Inclusion of real-time GNSS-based magnitude estimation in ShakeAlert

Jessica R. Murray, Carl W. Ulberg, Marcelo Santillon, Craig Scrivner, Brendan W. Crowell, & Timothy Melbourne

Published September 10, 2023, SCEC Contribution #12983, 2023 SCEC Annual Meeting Poster #059

To improve the ability of the ShakeAlert earthquake early warning system to alert for M7+ earthquakes, we adapt a magnitude estimation algorithm (“GFAST-PGD”, Crowell et al., SRL, 2016) based on peak ground displacement (PGD) measured in real-time using Global Navigation Satellite System (GNSS) data. As implemented for ShakeAlert, GFAST-PGD runs in parallel with two seismic algorithms, and the magnitude estimates from all three are combined in a weighted average (MAV). This implementation includes criteria to suppress inclusion of GFAST magnitude estimates (MPGD) that may be unreliable due to noise in the GNSS position estimates (see Murray et al., BSSA, 2023). Furthermore, new logic was introduced to determine which MPGD estimates are included in MAV each time that source parameters are updated during an earthquake. The criteria require that 1) ≥3 GNSS stations report PGD exceeding time-dependent threshold values determined empirically from analysis of GNSS position noise, 2) MPGD uncertainty ≤0.5 unit, 3) MAV ≥ 6.0 based on seismic data alone, and 4) MPGD ≥ 7.0.

The GFAST-PGD algorithm was submitted to ShakeAlert’s testing and certification system in February 2023. This formal evaluation involves both monitoring the real-time behavior of the algorithm and replaying data from historic earthquakes in simulated real-time mode. Methods that measurably improve alerting without introducing undue complication to the overall system are put forward for inclusion in production. Initial results indicated that inclusion of GFAST can improve warning times for large megathrust earthquakes but also revealed that the logic for combining magnitude estimates required further modification. Testing of an updated version is set to begin soon. Our ongoing monitoring of GFAST-PGD performance in real-time mode, as implemented on the non-production ShakeAlert servers, has highlighted the impact of nonearthquake-related steps and excursions in the GNSS time series that exceed the typical drift caused by colored noise. In parallel with ongoing testing, we are exploring ways to reliably flag problematic data and prevent its inclusion in MPGD estimation.

Key Words
Earthquake early warning, ShakeAlert, real-time GNSS

Citation
Murray, J. R., Ulberg, C. W., Santillon, M., Scrivner, C., Crowell, B. W., & Melbourne, T. (2023, 09). Inclusion of real-time GNSS-based magnitude estimation in ShakeAlert. Poster Presentation at 2023 SCEC Annual Meeting.

Related Projects & Working Groups
Tectonic Geodesy