水文模型中不同目标函数的影响分析比较

Effects of Objective Functions on Performance of Hydrological Models

在水文模型中,目标函数的选择对参数率定至关重要,不同的目标函数可以得到不同的模拟结果。本文以三水源新安江模型为例,采用SCE-UA算法,选定三个不同的目标函数(平方均方误,对数均方误和平方根均方误)最小为目标函数分别进行参数优化,比较其优化结果在高、低水期的精度,最后分析不同目标函数对模拟结果的影响。研究发现:以平方均方误为目标函数的时候,高水期的模拟效果较好;以对数均方误为目标函数的时候,低水期的模拟效果较好;以平方根均方误为目标函数的时候,在整体上的模拟效果较好。

Selection of an objective function is of great significance for the calibration of rainfall-runoff models. Different objective functions could cause different simulated results. This paper has used Xinanjiang model and SCE-UA algorithm, and chosen three different objective functions： mean squared error of square transformed, mean squared error of logarithmic transformed and mean squared error of square root transformed, to calibrate and simulate the observed hydrologic processes. The precision of the simulat- ed results in the periods of high-flow and low-flow are compared. The effects of different objective functions on the performance of hydrological models have been analyzed： MSSE puts more emphasis on high-flow simulations, MLSE on low-flow simulations and MSRSE gives an intermediate picture of the overall hydrograph fit.