This paper presents an integrated modeling framework aiming at accurate predictions of flood hazard from heavy rainfalls. The accuracy of such predictions generally depends on the complexity and resolution of the employed model components. We propose an integration of complementary models in one framework that facilitates GPUs to improve accuracy and simulation time. The spatially distributed runoff model integrates surface flow routing based on the full shallow water equations, infiltration based on the Green–Ampt equation, and interception. In urban areas, the runoff model is coupled with the Storm Water Management Model (SWMM). The integrated model is validated and tested on laboratory, rural and urban scenarios with regards to accuracy and computational efficiency. The GPU acceleration yields speedups of 1000 times compared to a CPU implementation and enables the coupled simulation of flash floods at 1 m resolution for an urban area of 200 km${}^2$ in realtime.

本文提出了一个综合建模框架，旨在准确预测暴雨造成的洪水灾害。这种预测的准确性通常取决于所采用的模型组件的复杂性和分辨率。我们建议在一个框架中集成互补模型，以促进 GPU 提高准确性和模拟时间。空间分布径流模型集成了基于全浅水方程的地表流路由、基于 Green-Ampt 方程的入渗和截流。在城市地区，径流模型与雨水管理模型 (SWMM) 相结合。该集成模型在实验室、农村和城市场景的准确性和计算效率方面进行了验证和测试。${}^2$实时。