基于随机森林与可变模糊集的城市洪涝脆弱性评估

洪涝灾害作为城市面临的重要自然灾害之一，对社会经济发展造成了重要的影响。首先从自然、经济、基础设施3个方面利用随机森林模型以数据驱动的方式筛选出了影响城市洪涝脆弱性的12项关键指标并据此构建了城市洪涝脆弱性评估指标体系。在此基础上，基于随机森林和熵权法计算指标体系权重，构建基于随机森林（RF）和可变模糊集（VFS）耦合的城市洪涝脆弱性评估模型（RF-VFS）。最后，以长江流域特大城市南京市为例，进行洪涝脆弱性评估，得到了影响南京市洪涝脆弱性的关键影响因素，并从完善城市洪涝管理政策、加强城市基础设施建设和防灾减灾应急措施3个方面提出了降低城市洪涝脆弱性的对策建议。结果表明：南京市洪涝脆弱性与汛期降雨量、硬地面积、地均GDP、城市道路网密度等指标呈较强关联；近20年南京市洪涝脆弱性等级呈上升趋势且高等级的年份出现频率愈加频繁，尤其是近5年洪涝脆弱性等级较高，研究结果与实际调研情况相符。该研究能为南京市相关管理部门加强洪涝管理和降低洪涝脆弱性提供决策参考，同时对其他城市的洪涝灾害管理也具有借鉴意义。

Urban Flood Vulnerability Assessment Based on Random Forests and Variable Fuzzy Sets

As one of the important natural disasters faced by cities, urban floods have important impacts on urban social and economic development. For example, Beijing in 2012, Wuhan in 2016, and northern Anhui in 2018 suffered severe economic losses due to floods. This paper established the assessment index system of urban flood vulnerability based on the random forest model from nature, economy, and infrastructure aspects; then combined with the random forest and entropy weight method, the weights of the index system were calculated. Finally, the urban flood vulnerability assessment model (RF-VFS) based on the coupling of random forest (RF) and variable fuzzy set (VFS) was constructed. As a typical flood area in the middle and lower reaches of the Yangtze river basin, Nanjing city is prone to severe flood disasters in summer due to the influence of topography and seasonal rainfall. Therefore, Nanjing was taken as a case study area to explore the dynamic characteristics and driving factors of flood vulnerability, and the corresponding vulnerability grade curve of Nanjing from 1996 to 2018 is analyzed. The study handled data and its digital mapping based on GIS. The countermeasures were proposed from three aspects: urban flood management policies, urban infrastructure construction, and emergency measures for disaster prevention and reduction. The assessment results showed that indicators, such as accumulated rainfall during flood season, impervious surface area, GDP per square kilometer of land and urban road network density have significant positive correlation with flood. And the flood vulnerability level has been on the rise for the past 20 years, especially in the past 5 years, when the degree of vulnerability to flood has been relatively high. The results are consistent with the actual situation of flood vulnerability in Nanjing city. This study therefore provides decision-making recommendations for the relevant management departments to improve the flood management capability, and reduce the flood vulnerability.