Analysis of spatial and temporal water pollution patterns in Lake Dianchi using multivariate statistical methods

Abstract

Various multivariate statistical methods including cluster analysis (CA), discriminant analysis (DA), factor analysis (FA), and principal component analysis (PCA) were used to explain the spatial and temporal patterns of surface water pollution in Lake Dianchi. The dataset, obtained during the period 2003–2007 from the Kunming Environmental Monitoring Center, consisted of 12 variables surveyed monthly at eight sites. The CA grouped the 12 months into two groups, August–September and the remainder, and divided the lake into two regions based on their different physicochemical properties and pollution levels. The DA showed the best results for data reduction and pattern recognition in both temporal and spatial analysis. It calculated four parameters (TEMP, pH, COD_Mn, and Chl-a) to 85.4% correct assignment in the temporal analysis and three parameters (BOD, NH$_{4}^{+}$–N, and TN) to almost 71.7% correct assignment in spatial analysis of the two clusters. The FA/PCA applied to datasets of two special clusters of the lake calculated four factors for each region, capturing 72.5% and 62.5% of the total variance, respectively. Strong loadings included DO, BOD, TN, COD_Cr, COD_Mn, NH$_{4}^{+}$–N, TP, and EC. In addition, box–whisker plots and GIS further facilitated and supported the multivariate analysis results.