Machine learning-based prediction of seasonal groundwater quality for urbanized parts of Melur (Tamil Nadu), India

• Groundwater quality assessment using WQI. • Machine learning algorithms for predicting groundwater quality. • Prediction performances of RF, XGBoost, DT and ANN models. • ANN based quality prediction performing better than other ML models. Groundwater contamination has become a critical global challenge due to population growth, industrialization and climate variability, while in India it is further intensified by urbanization, over-extraction and heavy metal pollution. Accurate prediction of water quality is critical for the sustainable groundwater management. This research used Artificial Intelligence (AI)- derived machine learning (ML) models such as Artificial Neural Network (ANN), Decision Tree (DT), Random Forest (RF) and eXtreme Gradient Boosting (XGBoost) to predict groundwater quality (bore wells and open wells) in two seasons in the urbanized area of Melur in the Tamil Nadu state of India. Hydrogeochemical parameters, heavy metal contents and weight-based groundwater quality index (WQI) were used to develop predictive models for both the seasons. ANN was the most effective among the prediction algorithms for the pre-monsoon, achieving the highest accuracy of (R² =0.95). Its performance, however, declined significantly (R² = 0.68) for the post-monsoon compared to XGBoost (R² = 0.87). This approach of using different machine learning models for accurate prediction of water quality in different seasons shows the robustness of the forecast for an efficient water resource management practice.