Fast disaster assessment in unconventional emergences (UEs) is difficult due to the uncertainties and incomplete information. This article explores the use of Bayesian network (BN), which facilitates the quantification of uncertainties, for fast disaster assessment. To guide and discipline the development of BN model, we proposed a procedure that consisted of two modules divided by the occurrence of emergency. The procedure was illustrated by a High-casualty fire (HCF) assessment model that contained three subnetworks: the direct result of a HCF, an internet public opinion (IPO) and a decision support subnetwork. This model not only assesses the effects of HCF and corresponding IPO but also enables users to make decisions based on assessment results. The results suggest that the model can perform fast disaster assessment with incomplete information. Conditional probabilities and prior probabilities learned from historical data and expert elicitation can compensate for the shortage of information. Furthermore, the model provides useful information for choosing effective risk control options using sensitivity analysis.
展开▼
