在极端降雨频发和城市化持续推进的背景下，越来越多的人口和财产被置于洪水的威胁之下。城市雨洪灾害已经演变为全球关注的焦点，城市雨洪模拟则成为了学界研究的热点。但城市降雨-下垫面条件复杂，空间变异性显著，这对模拟的精细化程度提出了要求，也为模拟的准确性带来了未知的影响因素。而由此产生的诸如应当如何考虑关键的城市要素、如何对精细化模型网格科学的进行升尺度、以及地面雨量计应该被部署为多大的密度等一系列关键问题迫切需要回答。围绕上述问题，本研究基于天空地高精度综合观测数据和成熟的城市雨洪模拟技术，在居住小区（0.04 km2）、清华校园（~3 km2）、以及清河流域（~200 km2）等不同的城市水文尺度上开展了系统的工作，旨在强化城市精细化水文过程认识，提高城市精细化水文模拟能力。研究发现：1）建筑屋面微尺度汇流过程有助于降雨产流汇聚，继而增大集水区径流响应峰值，在特定降雨-下垫面条件下该增幅将接近10%，有必要在城市雨洪模拟中做具体考虑；2）模型网格尺度的变化会影响地表汇流特性描述，继而改变洪水模拟结果，且在不同的降雨条件下表现出明显差异。模型关键汇流参数（透水区地表糙率和次网格汇流比例）具有显著的尺度（模型网格）依赖性，应根据特定网格尺度下对汇流过程的具体描述计算确定；3）低空间分辨率数据无法有效捕捉关键的降雨时空特性，继而将影响洪水模拟。总体来看，降雨数据空间分辨率降低会导致峰值流量的严重低估，具体低估程度则与所关注的水文尺度密切相关。以±20%作为径流模拟的最大允许偏差，排水片区和流域尺度分别要求降雨空间分辨率在500 m 和10 km以上，而在街区尺度则至少要达到300 m；4）相比于降雨空间特性，降雨时间特性在洪水预报中更加重要。就洪水峰值的模拟而言，30 min最大降雨强度是最为重要的降雨时空特性。本研究进一步厘清了城市雨洪模拟中可能的不确定性来源，有助于科学推动精细化城市暴雨洪水模拟技术以成本-效益平衡的方式持续发展。
In the context of frequent extreme rainfall and continuous urbanization, increased population and property are put under the threat of flooding. Urban flooding has become a global concern, while urban stormwater modeling has become a hotspot for research. However, the condition of both rainfall and underlying surface in urban environments present extreme complexity and significant spatial variability, which calls for refined modeling technology and brings numerous unknown influence factors to flood forecasting simultaneously. Also, a range of questions, e.g. how to represent the key elements of urban, how to conduct spatial upscaling for the refined model grid cells scientifically, and what should be the rational density of rain gauge deployment, are still waiting for answers. Based on the comprehensive observation data from multiple sources and mature urban stormwater modeling technology, this study carried out systematic work on different urban hydrological scales (e.g. a neighborhood of 0.04 km2, the campus of Tsinghua University (~3 km2), and the Qing river basin in Beijing (~200 km2)) concerning the above problems, aiming to strengthen the understanding of refined urban hydrological processes and enhance the capacity of refined urban hydrological simulation.Results show that 1) the microscale routing process on building roof conduce to the convergence of runoff and thus increase the peak flow at the catchment outlet, which may approximate 10% under given conditions and deserves special attention in the urban stormwater modeling; 2) changes in the spatial scale of model grid cells will affect the depiction of surface routing and then influence flood simulation. While the impacts on flood simulation show an obvious contradiction between heavy and light rainfall events. The key routing parameters (i.e. surface roughness of pervious area and subgrid routing percent) present obvious scale dependence, the value of which should be determined according to the surface routing characterization under the specific scale; 3) the data with coarse spatial resolution can not capture critical rainfall characteristics and will largely influence urban flood forecasting. In general, the decrease of rainfall spatial resolution leads to underestimated peak flow, while the underestimation degree is closely related to the hydrological scale of concern. Taking ±20% as the maximum permissible bias in discharge simulation, the spatial resolution of rainfall data should be higher than 500 m and 10 km at the district and catchment scale respectively, while at least 300 m for the block scale; 4) the temporal characteristics of rainfall are more important in flood forecasting compared to the spatial characteristics. For the simulation of flood peaks, the 30-min maximum rainfall intensity is the most important rainfall characteristic. This study further clarifies the potential uncertainty in urban flood modeling and helps to enhance the technology of refined urban stormwater modeling in a cost-benefit balanced way.
