Anatomy of an Earthquake Early Warning (EEW) Alert: Predicting Time Delays for an End-to-End EEW System

Abstract

Earthquake early warning (EEW) systems can play an important role in seismic‐risk mitigation, filling the gap between long‐term measures such as earthquake safe building design (e.g., Housner et al. , 1997) and medium‐to‐short term measures like operational earthquake forecasting (e.g., Gerstenberger et al. , 2005) and rapid notifications following a strong earthquake (e.g., Wald et al. , 1999). Throughout the last two decades algorithmic improvements, denser seismic networks, and new communication technologies have made EEW feasible. With the achievable warning times increasing continuously, EEW is becoming useful to a wider range of potential end users (Bose et al. , 2013). EEW algorithms can roughly be grouped into three categories: on‐site, regional, and front detection approaches. The onsite approach sends an alert when one or more thresholds of waveform parameters measured in real time are exceeded at one or two stations (Nakamura, 1988; Kanamori, 2005; Bose, Hauksson, Solanki, Kanamori, and Heaton, 2009). The regional approach waits for the earthquake to be detected at a number of sites to estimate its location and then uses an empirical relationship to infer the magnitude and expected ground motion at a target site from the first few seconds of waveforms (Allen and Kanamori, 2003; Cua and Heaton, 2007; Cua et al. , 2009; Satriano et al. , 2011). The front detection approach is a special case of a regional approach that targets a specific area known to generate strong seismicity, hence the origin is assumed and only the magnitude is inferred (e.g., Espinosa‐Aranda et al. , 1995; Marmureanu et al. , 2011). Usually seismicity cannot be assumed to only occur in a particular area, and therefore the majority of operational and in‐development EEW algorithms today run an implementation of the onsite approach (Nakamura et al. , 2011) or, more commonly, a combination of the on‐site and network‐based …