应用多元统计分析理论预测城市大气污染

本文分析研究地区31个环境,单元中4个环境因子与5个大气污染指标的相关性.通过对聚类中各因子最佳分割点的确定,对各子聚类的合并判断以及对各指标的判别筛选,探索大气污染的发生、发展与环境中诸因子存在状态的定量关系,从而实现对该区大气污染状况的定量化预测.

Application of the Theory of Multivariate Statistical Analysis to the Prediction of Air Pollution in Urban Environment

A stepwise cluster analysis method toward a multitude of environmental variables was advanced for the predic-tion of air pollution in urban environment. The method has improved monovariate AID(Automatic Interaction Detection) algorithm. It can effectively deal with continuous and discrete variables, as well as the nonlinear relations among variables, Compared with classical modelling theories, the method provides a new way for predicting air pollution.In a case study,the relationship between three air pollutant concexntrations and four impact factors in thirty-one grid squares in Xiamen was analyzed. Cluster threes concerning quantitative correlation between air quality and the distribution of relevant impact factors were derived from significant test, discriminant analysis arid cluster analysis to realize the aim of predicting air quality.