Assessment of temporal and spatial variations in water quality using multivariate statistical methods: a case study of the Xin’anjiang River, China

This study evaluated the temporal and spatial variations of water quality data sets for the Xin’anjiang River through the use of multivariate statistical techniques, including cluster analysis (CA), discriminant analysis (DA), correlation analysis, and principal component analysis (PCA). The water samples, measured by ten parameters, were collected every month for three years (2008–2010) from eight sampling stations located along the river. The hierarchical CA classified the 12 months into three periods (First, Second and Third Period) and the eight sampling sites into three groups (Groups 1, 2 and 3) based on seasonal differences and various pollution levels caused by physicochemical properties and anthropogenic activities. DA identified three significant parameters (temperature, pH and E.coli) to distinguish temporal groups with close to 76% correct assignment. The DA also discovered five parameters (temperature, electricity conductivity, total nitrogen, chemical oxygen demand and total phosphorus) for spatial variation analysis, with 80.56% correct assignment. The non-parametric correlation coefficient (Spearman R) explained the relationship between the water quality parameters and the basin characteristics, and the GIS made the results visual and direct. The PCA identified four PCs for Groups 1 and 2, and three PCs for Group 3. These PCs captured 68.94%, 67.48% and 70.35% of the total variance of Groups 1, 2 and 3, respectively. Although natural pollution affects the Xin’anjiang River, the main sources of pollution included agricultural activities, industrial waste, and domestic wastewater.