PhD Final Defense – Yi (Victor) Wang
Predicting Earthquake Casualty Rates Accounting for Community Vulnerability
Advisor: Professor Paolo Gardoni
Co-advisor: Professor Colleen Murphy
Date: Thursday May 24th, 1:30 pm
Location: Room 2311, Newmark Civil Engineering Building
Casualties are one of the most severe results of earthquakes. Prediction of casualty rates due to earthquakes plays an important role in seismic risk management. A predicted casualty rate reflects the vulnerability of a community of interest. An accurate prediction of earthquake casualty rates may facilitate the response immediately after a devastating earthquake. Predictions of earthquake casualty rates also offer insights for seismic risk analyses regarding future earthquakes.
Existing models for predicting earthquake casualties, however, had three general limitations. They usually required a detailed building inventory that might not be readily available. They tend to account insufficiently for socioeconomic factors that may affect earthquake casualties. Many overlooked data points with zero casualties, while omissions of zero observations may lead to selection bias.
Considering these three limitations, this dissertation presents a data-driven methodology of predicting earthquake casualty rates. The proposed methodology implements regression models that do not rely on detailed building inventories. The regressions are conducted with intensity measure and socioeconomic data that reflect the vulnerability of communities of interest. The dissertation uses regression approaches with zero-inflated techniques to take full advantage of the zero-casualty data points.
Through the development of a fragility formulation, the dissertation models the earthquake casualty rate of a community as the conditional probability that a standard person in the community is killed or injured for a given intensity measure of the earthquake at the site. It presents three case studies on the calibration and validation of the probabilistic fragility models. The three cases are based on the 2015 Gorkha earthquake in Nepal, 61 earthquakes affecting Taiwan from 1999 to 2016, and 902 earthquakes worldwide between 2013 and 2017, respectively. Using seismic hazard maps, the dissertation further applies the computed fragilities to conduct seismic risk analyses to predict the expected casualty rates and counts due to future earthquakes.
Although the proposed methodology is presented to predict casualty rates due to earthquakes, it is general and can be applied to other types of impacts as well as hazards. Future works may also extend this methodology to predict individual proneness to hazard impacts. The dissertation not only enriches the existing literature on rapid assessment of hazard losses for response, but also offers an evidence-based method to predict future hazard impacts that may be useful for the insurance and reinsurance industries.