WASMOD模型参数敏感性与相关性分析

Sensitivity and Correlation Analysis of Parameters in WASMOD

对水文模型参数进行敏感性和相关性分析,可以减少率定参数的数量、提高模型运行效率,并可以为优化和改进模型结构提供依据。本文采用局部分析法和相关系数法讨论了WASMOD模型中的五个参数对目标函数的敏感程度以及参数之间的相关程度。研究发现,在黑河流域上游山区的径流模拟过程中,WASMOD模型中影响降水形态的温度参数a1对目标函数最为敏感;其次是影响地表径流的参数a6和影响融雪过程的温度参数a2;基流参数a5最不敏感。模型参数之间的相关性不是很强,这也说明,模型结构已得到很好的优化;为保证模型的模拟效果,模型参数的数量不能减少。

Model parameter sensitivity and correlation have great effects on model calibration, runs and efficiency. Here, we focused on the sensitivity and correlation of parameters in a water balance model the water and snow balance modeling system （WASMOD）, a conceptual lumped model with a snowmelt module for water balance computation. It can be used in streamflow series extension, water balance calculation, hydrological responses to climate change simulation and runoff forecast. The sensitivity of parameters in WASMOD was analyzed using local sensitivity analysis. Correlations between model parameters were analyzed using the correlation coefficient method. The upper reach of the Heihe River basin was selected as the study area. Results showed that WASMOD can give a fairly satisfactory runoff simulation with coefficients of Nash-Sutcliffe reaching over 0.92 in both the calibration and validation periods. The most sensitive parameter to the objective function was temperature parameter a 1 , which is related to the fact that both rainfall and snowfall contribute to the recharge of runoff. The next most sensitive was the runoff parameter a 6 and temperature parameter a 2 . Base flow parameter a 5 was the least sensitive, indicating that base flow in the study area is fairly stable and variation in total runoff is mainly derived from variation in surface flow. No strong correlation was found, the two highest correlation coefficients reached -0.68 and 0.64 between parameters a 1 and a 2 , and a 5 and a 6 , respectively.