物种敏感性分布法拟合函数的拟合优度评价  

Evaluation of the goodness of fit of the species sensitivity distribution fitting function

作　　者：董明明 牟力言 秦莉 安毅 林大松 DONG Ming-ming;MOU Li-yan;QIN Li;AN Yi;LIN Da-song(Agro-Environmental Protection Institute,Ministry of Agriculture and Rural Affairs,Tianjin 300191,China)

机构地区：[1]农业农村部环境保护科研监测所,天津300191

出　　处：《农业环境科学学报》2021年第3期544-551,共8页Journal of Agro-Environment Science

基　　金：国家自然科学基金项目(41877403)。

摘　　要：为明确物种敏感性分布(SSD)法适用于酸性土壤条件下的最优拟合函数,以均方根(RMSE)和残差平方和(SSE)为评价指标,系统分析了不同pH和累积概率条件下5种常见拟合函数(Log-logistic、Gamma、Log-normal、Weibull和BurrⅢ)的拟合优度。研究表明:两种pH条件(pH 5.5和pH 6.5)下5种拟合函数的拟合优度无明显差异,其中Log-logistic、BurrⅢ函数的SSE值分别为0.021、0.024和0.169、0.191,RMSE值分别为0.038、0.040和0.106、0.113,两者拟合效果较好;但不同累积概率条件下拟合函数的拟合优度存在一定差异,其中低累积概率(p≤20%)条件下Log-logistic与Gamma的SSE值分别为2.45×10^(-4)和2.46×10^(-4),RMSE值分别为4.04×10^(-3)和4.05×10^(-3),两者拟合效果较好;中累积概率(20%80%)条件下BurrⅢ与Log-logistic的SSE值分别为0.151和0.203,RMSE值分别为0.100和0.116,两者拟合效果较好。研究表明,酸性土壤(pH≤6.5)中低、中累积概率(0≤p≤80%)条件下优先推荐使用Log-logistic拟合函数,而高累积概率(p>80%)条件下优先推荐使用BurrⅢ拟合函数。To utilize the optimal fitting function of the species sensitivity distribution(SSD)method for acidic soil conditions,this study uses the root mean square error(RMSE)and the sum of squares for error(SSE)as evaluation indicators to systematically analyze the goodness of fit of five common fitting functions(i.e.,Log-logistic,Gamma,Log-normal,Weibull,and BurrⅢ)under conditions of cumulative probability and different pH.The study have shown that there is no significant difference in the goodness of fit of the five fitting functions under two pH conditions(pH=5.5 and pH=6.5),among which the SSE values of Log-logistic and BurrⅢfunctions are 0.021,0.024 and 0.169,0.191,respectively,whereas RMSE values of the two functions are 0.038,0.040 and 0.106,0.113,respectively,the fitting effect for the two functions is better.But the goodness of fit of the fitting function under different cumulative probability conditions is different,under conditions of low cumulative probability(p≤20%),the SSE values of Log-logistic and Gamma are 2.45×10^(-4)and 2.46×10^(-4),and the RMSE values of the two functions are 4.04×10^(-3)and 4.05×10^(-3),respectively,and the fitting effect for the two functions is better.Under conditions of cumulative probability(20%80%),the SSE values of BurrⅢand Log-logistic are 0.151 and 0.203,and the RMSE values of the two functions are 0.100 and 0.116,respectively,and the fitting effect of the two functions is better.Based on the above results,the Log-logistic fitting function is preferred under acidic soil conditions(pH≤6.5)with low and medium cumulative probability(0≤p≤80%),and BurrⅢfitting function is preferred under conditions of high cumulative probability(p>80%).

关 键 词：酸性土壤 镉 物种敏感性分布 拟合函数 拟合优度 

分 类 号：X825[环境科学与工程—环境工程]