 6.0, consistent with the ANSS catalog.