Artificial Intelligence and Explainable AI for Groundwater Quality Assessment: A Comparative Review

Groundwater quality assessment is critical for determining the suitability of groundwater resources, especially in areas where groundwater resources are being affected by urbanization, industrialization, and other sources of contamination. Traditional evaluation methods relying on physicochemical parameters, statistical analysis, and Water Quality Index (WQI) can be helpful but may lack the power to reflect the complex interactions in large and heterogeneous groundwater data sets. In view of this, groundwater-quality assessment and/or prediction tasks have increasingly been studied using machine learning (ML), deep learning (DL), and Explainable Artificial Intelligence (XAI) approaches. This paper compares and contrasts the AI and XAI methods used for groundwater and related water-quality assessment. A set of studies focused on the dataset characteristics, groundwater quality parameters, modelling approaches, integration of the WQI, explainability, and implementation were analysed. Both the comparison and the results suggest that ML models and DL models can offer robust prediction capabilities, and XAI techniques like SHAP, LIME, and feature-importance methods can enhance the interpretability of complex models. The review additionally pinpoints the low level of integration of predictive modelling, explainability, environmental-health assessment, and actionable decision support. Geographically diverse data sources, multi-parameter modelling, and integrated systems need to be focused on in future studies.