基于门限极值理论的湖泊水质参照状态的确定

在前人工作的基础上,利用门限极值的广义Pareto分布理论和超出阈值峰(Peak Over Threshold,POT)方法,提出了一种确定湖泊参照状态浓度的新方法.该方法不仅能够给出更为精确的置信区间,而且克服了广义极值分布理论取用数据浪费等缺陷.将该方法应用到太湖的水质基准参照状态中,通过POT方法对太湖8个站点1995~2006年总氮(TN),总磷(TP)和叶绿素a(Chl-a)的数据进行预处理,分别以-1.0mg/L, -0.05mg/L与-4μg/L作为它们观测值相反数的门限值,结果表明观测值的相反数符合广义Pareto分布,验证了方法的可行性.推荐采用25%分位点的值作为太湖总氮,总磷和叶绿素a的参照状态,即太湖的参照状态是:总氮0.66mg/L;总磷0.023mg/L;叶绿素a为1.27μg/L.最后分别得出了它们各自的95%置信区间,而且其精度明显高于广义极值分布理论结果.

Estimation of the lake quality reference condition based on the threshold extreme theory

A new method was established to calculate the lake nutrient reference condition based on the extreme threshold theory combined with the Peak Over Threshold. The new method not only revealed statistical inferences, but also overcame the weakness such as wasted observational data in generalized extreme value (GEV) theory. More accurate confidence intervals for parameters and substance concentrations could be obtained using this new method. The method was used to estimate the reference conditions of Taihu Lake. The observed data for total nitrogen (TN), total phosphorus (TP) and chlorophylla (Chl-a) at eight sites located in Taihu Lake during 1995~2006 were pre-treated by the Peak Over Threshold. The results revealed that negative values of TN, TP, and Chl-a, which the threshold values are ?1.0mg/L, ?0.05mg/L, and ?4μg/L, fitted the generally Pareto model well. It was recommended to use 25th percentile as reference conditions. Thus, the values of the reference conditions for TN, TP and Chl-a were 0.66mg/L, 0.023mg/L and 1.27μg/L, respectively. The 95% confidence intervals were also obtained, more accurately than using generally extreme theory.