The potential for improved ShakeAlert earthquake early warning using real-time distributed slip models

Jessica R. Murray

Submitted September 7, 2025, SCEC Contribution #14483, 2025 SCEC Annual Meeting Poster #TBD

The ShakeAlert® earthquake early warning (EEW) system uses multiple algorithms based on seismic or geodetic data to infer earthquake magnitude and location. This information is used to calculate anticipated shaking intensity at user locations, and locations whose predicted modified Mercalli intensity (MMI) exceeds a pre-defined threshold will receive alerts. The seismic-based EPIC algorithm uses the first few seconds of the P-wave to rapidly characterize the source, but its magnitude estimates tend to saturate around M7.0, and its location estimates can be poorly constrained for off-shore events. To address magnitude saturation, ShakeAlert includes a geodetic method, GFAST-PGD, based on real-time, 1 Hz global navigation satellite system (GNSS) data. GFAST-PGD estimates magnitude using a scaling relationship between hypocentral distance and peak ground displacement (PGD), adopting the source location from the seismic algorithm(s). Its point source assumption poses a challenge for large subduction interface earthquakes due not only to location error but also the effect of fault finiteness on ground motion for very large events.

A separate geodetic EEW algorithm, BEFORES, infers temporally-evolving, three-dimensional distributed slip models from real-time GNSS data. Here I investigate the potential impact of including the BEFORES algorithm in ShakeAlert for slab interface earthquakes like those expected in the Cascadia subduction zone. Using high-rate (1 Hz sampling) displacement time series for eight historic earthquakes, I show that BEFORES can increase the amount of warning time to user locations that experience strong (MMI 5.5+) shaking relative to GFAST-PGD as implemented for ShakeAlert, especially in applications that use an alert threshold of MMI 4.5. I also modify the original BEFORES method (Minson et al., JGR, 2014) to incorporate a priori information on slab geometry; this reduces the extent to which locations experiencing only mild shaking would receive alerts. While the Cascadia subduction zone is largely outside of California, better characterization of events anywhere along the subduction zone will contribute to more accurate and timely alerts, as well as minimize unnecessary alerting, in California.

Key Words
earthquake early warning, real-time GNSS

Citation
Murray, J. R. (2025, 09). The potential for improved ShakeAlert earthquake early warning using real-time distributed slip models. Poster Presentation at 2025 SCEC Annual Meeting.

Related Projects & Working Groups
Tectonic Geodesy