Evaluation and estimation of surface water quality in an arid region based on EEM-PARAFAC and 3D fluorescence spectral index: A case study of the Ebinur Lake Watershed, China

Development of tools for real-time monitoring of water quality is crucial for management of stream water pollution. In this study, fluorescence excitation-emission matrix (EEM) with parallel factor analysis (PARAFAC) were used to estimate five-day biochemical oxygen demand (BOD5), chemical oxygen demand (COD), suspended sediment (SS), total nitrogen (TN), pH, total phosphorus (TP), ammonia nitrogen (NH3+-N), turbidity (NTU), and dissolved oxygen (DO) concentrations in inland freshwater streams. The study area was located in Ebinur Lake Watershed, Xinjiang Uygur Autonomous Region, China. Our results showed that the four fluorescence components were successfully extrapolated by the PARAFAC factor analysis modeling from the fluorescence EEM data including microbial humic-like (C1), terrestrial humic-like organic substances (C2, C4), and protein-like organic substances (C3). The four PARAFAC components were selected as the water quality monitoring index of river water samples. Three-dimensional fluorescence index analysis showed that terrigenous organic pollution was the main organic pollution type in the watershed. This indicates that the watershed is subject to human disturbance, given the large variation of organic pollution in the water body. There was a significant correlation between the three-dimensional fluorescence index, the fluorescence components, and surface water quality index. In addition, the application of three-dimensional fluorescence index and fluorescence component to establish water quality index produced a model fitting coefficient greater than 0.5, which demonstrated that the accuracy of the model was in line with monitoring requirements in practice. This implies that the use of three-dimensional fluorescence index and three-dimensional fluorescence component to monitor Ebinur Lake Watershed water quality has a great potential.