Enhancing source apportionment of carbon, nitrogen, and phosphorus through integrating PMF and observed source profiles in a subtropical river.

Apportioning pollution sources under compound pollution conditions is challenging in river pollution source analysis. The positive matrix factorization (PMF) model is widely used to analyze river pollution sources. However, the identification of pollutants in this model relies primarily on the subjective experience of the researchers, leading to ineffective identification of different contaminants from similar sources. In this study, we propose a comprehensive deviation index (CDI) to quantitatively identify pollution source types based on the PMF and observed source profiles. Taking the subtropical Xizhijiang River Basin as a case study, we quantitatively identified the pollution sources and their contributions to dissolved organic carbon (DOC), total nitrogen (TN), and total phosphorus (TP) using observed water quality and pollution sources data. The results showed that the eight major pollutants in the study region exhibited significant positive correlations, indicating the similarity of pollutant sources in the watershed. The PMF model identified three primary pollution sources with coefficients of determination for observed versus predicted concentrations ranging from 0.60 to 0.98. The CDI unveiled that the watershed's three pollution sources were farmland, rural, and wastewater treatment plants (WTPs). Farmland emerges as the predominant contributor to DOC (68.04 %), TC (63.29 %), and TDP (44.51 %). Rural notably contributes to NH3-N, PO4 3-, TDP, and TN, with percentages of 86.37 %, 57.65 %, 41.40 %, and 30.45 %, respectively. WTPs significantly contribute to NO2 -, NO3 -, and TN, accounting for 71.81 %, 57.39 %, and 37.26 %, respectively. Incorporating source fingerprints into the PMF model, the CDI can accurately identify pollution sources, improve the interpretability of source identification, and mitigate uncertainty in the multiple-source unknown receptor model. These findings have immediate and practical implications for river ecosystem management and pollution control, providing a more effective method for identifying and addressing pollution sources.