Research On Efficient High-resolution Simulation And Forecasting Method For Urban Inundation Process

Urban inundation forecasting is an important non-engineering measure to reduce the risk of flooding and ensure the safety of human and property.To address the problems that traditional urban inundation forecasting methods have low spatial resolution and short forecasting period,and cannot predict the inundation process accurately and timely,this paper constructs a urban inundation forecasting model with high-resolution and long forecasting period by integrating a numerical meteorological forecasting model and an efficient high-resolution numerical model of flooding process.The urban inundation forecasting model generates rainfall data by GRAPES MESO numerical meteorological model,which is a new numerical forecasting model system with multi-scale general dynamical model as the core and unified software programming standard as the platform,and can update the future 72h rainfall data of the simulated area twice a day(00:00 and 12:00)with a spatial and temporal resolution of 3h and 10km.The numerical model of flooding process adopts the Green-Ampt model to calculate the soil infiltration process;the A.P.J.D E ROO method integrates the Aston model and the vegetation canopy retention capacity equation to accurately calculate the plant retention process;the surface runoff and pipe network drainage process are calculated by numerically solving the 2D shallow water equation and the 1D St.Venant equation.In addition,the numerical model uses high-resolution DEM data to characterize the complex urban surface morphology,and introduces GPU par 2allel accelerated computing technology to achieve a significant increase in computational efficiency without reducing computational accuracy.In order to further improve the accuracy of rainfall forecast data,a data reconstruction formula based on regression analysis method is proposed.At the same time,an analysis method is constructed to evaluate the flooding uncertainty caused by the spatial non-uniformity of rainfall,and the deviation amount of simulation results caused by the uneven spatial distribution of rainfall under different conditions of inhomogeneity coefficient Cv,rainfall recurrence period and rainfall peak coefficient is analyzed.The results show that the spatially non-uniform distribution of rainfall has an impact on the spatial distribution of inundation,inundation volume,hazard level and inundation time,and the inundation space will become more concentrated;the peak inundation volume has a tendency to increase;the more hazardous area will expand;and the peak inundation time will lag behind.Based on the modeling completion,the accuracy,stability and efficiency of the model were verified by a typical urban inundation process under six different characteristic rainstorm conditions.The research work adopts the main flooding influencing factors(inundation depth and inundation area)affecting urban socio-economic activities as analytical indicators to evaluate the forecast accuracy,and the model is driven by three types of rainstorm data(observed forecast rainfall data,original forecast rainfall data,and reconstructed forecast rainfall data)for simulation calculations.The comparative analysis of the simulation results under the three types of storm data shows that the forecast model can correctly forecast the inundation locations and is computationally efficient,taking only 2.45 hours to complete the simulation of the 6-hour inundation process in an area of 5.37 million grid cells.The average relative errors of inundation area and water depth are 35%and 2.5%,respectively.It can be seen that the flood forecasting model has a good forecasting performance,which is important for guiding flood control and sustainable urban stormwater management.