Retrospective and real-time event-based performance of the PLUM earthquake early warning algorithm for the West Coast, USA

Elizabeth S. Cochran, Debi Kilb, Julian Bunn, Sarah E. Minson, Annemarie S. Baltay, Mitsuyuki Hoshiba, & Yuki Kodera

Submitted August 12, 2019, SCEC Contribution #9432, 2019 SCEC Annual Meeting Poster #TBD

Using both retrospective and real-time event-based performance assessments, we explore the effectiveness of the Propagation of Local Undamped Motion (PLUM) earthquake early warning (EEW) algorithm, which we previously optimized for southern California data, notably by requiring two alerting stations (Cochran et al., 2019). The PLUM algorithm differs from typical EEW algorithms in that it is designed to detect and forecast strong ground motions, not earthquake source characteristics. We first investigate a retrospective 6-year (2012-2017) test suite of 558 M≥3.5 US earthquakes, within the California, Oregon, and Washington recorded by ~750 stations. PLUM alerts using Modified Mercalli Intensity (MMI) trigger thresholds of 4.0 and 2.5 for the first and second stations, respectively. The latencies in detecting M≥3.5 events range between 3-37 seconds (mean 8-seconds) when observed ground motions exceed MMI 4. Of the 20 M≥5 events, PLUM detects 10 events. Eight of the undetected events were located offshore, and the other two were in regions of sparse station coverage. PLUM successfully detected two offshore events, a 2014 M6.8 event ~100 km offshore and a 2015 M5.7 event within 30 km of the coastline. In January 2019, we began testing PLUM in real-time, using the same 4.0 and 2.5 MMI thresholds. PLUM issued alerts for the July 2019 Ridgecrest M6.4 and M7.1 earthquakes, with latencies (including telemetry) of 5 and 7 seconds. PLUM latencies are lower than those of the production ShakeAlert source-based algorithms (EPIC and FinDer), but ground motion-based PLUM warning times are currently limited to approximately 10 s as they are extrapolated from nearby observations. PLUM correctly forward-predicted that ground motions would exceed MMI 4.0 in Los Angeles, in contrast to the existing ShakeAlert system that underestimated both the mainshock magnitude, and the ground motions. In its current configuration, we conclude that PLUM could provide reliable strong motion alerts for EEW in regions where the seismic station network is dense and where the inter-station spacing is usually less than 30 km.

Key Words
Earthquake early warning, real-time seismology, ground motion

Citation
Cochran, E. S., Kilb, D., Bunn, J., Minson, S. E., Baltay, A. S., Hoshiba, M., & Kodera, Y. (2019, 08). Retrospective and real-time event-based performance of the PLUM earthquake early warning algorithm for the West Coast, USA. Poster Presentation at 2019 SCEC Annual Meeting.


Related Projects & Working Groups
Seismology