洪水灾害危险性评估方法研究综述

Flood Hazard Assessment Methods: Research Review

对洪水危险性的合理分析与预判是人类管理洪水灾害的重要途径之一。为此,本研究系统地回顾了世界范围内洪水灾害危险性评估的相关研究,总结并对比了四种主要评估方法,即数值模拟分析法、多指标决策分析法、机器学习法、遥感遥测法。从模型的结构、参数及输入三个角度分析了目前洪水危险性评估系统中存在的不确定性。相关研究需要加强水文等相关数据的收集与积累,深入洪水机理研究。数值模拟分析需关注复合洪水的模拟并开发综合性多向耦合模型;多指标决策分析应关注数据分解技术和综合权重法的发展;机器学习需建立一套高效率求解优化参数的算法,同时加强与洪水形成物理过程的机理性连接;遥感遥测法则应关注多源数据的融合和分解,以及对异常水体识别算法的开发与改进。

Rational analysis and prediction of the flood hazard are critical for the management of flood disasters. This study systematically and widely reviews the relevant studies on flood-hazard assessment (FHA). Four primary methods of FHA, including numerical simulation analysis (NSA), multi-criteria decision analysis (MCA), machine learning (ML), and remote sensing (RS), are summarized and compared. The uncertainty of model structure, parameters, and inputs is also analyzed. Finally, the advice for current research is as follows: the collection of hydrological data and the study of the flooding mechanism are needed to be strengthened. Specifically, more attention should be paid to the simulation of the com-pound flood, and the multi-directional coupling model should be developed comprehensively. The da-ta-decomposition technology and weight method should be improved, an optimization parameter algo-rithm with higher efficiency is essential to be established, and the relevant physical process of the flood needs to be elaborated. The decomposition and fusion of multi-source data should be developed in the future, and the identification of abnormal water bodies needs to be explored.