ElarmS Earthquake Early Warning System Enhancements and Performance Evaluation

ElarmS is an earthquake early warning system that contributes alerts to CISN ShakeAlert, a prototype end-to-end earthquake early warning system being developed and tested by the California Integrated Seismic Network (CISN). ElarmS is one of several systems that contribute to CISN ShakeAlert. The UC Berkeley ElarmS system consists of multiple continuous-waveform processors and trigger-association processors running at three geographical locations and communicating via the Apache ActiveMQ Messaging system. With performance monitoring tools, we review the system component latencies and earthquake hypocenter estimations. Statistics are presented for hypocenter and origin time accuracy and alert times latencies for the previous year by geographic regions. We are also testing the ElarmS system by replaying the raw waveform data for historical earthquakes of interest. This allows us to check the signal detection algorithm, the trigger associator and the magnitude estimator. Over the last year, ElarmS successfully alerted on all significant earthquakes (M>=4.5, 10 earthquakes) and generated no false alarms. This includes events in Los Angeles, Santa Rosa, offshore Eureka, and the southern Sierra. The magnitude estimates were within 0.5 magnitude units in all cases. The largest event was the M6.8 event offshore of Eureka. The initial ElarmS magnitude estimate was M6.3 with a location 7 km from the true (offshore) location. ElarmS also alerted on all of the larger (M>=4.5) aftershocks. The speed of the alert depends on the quality of the seismic network around the epicenter. For the M5.1 Los Angeles Le Habra event the alert was 4.2 sec after the origin time, this is typical for events in the LA region. For the Santa Rosa event the alert was 6.8 sec after the origin time, 10.8 for the southern Sierra event, and 25.7 for the (offshore) Eureka earthquake. Comparing the performance of ElarmS across the state for a wider range of magnitudes (M>=3) the effect of variable seismic network quality is very apparent. The performance is good to excellent in the SF Bay and LA areas where station coverage is densest. In the greater SF Bay Area of the 38 M>=3.0 events reported by the ANSS, 5 were missed (all in the Geysers Geothermal Area) and there were no false events. In the LA region, of the 35 M>=3.0 events, 4 were missed and there were 2 false events.