The Estimation of Chemical Oxygen Demand of Erhai Lake Basin and Its Links with DOM Fluorescent Components Using Machine Learning

• Published in: Water (Basel) Vol. 13; no. 24; p. 3629
• Main Authors: Zhao, Yuquan, Shen, Jian, Feng, Jimeng, Sun, Zhitong, Sun, Tianyang, Liu, Decai, Xi, Mansong, Li, Rui, Wang, Xinze
• Format: Journal Article
• Language: English
• Published: Basel MDPI AG 01.12.2021
• Subjects: Algorithms; Chemical oxygen demand; COD; Dissolved organic matter; Dissolved oxygen; DOM; Drinking water; EEM–PARAFAC; Emission analysis; Factor analysis; Fluorescence; Hydrology; Lake basins; Lakes; Learning algorithms; Machine learning; machine learning models; Methods; Nitrogen; Parameter estimation; Pollution; Pollution monitoring; Potassium; Quality management; random forest; Real time; Rivers; Water inflow; Water management; Water quality; water quality estimation; Water quality management; Water temperature; Watersheds; United States > US; China; Yunnan China
• ISSN: 2073-4441; 2073-4441
• DOI: 10.3390/w13243629

Water quality estimation tools based on real-time monitoring are essential for the effective management of organic pollution in watersheds. This study aims to monitor changes in the levels of chemical oxygen demand (COD, CODMn) and dissolved organic matter (DOM) in Erhai Lake Basin, exploring their relationships and the ability of DOM to estimate COD and CODMn. Excitation emission matrix–parallel factor analysis (EEM–PARAFAC) of DOM identified protein-like component (C1) and humic-like components (C2, C3, C4). Combined with random forest (RF), maximum fluorescence intensity (Fmax) values of components were selected as estimation parameters to establish models. Results proved that the COD of rivers was more sensitive to the reduction in C1 and C2, while CODMn was more sensitive to C4. The DOM of Erhai Lake thrived by internal sources, and the relationship between COD, CODMn, and DOM of Erhai Lake was more complicated than rivers (inflow rivers of Erhai Lake). Models for rivers achieved good estimations, and by adding dissolved oxygen and water temperature, the estimation ability of COD models for Erhai Lake was significantly improved. This study demonstrates that DOM-based machine learning can be used as an alternative tool for real-time monitoring of organic pollution and deepening the understanding of the relationship between COD, CODMn, and DOM, and provide a scientific basis for water quality management.