基于多模型适用的大通河流域洪水临界雨量阈值研究与比较分析

Comparative analysis of three models to the study on risk precipitation threshold value in Datong river basin

基于大通河流域地理信息资料、气象资料、水文资料,采用SWAT模型、HBV水文模型、统计模型对大通河流域进行临界雨量适用研究。结果表明,大通河流域中雨以上量级降水主要出现在5—9月,2007—2011年中雨以上量级降水累计次数为41次,频次和强度增加明显。通过大通河流域3种模型适用的结果以及实际洪水发生的检验情况来看,SWAT水文模型在大通河流域适用效果最好,率定期模型效率系数为0.87,相关系数为0.94,验证期模型效率系数为0.62,相关系数为0.83,模拟流量和观测流量的拟合率满足模型检验的标准;HBV模型率定期确定性系数为0.84,NSE为0.61,验证期确定性系数为0.81,NSE为0.51,率定后HBV水文模型在大通河流域具有较强的适用性,但与SWAT模型拟合结果相比,HBV水文模型模拟流量和实际流量拟合率较低。在流域内如果有较长时段的水文资料,采用水文模型确定临界雨量较为科学,而统计模型基于降水量与水位差的回归关系,没有考虑从降雨形成地表径流的过程,确定的阈值偏大。

Based on data from GIS information,meteorological stations and hydrological stations,the applicability study of risk precipitation in Datong river basin was carried out using SWAT model,HBV model and statistical model.The results indicated that,the middle rain and above the middle rain in Datong river mainly appeared from May to September,the cumulative number of times was 41 from 2007 to 2011,and the trend of frequency and intensity showed an obvious increasing.Through the results of the application of three models in Datong river basin and the test of the actual flood,the SWAT model was the best method to modify the runoff in the Datong river basin,with efficiency coefficient of 0.87 and R2 of 0.94 during the validation period,and with efficiency coefficient of 0.62 and R2 of 0.83 during the calibration period,and simulating runoff and observed runoff reached the testing standard;The coefficients of determination and NSE were 0.84 and 0.61 during the validation period with HBV model,but they were 0.81 and 0.51 during the calibration period,there was no good fitting rate using HBV model compared with SWAT model.If there is a long period of hydrological data in the basin,it is more scientific to use hydrological model to determine the critical rainfall.However,the statistical model,based on the regression relationship between precipitation and water level difference,does not consider the process of surface runoff formed from rainfall,so the threshold value determined is too large.