The Extended Integrated Particle Filter Method (IPFx) as a High-Performance Earthquake Early Warning System

Abstract

ABSTRACT An earthquake early warning (EEW) system rapidly analyzes seismic data to report the occurrence of an earthquake before strong shaking is felt at a site. In Japan, the integrated particle filter (IPF) method, a new source-estimation algorithm, was recently incorporated into the EEW system to improve the source-estimation accuracy during active seismicity. The problem of the current IPF method is that it uses the trigger information computed at each station in a specific format as the input and is therefore applicable to only limited seismic networks. This study proposes the extended IPF (IPFx) method to deal with continuous waveforms and merge all Japanese real-time seismic networks into a single framework. The new source determination algorithm processes seismic waveforms in two stages. The first stage (single-station processing) extracts trigger and amplitude information from continuous waveforms. The second stage (network processing) accumulates information from multiple stations and estimates the location and magnitude of ongoing earthquakes based on Bayesian inference. In 10 months of continuous online experiments, the IPFx method showed good performance in detecting earthquakes with maximum seismic intensity ≥3 in the Japan Meteorological Agency (JMA) catalog. By merging multiple seismic networks into a single EEW system, the warning time of the current EEW system can be improved further. The IPFx method provides accurate shaking estimation even at the beginning of event detection and achieves seismic intensity error <0.25  s after detecting an event. This method correctly avoided two major false alarms on 5 January 2018 and 30 July 2020. The IPFx method offers the potential of expanding the JMA IPF method to global seismic networks.