The Plum Earthquake Early Warning Algorithm: A Case Study of Two West Coast, USA, Datasets

Abstract:

The PLUM (Propagation of Local Undamped Motion) earthquake early warning (EEW) algorithm differs from typical source-based EEW algorithms in that it forward-predicts shaking directly from observations. We apply PLUM to two retrospective datasets: 558 M3.5+ earthquakes within California, Oregon, and Washington (2012-2017), and a test suite of historical and other signals (1999-2014). We compute Modified Mercalli Intensity (MMI) using velocity and acceleration data, and leverage co-located sensor data to avoid problematic signals. We favor a two-station alerting method, using MMI trigger thresholds of 4.0 and 3.0 for the first and second stations, respectively. This configuration reduces the detection of M<4 events (where alerts for MMI4+ shaking are usually too late) without significantly increasing detection latencies of M5+ events. Based on existing station density, of the 558 M3.5+ earthquakes only 85 are theoretically detectable. PLUM correctly detected 79 events, with detection times of 1.5-37s (mean 8.0s; median 5.8s). Detection times do not include telemetry latencies or the time required to communicate alerts to people. No false alerts were issued at the MMI4+ level. PLUM detected all 10 of the theoretically detectable M5+ events. These included the 2014 M6.8 event ~100 km offshore that was problematic for other EEW algorithms. Of the 79 detected events, however, only two have the potential to provide timely warnings to regions of MMI4+ shaking: the 2014 M6.8 offshore earthquake and the M6.02 2014 Napa earthquake. For the test suite data, PLUM correctly identified 14/15 theoretically detectable M5+ local events and had a 100% success rate avoiding calibration pulses, anomalous signals, and regional and teleseismic events. We conclude PLUM can provide reliable EEW strong-motion alerts in regions where the seismic station network is dense (inter-station spacing ≤30 km), and is a promising candidate for integration into the U.S. ShakeAlert EEW system.