观测误差对非线性环境模型参数识别的影响

Effects of observation errors on parameter identification of a nonlinear environmental model

采用SCE UA方法和RSA方法 ,比较了在不同的观测误差条件下 ,优化方法和不确定性分析方法对于非线性模型参数识别的影响 .分析表明 ,RSA方法是在具有观测误差的数据条件下进行参数识别的一种有效手段 .同时 ,通过比较分析 ,发现RSA方法预测的浓度平均值和概率最大值与“真实值”并不完全一致 .因此 ,在使用RSA方法时 ,应该充分考虑预测区间 。

Parameter identification is a crucial procedure for model application.Observation errors are widely present in environmental data.This paper focused on the effects of observation errors on parameter identification of nonlinear environmental models.It specifically compared the identification results by the optimization and uncertainty analysis approaches,taking the SCE-UA and RSA methods as representatives respectively.It was found that the RSA was an effective approach to reduce the impacts of observation errors.It was also showed that the idntified average values or the statistically most likely values of the RSA were often not in consistence with the “true values”.It is there fore suggested that the intervals be considered as identified parameter uncertainty so as to reduce the decision-making risks.