Emergency response process for sudden rainstorm and flooding: scenario deduction and Bayesian network analysis using evidence theory and knowledge meta-theory

The traditional "forecast-response" paradigm is facing significant practical challenges. For example, when different scenarios require different reaction mechanisms, the applicability of this method is weak since this paradigm describes the crisis itself from a macroperspective, neglecting analyzing emergency response measures from a microperspective. Hence, in this paper, the research paradigm will be shifting to "scenario-response" paradigm, which analyzes the impact of corresponding measures on events from a microscopic perspective. In 2021, an expected and torrential rainfall in Zhengzhou, Henan, China, causing caused 398 deaths and an estimated direct economic loss of 120.6 billion RMB. Accordingly, an empirical analysis was conducted for this heavy rain event to examine the intricate evolution of emergency response utilizing a constructed scenario Bayesian network. This network was constructed by fusing the knowledge meta-theory, scenario evolution and Dempster's rule of combination along with 362 relevant historical representative events, and it has the capability to identify the development of the various emergency events and fuse the assessments of different experts. The effects of each measure on the probability of response outcomes were analyzed in an event-driven Bayesian network. The counterfactual outcomes of the interventions were also explored through causal inference to determine the priority measures in the emergency response process. The similarity between each target scenario and each source scenario reached more than 0.7, among which the highest similarity reached 0.78. The evolutionary accuracy of the incident response process was examined by comparing the scenario similarity. The method proposed in this study can help as a theoretical basis for implementing the "scenario response" paradigm.