城市暴雨内涝模拟模型优化与精度验证

This paper aims to optimize the performance of a previously developed hydrodynamic model for urban flood simulation. Our major task includes calibration of two key parameters in the runoff generation and flood routing modules, and verification of the precision of the model output. In the runoff producing module, we focused on optimization of Curve Number (CN) values. To achieve this purpose, the method of Linear Spectral Mixture Analysis (LSMA) was employed to extract terrestrial information of vegetation coverage, soil categories and impervious land use from Landsat TM images, based on which a specific CN value could be defined for each unit in the hydrodynamic model. As for the flood routing module, we reset the manning coefficient via integrating previous empirical value and findings from calibration experiments conducted in this study. Verification experiments show both the calibration of CN values and manning coefficient promotes the model's simulation precision. Using the Vegetation-Impervious Surface-Soil (V-I-S) raster layers, in which the CN values incorporate more accurate information of vegetation coverage and soil categories, as input for the hydrodynamic model, are able to lower the extreme abnormal values of simulated water depth, and provide more reasonable estimation of water volume and inundation area. After resetting the manning coefficient for different land uses, the simulated maximum water depth increased notably (almost 100 mm), compared with previous model outputs without calibration of this parameter. Through our calibration study, it is safe to say that manning coefficient is a sensitive and critical parameter and deserves further attention in the extension research for optimization of the flood routing module.