Application of Multivariate Statistical Techniques and Water Quality Index for the Assessment of Water Quality and Apportionment of Pollution Sources in the Yeongsan River, South Korea

This study assessed spatial and temporal variations of water quality to identify and quantify possible pollution sources affecting the Yeongsan River using multivariate statistical techniques (MSTs) and water quality index (WQI) values. A 15 year dataset of 11 water quality variables was used, covering 16 monitoring sites. The nutrient regime, organic matter, suspended solids, ionic contents, algal growth, and total coliform bacteria (TCB) were affected by the summer monsoon and the construction of weirs. Regression analysis showed that the algal growth was more highly regulated by total phosphorus (TP; R(2) = 0.37) than total nitrogen (TN, R(2) = 0.25) and TN/TP (R(2) = 0.01) ratios in the river after weir construction and indicated that the river is a P-limited system. After constructing the weirs, the mean TN/TP ratio in the river was about 40, meaning it is a P-limited system. Cluster analysis was used to classify the sampling sites into highly, moderately, and less polluted sites based on water quality features. Stepwise discriminant analysis showed that pH, dissolved oxygen (DO), TN, biological oxygen demand (BOD), chemical oxygen demand (COD), chlorophyll-a (CHL-a), and TCB are the spatially discriminating parameters, while pH, water temperature, DO, electrical conductivity, total suspended solids, and COD are the most significant for discriminating among the three seasons. The Pearson network analysis showed that nutrients flow with organic matter in the river, while CHL-a showed the highest correlation with COD (r = 0.85), followed by TP (r = 0.49) and TN (r = 0.49). Average WQI values ranged from 55 to 141, indicating poor to unsuitable water quality in the river. The Mann–Kendall test showed increasing trends in COD and CHL-a but decreasing trends for TP, TN, and BOD due to impoundment effects. The principal component analysis combined with factor analysis and positive matrix factorization (PMF) showed that two sewage treatment plants, agricultural activities, and livestock farming adversely impacted river water quality. The PMF model returned greater R(2) values for BOD (0.92), COD (0.87), TP (0.93), TN (0.91), CHL-a (0.93), and TCB (0.83), indicating reliable apportionment results. Our results suggest that MSTs and WQI can be effectively used for the simple interpretation of large-scale datasets to determine pollution sources and their spatiotemporal variations. The outcomes of our study may aid policymakers in managing the Yeongsan River.