Under the framework of sustainable development, ensuring the quality of water for drinking and irrigation purposes presents a complex challenge, as it is influenced by the interrelated effects of multiple surface water management parameters. Therefore, the present study aims to improve the assessment of surface water quality and to evaluate the suitability of surface water networks for domestic and industrial use in Rourkela and its surrounding areas within Rourkela City, Odisha. This region exhibits a high dependence on surface water resources, which are limited in availability, underscoring the need for comprehensive quality evaluation. The WQI (water quality index) of the study region was trained and tested using various ML (machine learning) algorithms employing Python application software. These ML models are: Support Vector Machine (SVM), Random Forest (RF), Extra Trees (ET), Multiple Linear Regression (MLR), Decision Tree (DT), Logistic Regression, K-Nearest Neighbors (KNN), Gradient Boosting (GB), Naive Bayes (NB), and AdaBoost (AB). To achieve this objective, a total of 12 water samples were collected from ten designated monitoring sites over a two-year period (2023–2025) during the pre-monsoon season. The samples were analyzed for key physicochemical parameters, including pH, electrical conductivity (EC), total dissolved solids (TDS), alkalinity, total hardness (TH), copper (Cu2⁺), zinc (Zn2⁺), sodium (Na⁺), potassium (K⁺), lead (Pb2⁺), phosphate (PO43⁻), and iron (Fe2⁺). The Water Quality Index (WQI) approach was applied to assess overall suitability, while Pearson Correlation analysis helped to identify relationships among parameters. Radar distribution maps were generated using Python software to visualize the variability of water quality across the study area. Observed physicochemical results depicts that, localized exceedances in EC, TH, TDS, and trace metals concentrations (Cu2+, Zn2+, Pb2+, Fe2+) highlight potential risks, particularly in areas with higher drinking and agricultural activity. The WA (weighted–arithmetic) WQI results spanned between 40 and 361, indicating that most samples fall within the Poor/ Very Poor/ Unsuitable” WQI classes, suggesting that surface water in the area is generally unsuitable for drinking and household purposes. Pearson’s coefficient visualized that Alkalinity, TH, and K+, are the most influential parameters controlling WQI, while parameters like EC, TDS, Na+, and pH, have a less minimal impact. MLR (Multiple Linear Regression) analysis reveals that, although all examined parameters are statistically significant, sodium (Na⁺), iron (Fe2⁺), potassium (K⁺), and alkalinity exhibit the most substantial influence on surface water quality. This is indicated by their relatively lower standardized beta coefficients, which are closely linked to the observed classification of water as moderately hard in the study area. The statistical ML models' performance was assessed using RMSE, MSE, NSE, and R2. The class-wise ML models' performance was assessed using F1 Score, Precision, Accuracy, Recall, and AUC. The results of this study showed that the highest values of R2, MSE, RMSE, and NSE during the testing and training of models were 0.978, 5.65, 6.85, and 0.98, and 0.988, 4.25, 4.85, and 0.965, respectively. The highest results of class-wise model performance evaluation of F1-Score, accuracy, recall, and precision were 0.969, 0.999, 0.978, and 0.985 respectively. The highest value of the ROC-AUC curve in this study was 0.93 during GB and RF; this indicates that these models were the best for forecasting the WQI of this study. Overall, the findings indicate that surface water in the Rourkela City, is chemically suitable for use at some sampling locations, but continued monitoring is necessary to address localized contamination and to ensure long-term water security for the growing population.
Abhijeet Das (Mon,) studied this question.