城市暴雨内涝综述：特征、机理、数据与方法

建成环境的高空间异质性与致灾过程的复杂性给城市暴雨内涝研究带来巨大的挑战,具体表现为模型代表性不够、计算效率低、基础数据和验证数据匮乏。以机器学习为代表的人工智能技术、高分遥感和互联网大数据的快速发展则为城市暴雨内涝研究提供了新的契机。论文结合人工智能、高分遥感和互联网大数据等新技术发展,从特征、机理、数据与方法4个维度对暴雨内涝的研究现状和发展趋势进行了系统总结,主要结论包括：① 暴雨内涝具有短历时性、空间散布性、连锁性和突变性,其热点呈现空间上的动态迁移特征。② 降雨时空特征和城市化程度决定暴雨内涝灾害的量级,地形条件尤其是微地形则决定发生位置和内涝频率。地形控制作用指数(topographic control index, TCI)对暴雨内涝发生位置具有良好的指示能力。③ 排水管网、高精度地形和不透水面分布是暴雨内涝模拟的关键基础数据;降雨过程的高时空变异性是暴雨内涝近实时预报预警的主要瓶颈,需要充分利用天气雷达观测提高其精准度;互联网众包大数据是获取高空间覆盖度暴雨内涝灾情信息的新途径,但也面临不同类型信息融合、提炼和质量控制的挑战。④ 结合水动力模拟与机器学习可建立兼具物理基础和计算效率的暴雨内涝模拟方法,是实现近实时模拟与快速预报预警的有效途径。

A review on urban pluvial floods:Characteristics,mechanisms,data,and research methods

The spatial heterogeneity of urban built environments is a primary challenge in the research of urban pluvial floods in terms of the model representativeness, computational efficiency, and data requirements. The development of new technologies, including artificial intelligence, big data, and remote sensing, provides opportunities for the research of urban pluvial floods, such as efficient approaches and high-resolution data. This study conducted a comprehensive review of the research progress on urban pluvial floods from four perspectives—flood characteristics, mechanisms, data, and research methods, and finally came to four conclusions: 1) Urban pluvial floods have typical features such as short duration, scattered and evolving spatial distribution, chain effect, and sharp increase of losses at the critical scenario. 2) Micro-topography plays an important role in the spatial distribution of urban pluvial floods, and the topographic control index shows the potential of identifying frequently flooded areas. 3) The highly variable rainfall processes are the bottleneck in the near-real-time flood simulation, and the radar rainfall data provide a solution. Internet-based big data provide a new way to extract flood inundation data with high spatial coverage, but still face the problems of quality control and fusion with multi-sources data. 4) Machine learning could be coupled with hydrodynamic models to improve the efficiency of near-real-time flood simulation.