Stochastic Modeling of Typhoon-induced Flood in Taiwan

Tropical cyclones always bring precipitation to Taiwan, and they often cause significant flood resulting in economic losses for the insurance industry. The 2001 typhoon Nari produced record-breaking rainfalls in various parts of Taiwan, and more than 90% of the insured losses were attributable to precipitation-induced flood. Given that reliable historical data on precipitation is limited, a stochastic model has been developed to simulate tropical-cyclones-related-precipitation along with the resulting insured losses. The purpose of the model is to evaluate typhoon-induced flood risk. The model consists of three main components: hazard, vulnerability, and insurance loss estimation. In hazard component, auto-regressive model was used to generate ten thousand years of synthetic tropical cyclones in the North West Pacific. Based on studies of Frank (1977) and Simpson and Riehl (1981), a simple parametric rainfall rate model was proposed to calculate rainfall rate associated with each synthetic event. Given available historical data, various statistical tests were then carried out to evaluate results from the hazard component. The tests performed indicate that the proposed methodology has the capability to simulate the frequency and severity for wind and rainfall. After completion of the hazards component, vulnerability functions were developed on the basis of building attributes such as occupancy, construction materials, and building height. Vulnerability functions for contents were also developed. Results from the loss estimation show that the model is capable of reproducing the flooding and insured losses from Nari and other historical typhoons. Future work will focus on physical enhancements of the simple parametric rainfall rate model as well as examination of the impact of global warming on tropical cyclone precipitation.