Abstract: In order to improve emergency management and control ability of huge crowds in scenic spots, a combination of scenario analysis and dynamic Bayesian network was used to study evolution process of stampede accidents. Situation state, disaster-causing body, disaster-bearing body and driving factors were selected as key elements to explore law and path of accident scenario evolution. Then, the dynamic Bayesian network was used to construct a scenario network of huge crowd stampede incidents, and development trend of the scene was deduced by using situation state probability. The results show that scenarios with high risk of accidents are population fall, crowd stalemate, population flow increase and crowd gathering in turn. And the major reasons for them are interventions of corresponding driving factors such as inadequate preparation for prevention, inadequate security management of scenic spots, imperfect human flow monitoring equipment and crowd diversion facilities.

Key words: scenic spots, intensive crowd, stampede accidents, scenario analysis, dynamic Bayesian network