ASSESSMENT OF URBAN LANDSCAPE WATER USING MULTIVARIATE STATISTICAL TECHNIQUES

Urban landscape water has become serious social environmental problems in recent years. This study presents the application of multivariate statistical techniques, namely, principal component and factor analysis (PCA/FA), multiple regression analysis (MRA), cluster analysis (CA) and discriminant analysis (DA), to classify and evaluate urban landscape water pollution levels and possible pollution sources in order to provide a scientific basis of prevention and remediation of urban water pollution in Shanghai. PCA evolved three principle components explaining about 94.365% of the total variance. FA was used by principal components (PCs) extracted from the data to get three factors which are considered to be a nutrition factor, an organic pollution factor and a phytoplankton factor in urban landscape water, respectively. Urban landscape water is mainly under the control of phosphorus and nitrogen pollution. PCA/FA was supported with multiple regression analysis to determine the most important parameter in each factor. The cluster analysis and discriminant analysis show that the studied urban landscape water can be grouped into three clusters i.e., the highly pollution level (HPL), medium pollution level (MPL) and less pollution level (LPL). The recognition capacities of the two discriminant functions were 93.3% and 6.7%, respectively. The multivariate statistical techniques are useful tools for evaluation and classification of spatial pollution differences in urban landscape water and may be applicable to assessment of other water bodies.