Investigation of the Environmental Quality of Watershed Prediction System Based on an Artificial Intelligence Algorithm

• Published in: Water, air, and soil pollution Vol. 236; no. 2; p. 142
• Main Authors: Liu, Zian, Ren, Lingwei, Ke, Zhonghao, Jin, Xizheng, Rui, Shuya, Pan, Hua, Ye, Zhiping
• Format: Journal Article
• Language: English
• Published: Cham Springer International Publishing 01.02.2025; Springer Nature B.V
• Subjects: air; Algorithms; Artificial intelligence; Artificial neural networks; Atmospheric Protection/Air Quality Control/Air Pollution; Back propagation; Back propagation networks; Catchment areas; Chemical oxygen demand; China; Climate Change/Climate Change Impacts; Correlation coefficient; Correlation coefficients; Current prediction; Earth and Environmental Science; Environment; Environmental monitoring; Environmental quality; Hydrogeology; Neural networks; Oxygen; Oxygen requirement; Pollution monitoring; prediction; Prediction models; Predictions; Propagation; Root-mean-square errors; soil; Soil Science & Conservation; Training; water; Water management; Water monitoring; Water pollution; Water quality; Water Quality/Water Pollution; Watersheds; China; Back-propagation neural network; NH; Water quality prediction; Chemical oxygen demand
• ISSN: 0049-6979; 1573-2932
• DOI: 10.1007/s11270-025-07778-6

Monitoring and predicting the environmental quality of watersheds is essential for understanding and managing water pollution. Current prediction models often suffer from limitations, including the need for excessive information, complex architectures, and extensive computational resources. To address these challenges, this paper proposes a water pollution prediction system using artificial neural network trained by the back-propagation algorithm with a 2–6-2 structure. The model was developed using chemical oxygen demand and NH₄⁺ concentration data collected from the catchment areas of Kaihua and Anji counties in Zhejiang Province between November 2020 and October 2021. The average relative errors of the neural network training for chemical oxygen demand and NH 4 + were -4.59% and -2.65%, the correlation coefficients were 100% and 98%, and the root-mean-square errors were 7.83% and 0.14%, which confirmed the effectiveness of the back-propagation neural network training. The average relative errors between the predicted and observed values of chemical oxygen demand and NH 4 + by the neural network were -4.46% and 2.34%, respectively, with correlation coefficients of 100% and 88%, coefficient of determination of 0.94, and root-mean-square errors of 7.72% and 0.11%, which indicated that the predicted values of the back-propagation neural network on the quality of the water were highly significant correlated with the measured values. This study highlights the potential of artificial neural network models to offer efficient, accurate, and computationally streamlined solutions for water pollution monitoring.