偏最小二乘回归法在非点源污染年负荷量估算中的应用

针对非点源污染数据资料稀缺，机理模型难以建立，而统计模型所需资料相对比较易于获得，计算简便直观，并且在水环境规划管理中对非点源的年负荷量预测、估算的指导意义要大于对某一时点、某一断面的负荷量的模拟预测的情况，尝试将综合了多元回归、主成分分析、典型相关分析的偏最小二乘回归法用于建立流域非点源污染年负荷量的模拟估算模型，同时运用水文分析法以淮河淮南段为例进行计算并对两种方法进行对比，分析各自优缺点及其适用范围。应用表明，水文分析法适用于一年以上的非点源污染负荷量估算而对单次或短期的估算误差较大。偏最小二乘回归法解决了多元回归时变量具有多重相关性的问题，能够在样本数量少的情况下建模，提高相关分析的精度，具有较强的解释能力。

APPLICATION OF PARTIAL LEAST SQUARE REGRESS METHOD TO ESTIMATE THE ANNUAL LOAD OF NON-POINT SOURCE POLLUTION

The instructional meaning of estimating the annual load of non-point source pollution is higher than forecasting the load of some time or some section.Different from the mechanism model that is difficult to be built because of the scarcity of data,the statistic model needs less data.The partial least square regress method integrating the multi-regression,the principal component analysis and canonical correlation analysis,were experimentally employed for estimating the annual load of non-point source pollution.At the same time,hydrological method was also used.Illustrated with Huaihe River in Huainan,the two methods were compared with their strongpoints and weaknesses.The hydrological method is applicable to the estimation longer than one year.It shows that the partial least square regress method can solve the problem of multi-relativity and can be used without enough data.It can improve the precision of relativity analysis and has better interpretative ability.