Performance of On-Site Earthquake Early Warning System Using Strong-Motion Records from Recent Earthquakes

Abstract

The method of on-site earthquake early warning was proposed in the past decade to enhance seismic preparedness and safety measures for areas close to the earthquake epicenter. The method leverages critical information contained in the initial 3 s of a P wave that arrives early to predict possibly intense ground shaking at a site. In this study, two types of empirical relationships are developed: (1) Pd3-PGV relationship for predicting the horizontal peak ground velocity (PGV) based on the peak displacement in the first 3 s of the vertical motion (Pd3); and (2) τc−Mw relationship for estimating the moment magnitude (Mw) of an earthquake based on a ground-motion period parameter (τc). Performance of these empirical relationships in on-site earthquake early warning framework is examined using global earthquake records from the most updated Next Generation Attenuation (NGA-West2) strong-motion database. Region-specific empirical predictive equations for California and Japan are proposed based on the state of the art mixed-effect regression to separate ground-motion variability between different earthquakes and that within an earthquake. Applicability and inapplicability of the empirical models in the near-fault condition are investigated using ground motions with strong velocity pulses. The overarching goal of study is to leverage the comprehensive database and state of the art regression techniques to facilitate understanding and engineering practice of on-site earthquake early warning in a variety of regions and in the near-fault condition.