A Bayesian Network-Based Risk Assessment Model for the Typhoon-Storm Surge-Flood-Dike Burst Disaster Chain

Abstract

The disaster chain caused by  typhoon has caused huge social economy and environment impact. In this study, a Bayesian network-based risk assessment model is proposed for the typical typhoon-storm surge-flood- dike burst (TSFD) disaster chain based on its uncertainty relationship of disaster elements in the disaster process. Firstly, thirteen main variables which affecting disaster events are mined and screened to construct the Bayesian network. Then, based on the historical TSFD disaster events, the chain transfer relationships between all of the thirteen variables are calibrated by calculating the conditional probability table. And three typical typhoon events are chosen to verify the Bayesian network using the Brier score. Finally, twelve disaster situations composed of three typhoon scenarios and four variables are applied to assess the TSFD disaster chain risk in the Greater Bay Area of China. The results reveal that typhoon intensity, typhoon path and astronomical tidal cycle can have significant impacts on the risk of the TSFD disaster chain. Astronomical tidal cycle and typhoon path have a more significant impact on storm surge disasters than on flood disasters, while the opposite effect is found for typhoon intensity. It is also noted the timely supply of flood-control materials can reduce the degree and risk of dike burst, but the probability of dike burst is reduced by approximately 2% in a high-intensity typhoon event or during spring tide.

Keywords: disaster chain, Risk assessment, multi-situation analysis, Greater Bay Area, Bayesian network