基于不确定性分析的IMPULSE模型预测能力评价

Assessment of IMPULSE model simulation capability based on uncertainty analysis

为解决分布式参数非点源污染(IMPULSE)模型不确定性分析中采样量和计算量过大的问题,在Bayes概率理论基础上,构建了基于Sobol序列的GLUE算法,用来描述多种扰动因素共同作用下的全局参数不确定性,从而对模型预测能力进行全面评价。将该方法应用于IMPULSE模型,对分布式参数的全局进行不确定性分析。结果表明:该模型结构优良,具有良好的预测能力,对空间不确定性有较高的预测稳定性和鲁棒性,可以满足实际流域污染模拟需要。

A Sobol-sequenee-based GLUE （generalized likelihood uncertainty estimation） algorithm was developed based on Bayesian probability theory to improve analysis of huge samples for the uncertainty analysis of the distributed parameters nonpoint source pollution model. The method describes the global uncertainties of the distributed parameters for multiple disturbances to assess the model＇s simulation capability. The method was applied to IMPULSE （integrated model of non point sources pollution processes） model. The results show that the model has good simulation capability, with steady, robust results for spatially distributed uncertainties and can be applied to watershed simulations with spatial uncertainties.