在水文模型率定过程中,不同目标函数侧重于径流模拟的不同层面。为了探究目标函数选择的不确定性对模型参数和径流模拟的影响,将HyMod模型应用于黄河源区,选择纳什效率函数f_NS、总水量平衡误差函数f_t、低水量误差函数f_d和高水量误差函数f_g作为目标函数,采用遗传算法(GA)分别率定出不同目标函数下对应的最优参数值,并依次代入水文模型模拟水文过程;通过对比分析年尺度及年内各月实测与模拟径流过程、纳什效率系数NSE、决定系数R²和均方根误差RMSE评价指标来探究参数率定的目标函数不确定性对年尺度水资源演变过程的影响。结果表明:当HyMod模型应用于黄河源区水文模拟的率定期和验证期时,目标函数选择的不确定性对各评价指标的影响差异明显,如f_NS目标函数下NSE值最大,在f_g下次之,在f_d下最小,此外率定期模拟精度优于验证期;同样,目标函数不确定性对不同特征时期径流的影响差异显著,其中,f_NS和f_t目标函数下,非汛期分别高估和低估模拟流量。研究成果可为水文模型参数率定目标函数的选择提供理论参考。

In the process of hydrological model calibration, different objective functions focus on different aspects of runoff simulation. In this paper, the HyMod model is applied to the source region of the Yellow River to explore the uncertainty of choosing objective functions and the influence of such uncertainty on runoff simulation. The Nash efficiency function f_NS, the total error of water balance function f_t, the low water error function f_d and high error function f_g are selected as objective functions. The optimal parameter sets corresponding to different objective functions are calibrated by using the genetic algorithm (GA), and are put into hydrological model in turn to simulate the hydrological processes. By comparing the yearly scale and monthly scale measured and simulated runoff processes as well as evaluation indicators such as Nash-Sutcliffe efficiency coefficient(NSE), cofficient of determination(R²) and Root Mean Square Error(RMSE), the impact of uncertainty of objective function on water resource evolution in yearly scale can be obtained. Results demonstrate that the impact of uncertainty on evaluation indicators differs remarkably when HyMod model is applied to the hydrological simulation in the source area of Yellow River in calibration and validation periods. For example, under f_NS objective function, the value of NSE is the largest, followed by f_g and f_d in sequence. In addition, the accuracy of calibration period is superior to that of verification period. Similarly, the impact of uncertainty differs in various characteristic periods. For example, under f_NS and f_t objective functions, the simulated flow values in non-flood season are respectively overestimated and underestimated. The research findings offer theoretical reference for the selection of objective functions in calibrating hydrological model parameters.