Climate change and human activity are posing an increasing danger to water quality. Surface water in Algeria, especially in semi-arid areas, is under increasing pressure from industrial and urban discharges, and there is still a dearth of scientific information regarding its quality and appropriateness for irrigation. This study’s goal was to assess the physico chemical properties and surface water irrigation quality at nine locations along the Seybouse River in northeastern Algeria, thereby addressing an important lack of information in an area with little monitoring. The current study looked at nine sampling sites to assess the river Water Quality Index (WQI), sodium absorption rate (SAR), soluble sodium percentage (SSP), and Kelly Ratio (KR). Artificial Neural Network (ANN) and Response Surface Methodology (RSM) inspired machine learning (ML) approaches were used to enhance the evaluations utilizing Principal Component Analysis (PCA) data, representing 82.08% of the entire collection of data. The study’s findings demonstrate that the ANN forecasting of WQI, SSP, SAR, and KR is more accurate than the RSM models. The ANN models achieved superior predictive performance with R2 values of 98.20%, 98.39%, 99.85%, and 97.24% and remarkably low Mean Squared Errors (MSE) of 0.0451, 0.0231, 0.0241, and 0.0167, respectively. Furthermore, the low Root Mean Square Error (RMSE) values for ANN (ranging from 0.0211 to 0.0521) confirm its high precision compared to RSM. In comparison, RSM models yielded R2 values of 97.10%, 96.37%, 97.06%, and 97.20% with slightly higher error margins. According to this result, ANN and RSM-BBD forecasting methods provide an accurate alternative in forecasting. Furthermore, ML improves the monitoring of water quality and gives water managers the ability to control water quality indicators. Developers may find this research useful if they require accurate water quality data to assist them create plans for managing water supplies.
Kherouf et al. (Thu,) studied this question.