An Earthquake Early Warning Method Based on Huygens Principle: Robust Ground Motion Prediction Using Various Localized Distance-Attenuation Models

摘要

For the stable operation of earthquake early warning systems under complex earthquake scenarios such as large earthquakes with complicated rupture processes and multiple simultaneous events during intense aftershock activity, wavefield-based ground motion prediction approaches that can predict shaking directly from the observed shaking, without source parameter characterization, have been proposed. These existing algorithms, however, have weak points, including the employment of inflexible wave propagation models and limited available lead times for sites that have a potential to obtain long lead times. To address these shortcomings, I developed the Approximation by Local Pseudo-Hypocenter Attenuation (ALPHA) method, a new wavefield-based algorithm that predicts ground motion by estimating the distance-attenuation relation from direct observations of the ongoing wavefield in real time. For this estimation, ALPHA uses multiple point-source models located below each observation station to simulate elementary wave propagation in accordance with Huygens principle. Application to the 2011 M_w9.0 Tohoku-Oki and 2016 M7.3 Kumamoto earthquakes and 21 additional earthquakes with M ≥ 7 showed that ALPHA stably calculated accurate ground motion predictions for those events and provided long lead times for sites with an intensity threshold on bedrock of three on the Japan Meteorological Agency scale (~IV-VI on the modified Mercalli intensity scale). These findings indicate that ALPHA would enable early warning systems to issue accurate warnings to wider areas at an early stage than existing algorithms.