基于全局敏感性分析方法的WASP模型不确定性分析

为了定量讨论WASP(The water quality analysis simulation program,水质分析模拟程序)模型中各参数对模拟结果的影响及其不确定性,本文在Simulink环境下对模型进行重组,利用Sobol方法,对影响DO、CBOD、NH3-N、NO3--N等4项输出的各参数敏感性进行了研究,重点讨论了输入条件如时空变化对参数敏感性的影响.结果表明:Sobol全局敏感性分析方法能够有效地筛选出WASP模型中对DO、CBOD、NH3-N、NO3-变化模拟较为敏感的参数,在实现模型参数“本土化”中具有较大的应用潜力;对比总敏感性与一阶敏感性指数之间的差异, WASP模型当中的若干参数(如GP1、PNH3、anc等)耦合性较强,不宜使用一次一个变量法对这些变量进行敏感性分析;在参数敏感性变化方面,即使对于同一条河流,WASP的参数敏感性指数会因边界条件的时空变化而发生改变.敏感性受空间变化的影响不大,但受时间(季节)变化的影响显著;输入变量随时间的变化,会引起模型不确定性在各季节上的较大的差异;WASP模型当中存在明显的“异参同效”作用,仅靠传统方法进行全局寻优率定,并不能体现水体各组分之间相互转化的过程,还需要结合实验对关键参数进行率定.

Uncertainty analysis of WASP based on global sensitivity analysis method

To quantitatively evaluate parameters influence and uncertainty of model, WASP model was reorganized by Simulink in this paper. Sensitivity of the parameters related to model output of DO, CBOD, NH3-N and NO3--N was studied based on Sobol method. In particular, sensitivity changes with temporal and spatial variations of input were discussed. Global sensitivity analysis of Sobol method can identify the most sensitive parameters of the process simulation; The significant difference between the total sensitivity and first-order sensitivity index showed that some parameters (e.g. GP1, PNH3, anc, etc.) of the WASP model were strongly coupling, and“One variable at a time”method was inappropriate to evaluate the sensitivity of these parameters; Sensitivity index of WASP parameter changed with the temporal and spatial variation of boundary conditions, even for the same river; WASP model showed an obvious equifinality for different parameters, so the traditional methods such as the global optimization calibration, failed to simulate the mechanism process. Experiments were required as a verification for calibration of key parameters.