GIS-integrated flood risk assessment for metro systems based on bayesian cosine maximization method: a case study in Beijing

Against the backdrop of global climate change, extreme rainfall events are becoming increasingly frequent, leading to a continuous worsening of urban flooding and posing a serious threat to the operational safety of underground rail transit systems. To increase the objectivity and spatial accuracy of flood risk assessment for metro systems, this study proposes a comprehensive assessment model that integrates the Bayesian cosine maximization method (BCMM) with geographic information system (GIS) technology and applies it to the Beijing metro system as a case study. A 13-indicator evaluation system was established, covering three dimensions: hazard, exposure, and vulnerability. On the basis of the GIS platform, spatial quantification of multisource data such as topography, rainfall, and socioeconomic factors was carried out. Finally, the pairwise comparison matrix of expert scores was corrected via Bayesian theory, and the indicator weights were optimized via the cosine maximization method. A systematic comparison was then conducted with the traditional fuzzy analytic hierarchy process (FAHP). The results revealed that the central urban areas of Beijing and Tongzhou Districts are extremely high-risk flood areas, accounting for 4.89% of the total area. Within the metro system, 12.45% of the area falls into the extremely high-risk category, whereas 21.97% is classified as high risk. Notably, more than 50% of the stations on five lines, including Line 1 and Line 2, are exposed to high or extremely high flood risks. Compared with the FAHP method, the BCMM reduces the Euclidean distance in the weight calculation by 37.5%, demonstrates improved consistency and aligns better with historical flood-prone locations such as Jinanqiao. Under a simulated once-in-200-year extreme rainfall scenario, the overall flood risk level of the metro system increases significantly, with the proportion of high-risk areas and with the cumulative proportion of high- and extremely high-risk areas reaching 92.3%.