Early Warning Performance of FinDer, Virtual Seismologist, and eBEAR Systems in the 2024 ML 7.2 Hualien, Taiwan, Earthquake

Résumé fourni par la source

Abstract Public earthquake early warning (EEW) in Taiwan has been operating for more than a decade with a single point-source algorithm, existing Earthworm-Based Earthquake Alarm Reporting (eBEAR). For large earthquakes with finite rupture, point-source EEW algorithms will fail to accurately predict ground motion for all affected regions, even when the magnitude is correctly estimated. Further, single-algorithm EEW systems are more likely to fail than those that operate multiple independent algorithms. Since mid-2023, the Central Weather Administration (CWA) has been testing the Swiss Seismological Service (SED) at ETH Zurich SeisComP EEW system, developed at the Swiss Seismological Service at ETH Zürich, to explore the benefits of integrating Finite-fault rupture Detector (FinDer) and virtual seismologist (VS) using the SeisComP platform. This study aims to evaluate whether combining line-source and point-source models can overcome limitations observed in the existing EEW strategy, particularly during large earthquakes. During the recent ML 7.2 earthquake on 3 April 2024, in Hualien County, the original eBEAR faced challenges with its alert updating mechanism, issuing high-level Public Warning System alerts to only 12 counties and cities, notably missing five highly populated northern areas, including the capital, Taipei. To address this issue, CWA lowered the magnitude change threshold for updating alerts from 0.5 to 0.1 magnitude units. This adjustment improved the responsiveness of initial alerts; however, its point-source assumption still resulted in limited spatial coverage. In contrast, FinDer accurately captured rupture geometry and achieved the highest accuracy in ground-motion prediction as well as broader spatial alert coverage. Although the VS algorithm initially suffered from a configuration issue, the offline-corrected version yielded results similar to eBEAR, reflecting shared limitations of point-source methods. Our results demonstrate the complementary characteristics of point-source and finite-fault models in EEW applications. We suggest that integrating rapid point-source methods such as eBEAR or VS with finite-fault approaches like FinDer may significantly enhance both ground-motion prediction and reliability of EEW systems in future large earthquakes.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.