Investigations on Real-time GPS for Earthquake Early Warning

Abstract:

The Geodetic Alarm System (G-larmS) is a software system developed in a collaboration between the Berkeley Seismological Laboratory (BSL) and New Mexico Tech (NMT) primarily for real-time Earthquake Early Warning (EEW). It currently uses high rate (1Hz), low latency (< ~5 seconds), accurate positioning (cm level) time series data from a regional GPS network and P-wave event triggers from existing EEW algorithms, e.g. ElarmS, to compute static offsets upon S-wave arrival. G-larmS performs a least squares inversion on these offsets to determine slip on a finite fault, which we use to estimate moment magnitude. These computations are repeated every second for the duration of the event. G-larmS has been in continuous operation at the BSL for over a year using event triggers from the California Integrated Seismic Network (CISN) ShakeAlert system and real-time position time series from a fully triangulated network consisting of BARD, PBO and USGS stations across northern California. Pairs of stations are processed as baselines using trackRT (MIT software package). G-larmS produced good results in real-time during the South Napa (M 6.0, August 2014) earthquake as well as on several replayed and simulated test cases. We evaluate the performance of G-larmS for EEW by analysing the results using a set of well defined test cases to investigate the following: (1) using multiple fault regimes and concurrent processing with the ultimate goal of achieving model generation (slip and magnitude computations) within each 1 second GPS epoch on very large magnitude earthquakes (up to M 9.0), (2) the use of Precise Point Positioning (PPP) real-time data streams of various operators, accuracies, latencies and formats along with baseline data streams, (3) collaboratively expanding EEW coverage along the U.S. West Coast on a regional network basis for Northern California, Southern California and Cascadia.