Mitigating Noise in Real-time GPS Positions to Improve Reliability of Geodetic Magnitude Estimates in the ShakeAlert Earthquake Early Warning System
Description:
The ShakeAlert® earthquake early warning system uses two algorithms based on seismic data to infer source parameters such as magnitude and location from which it calculates expected ground motion at user locations. A third algorithm, GFAST-PGD, is triggered when the seismic algorithms detect an earthquake and estimates magnitude (MPGD) using peak ground displacement (PGD) measured with real-time, high-rate (1 Hz) Global Positioning System (GPS) data. GFAST-PGD calculates MPGD as it receives new positions, but for any given estimate, ≥3 of the contributing GPS stations must have observed PGD meeting an empirical threshold that grows with time due to nonstationary noise (evident as drift or steps in PGD time series). Ideally, PGD thresholds are only met during displacement due to a large earthquake. However using real-time data recorded in the absence of earthquakes large enough to generate GPS-detectable motion, we found that PGD calculated relative to hypothetical earthquake origin times often still exceeds thresholds. When PGD threshold exceedances due to noise coincide with ShakeAlert’s detection of a small earthquake, GFAST-PGD could produce spuriously large MPGD. Processing raw GPS data to find an individual station’s position requires externally provided real-time satellite orbit and clock correction products. We assess the effect of new products that are lower latency and more frequently updated. By quantifying how often stations in the ShakeAlert network exceed PGD thresholds in the absence of earthquakes before and after switching products, we find fewer exceedances with the new products, reflecting a noise reduction. This improvement may allow lowering PGD thresholds so that GFAST-PGD can contribute to more events, like the December 2024 M7.0 Cape Mendocino earthquake. Although GFAST-PGD did not contribute to ShakeAlert’s magnitude estimate for this event because <3 stations met existing PGD thresholds, retrospective analysis shows GFAST-PGD would have yielded a MPGD estimate of ~M6.5 in real-time.
Session: Performance and Progress of Earthquake Early Warning Systems Around the World [Poster]
Type: Poster
Date: 4/16/2025
Presentation Time: 08:00 AM (local time)
Presenting Author: Adam
Student Presenter: No
Invited Presentation:
Poster Number: 120
Authors
Adam Manaster Presenting Author aemanaster@contractor.usgs.gov Universities Space Research Association |
Jessica Murray Corresponding Author jrmurray@usgs.gov U.S. Geological Survey |
Mark Murray mhmurray@usgs.gov U.S. Geological Survey |
Carl Ulberg ulbergc@uw.edu University of Washington |
Marcelo Santillan marcelo@geology.cwu.edu Central Washington University |
Craig Scrivner scrivner@geology.cwu.edu Central Washington University |
Timothy Melbourne tim@geology.cwu.edu Central Washington University |
Walter Szeliga walter@geology.cwu.edu Central Washington University |
Brendan Crowell crowell.97@osu.edu Ohio State University |
Mitigating Noise in Real-time GPS Positions to Improve Reliability of Geodetic Magnitude Estimates in the ShakeAlert Earthquake Early Warning System
Session
Performance and Progress of Earthquake Early Warning Systems Around the World