Efficient simulation and prediction method for extreme rainfall-induced flooding processes in ultra-large cities
HOU Jingming
WANG Tian
LI Donglai
PAN Xinxin
YANG Yongping
ZHANG Shijie
CHEN Guangzhao
MA Liping
LYU Jiahao
GUAN Baojun
Abstract：Under the background of global climate change,the risk of extreme rainfall-induced flooding in cities is increasingly intensifying.How to achieve efficient and accurate prediction of such flooding processes in ultra-large cities has become a core scientific issue and engineering demand for disaster prevention,mitigation,and enhancing urban resilience.To address the problems of low computational efficiency and poor real-time performance of traditional hydrodynamic models,a dual-driven high-efficiency simulation and prediction method that integrates physical mechanism models with artificial intelligence(AI)algorithms is proposed.By incorporating runoff generation calculations,two-dimensional hydrodynamic routing,and the coupled mechanism of pipe networks and surface flow,a high-precision numerical model of rainfall-flood processes is constructed.Through non-uniform grid optimization and multi-GPU parallel computing,efficient and accurate simulation of extreme rainfall-induced flooding in ultra-large cities is achieved.Using training data generated by the physical model to drive the AI prediction model enables rapid forecasting of flooding processes.Taking Xi'an City in Shaanxi Province as an example,the dual-driven prediction model based on the integration of physical mechanisms and AI algorithms achieves a computational speed about 287 times faster than traditional hydrodynamic models,with relative error below 10%,and realizes rapid classification of waterlogging risks in the main urban area under extreme rainfall.This method provides efficient technical support for rapid early warning and scientific response to urban waterlogging caused by extreme rainfall in ultra-large cities.
Keywords：ultra-large citiesextreme rainfallwaterloggingnumerical simulationmechanism-AI dual-drivenrapid and accurate predictionhydrologic-hydrodynamic model
Publication Date：2025-09-30
Online Publishing Date：2025-10-21(First online date of this platform, not the publication date of the document)
Pages：10( 19-28 )