A Comparative Analysis of Groundwater Quality and ItsPrediction Using Machine Learning Techniques

Groundwater is a key water source; even small variations in its quality may result in ‎detrimental impacts, necessitating its regular monitoring for effective water management. This study emphasized monitoring and assessing groundwater from Mudirajupalem, Madalvarigudem, and the Lingayas Institute of Management and Technology, Vijayawada campus, Andhra Pradesh, India. The groundwater was analyzed for quantifying acidity, ‎alkalinity, pH, total dissolved solids (TDS), and total hardness according to standard methods. ‎All the groundwater sources showed high alkalinity values, whereas Mudirajupalem ‎groundwater exhibited high TDS (962±20 mg/L) and pH (9.2±0.3) values due to proximity to ‎farming fields. The obtained water quality index (WQI) values (358.34±15.2, 59.29±4.8, and ‎‎20.32±3.1 for the three sites, respectively) indicate that Mudirajupalem groundwater is unfit ‎for public intake. This study demonstrated promising results in forecasting water quality ‎parameters using a hierarchical reconciliation algorithm and predicted the WQI using gradient ‎boosted tree (GBT), random forest (RF), and decision tree (DT) techniques. The predicted ‎WQI values closely matched the experimental results, confirming that Mudirajupalem ‎groundwater is not fit for drinking. The GBT demonstrated superior performance (R²=0.95-‎‎0.98) compared with RF (R²=0.89-0.93) and DT (R²=0.84-0.87) for the selected study area ‎, and this study demonstrates that the application of advanced machine learning enables a proactive ‎approach for better water quality management to address the future water needs‎.