Future of Earthquake Early Warning: Quantifying Uncertainty and Making Fast Automated Decisions for Applications

摘要: Earthquake early warning (EEW) systems have been rapidly developing over the past decade. Japan Meteorological Agency (JMA) has an EEW system that was operating during 2011 M9 Tohoku earthquake in Japan, and this increased awareness of around world. While longer-time prediction still faces many challenges to be practical, availability shorter-time opens up a new door for loss mitigation. After fault begins rupturing, utilizes first few seconds recorded seismic waveform data quickly predict hypocenter location, magnitude, origin time expected shaking intensity level region. This information is broadcast different sites before strong arrives. The lead such short, typically minute or so, uncertain. These factors limit human intervention activate mitigation actions must addressed engineering applications EEW. study applies Bayesian probabilistic approach along with machine learning techniques decision theories from economics improve aspects operation, including extending it applications. Existing are often based on deterministic approach. Often, they assume only single event occurs within short period time, which led false alarms after Japan. develops probability-based algorithm existing model extend case concurrent events, observed aftershock sequence large earthquake. To overcome challenge uncertain EEW, also automated decision-making (ePAD) framework make robust A cost-benefit can capture uncertainties process used. called Performance-Based Early Warning, PEER Engineering method. Use surrogate models suggested computational efficiency. Also, proposed add influence into analysis. For example, value used quantify potential delaying activation action possible reduction uncertainty next update. Two practical examples, evacuation alert elevator control, studied illustrate ePAD framework. Potential advanced applications, as multiple-action decisions synergy structural health monitoring systems, discussed.