An Earthquake Early Warning Method Based on Bayesian Inference

Abstract

Earthquake early warning (EEW) systems can give people warnings before damaging ground motions reach their site and can reduce earthquake-related losses. It has been shown that Bayesian reference can be used to estimate earthquake magnitude, location, and the distribution of peak ground motion using observed ground-motion amplitudes, predefined prior information, and appropriate attenuation relationships. In this article, we describe a Bayesian approach to earthquake early warning (EEW) systems for the estimation of magnitude (M) and peak ground-motion velocity (PGV) using the Gutenberg–Richter relation as a priori probability derived from statistical historical earthquake information from the Japanese KiK-net. Moreover, using the magnitude Mj = 4.5 and PGV = 0.4 cm/s as the warning threshold, an earthquake hazard discrimination model was established. Following this, the proposed approach was compared with the traditional fitting method to analyze the earthquake hazard discrimination using an Mj = 4.3 earthquake in the Mt. Fuji area and an Mj = 5.5 earthquake in the Ibaraki area of Japan as examples. The results are as follows: (1) The probabilistic algorithm founded on the predictive model of the magnitude from the average period (τc) and PGV from the displacement amplitude (Pd) from the initial 3 s of the P wave is able to provide a fast and accurate estimation of the final magnitude and location. (2) The Bayesian inference approach for the earthquake hazard discrimination model has a 2.94% miss rate and a 5.88% false alarm rate, which is lower than that of the traditional fitting method, thus increasing the accuracy and reliability of earthquake hazard estimation.