降雨时间分辨率与插值方法对城市内涝模拟影响研究

Effect of rainfall time resolutions and interpolation methods on urban waterlogging simulations

为研究不同时间分辨率降雨资料对模拟结果的影响,使用不同插值方法提高降雨时间分辨率进行城市内涝模拟并确定最佳插值方法。以西安市天福和园小区为研究区域,利用GAST-SWMM耦合模型,选取纳什系数和峰值误差作为评判标准,对研究区域模拟和实测流量过程线进行对比。研究表明:经两场实测降雨过程验证,模型纳什系数分别为0.85和0.81,表明构建的一二维耦合模型精度较高;在本文设置的五种时间分辨率中,当降雨时间分辨率为5 min时,模拟精度最高,随着降雨时间分辨率的降低,模拟流量与实测流量过程线吻合程度逐渐变差,且时间分辨率为60 min时模拟结果最差;选取分形插值、拉格朗日插值、线性插值三种方法,将实测降雨数据时间分辨率从30 min和60 min插值为5 min,模拟结果均显示与实测吻合程度最好的是分形插值方法,拉格朗日插值方法次之,线性插值方法最差。本研究可为用于城市内涝模拟的降雨资料选取及处理提供依据。

This study examines the influence of rainfall data with different time resolutions on urban waterlogging simulations using a 1D-2D coupling model,namely GAST-SWMM(GPU Accelerated Surface Water Flow coupled with Transport Model-Storm Water Management Model).Different interpolation methods are used to improve time resolution,and the best one is determined.Nash coefficients and peak errors are compared to evaluate the simulated flow hydrographs against the measurements for Tianfu Heyuan District,Xi'an City.We examine three methods-fractal interpolation,Lagrange interpolation and linear interpolation-using a time interval range of 5 min to 60 min.The results show the Nash coefficient of the model is 0.85 and 0.81 for two rainfall events,indicating its high accuracy.Of the five sets of different rainfall time resolutions studied,the one of 5min gives the highest simulation accuracy.A lower resolution results in a greater deviation of the simulation from measurements,and the worst case occurs when the interval is 60 min.We find that fractal interpolation has the best agreement with measurements,followed by Lagrange interpolation,and linear interpolation is the poorest.This study can provide a basis for the selection and processing of rainfall data for urban waterlogging simulation.