Lagoon water quality monitoring based on digital image analysis and machine learning estimators

• Published in: Water research (Oxford) 2020-04, Vol.172, p.115471-115471, Article 115471
• Main Authors: Li, Yuanhong, Wang, Xiao, Zhao, Zuoxi, Han, Sunghwa, Liu, Zong
• Format: Article
• Language: English
• Subjects: Animals; Environmental Monitoring; Escherichia coli; Image processing; Lagoon water monitoring; Linear regression; Machine Learning; Spectrum analysis; Waste Water; Water; Water Quality
• ISSN: 0043-1354; 1879-2448
• DOI: 10.1016/j.watres.2020.115471

Lagoon has been widely used to treat animal wastewater. However, because lagoon effluent often fluctuates in water quality, land application of the effluent may pose a risk to the environment and/or public health. It is necessary to monitor the quality of lagoon water to reduce the risk of its land application. This paper proposes an innovative monitoring method for animal wastewater in lagoons. We implemented spectral processing techniques to analyze the reflectivity of wastewater samples from lagoons, and applied machine learning methods to estimate the water quality parameters of the effluents, including the levels of nitrogen, phosphorus, bacteria (total coliform and E. Coli), and total solids. This study found significant correlations between the spectral rate of emission and above water quality parameters. We used machine learning to train three types of estimators, normal equation linear regression (LR), stochastic gradient descent (SGD), and Ridge regression to quantify these relations. The model performance was evaluated by weight coefficient, function intercept, and mean squared error (MSE). The model showed that TS level and the blue band of spectral reflectance of samples have a relatively good linear relationship, and the MSE of prediction set and decision coefficient were 0.57 and 0.98, respectively. For bacteria level, the MSE of prediction set was 0.63, and coefficient R2 was 0.96. The results from this study could provide a versatile method for remote sensing of animal waste water. •Linear model between water quality parameters and wastewater spectral reflectance values.•The ways to optimize machine learning model performance on spectral data training.•An image-based technology for monitoring performance of animal wastewater lagoon.•Data reference for spectroscopy applications in agricultural wastewater treatment.