Abstract
Source apportionment of river water pollution is critical in water resource management and aquatic conservation. Comprehensive application of various GIS-based multivariate statistical methods was performed to analyze datasets (2009-2011) on water quality in the Liao River system (China). Cluster analysis (CA) classified the 12 months of the year into three groups (May-October, February-April and November-January) and the 66 sampling sites into three groups (groups A, B and C) based on similarities in water quality characteristics. Discriminant analysis (DA) determined that temperature, dissolved oxygen (DO), pH, chemical oxygen demand (COD(Mn))(,) 5-day biochemical oxygen demand (BOD₅), NH₄⁺-N, total phosphorus (TP) and volatile phenols were significant variables affecting temporal variations, with 81.2% correct assignments. Principal component analysis (PCA) and positive matrix factorization (PMF) identified eight potential pollution factors for each part of the data structure, explaining more than 61% of the total variance. Oxygen-consuming organics from cropland and woodland runoff were the main latent pollution factor for group A. For group B, the main pollutants were oxygen-consuming organics, oil, nutrients and fecal matter. For group C, the evaluated pollutants primarily included oxygen-consuming organics, oil and toxic organics.