Novel groundwater quality index (GWQI) model: A reliable approach for the assessment of groundwater

The groundwater is crucial for drinking, irrigation, river flows, lake levels, and ecosystem health. Regular monitoring and reporting of groundwater are required for sustainable aquifer management. The goal of this study is to assess the reliability of the newly developed Groundwater Quality Index (GWQI) model using water quality (WQ) data from 58 monitoring sites in three Irish domains. The study utilized the novel GWQI model which incorporates trace/heavy metals and a broad classification scheme to compute and nuanced interpretation of the GWQI scores. The model's performance, encompassing uncertainty and sensitivity, was evaluated using nine ML/AI models. Most monitoring sites demonstrated a ‘Fair’ to ‘Good’ WQ rating, while the WFD suggested EPA approach (e.g., ‘One Out, All Out’—OOAO) rated it from ‘Good’ to ‘Poor’. This suggests only a few indicators breached WHO (2022) and EPA (2014) guidelines. In addition, the Gradient Boosting model with the OPTUNA optimizer outperformed in predicting groundwater quality with high accuracy (RMSE‾ = 0.73, MSE‾ = 0.69, MAE‾ = 0.41, PABE‾ = 0.41), efficiency (MEF‾ = 0.63) and low prediction error (PBIAS‾ = 0.03; PREI‾ =0.02); whereas the model high sensitivity (R2‾≥ 0.91) indicated the strong relationship between model input and output with less uncertainty (<2 %). Compared to other exiting approaches, the GWQI model provides the accurate rating of groundwater quality, whereas the EPA approach misinterpreted the WQ in several monitoring sites in domain 2 (D02, D03, D07, D08, D10, D11, D12, and D13), and domain 3 (C01, C02, C05, C11, C15, and C17), respectively, without breaching number of indicators by following guidelines. Therefore, relatively the GWQI model shows outstanding performance to correct WQ classification in terms of reliability and reducing uncertainty in the groundwater quality assessment. However, the result of the research revealed that the GWQI model could be effective to classify groundwater quality more accurately. Since this study utilized single time-span dataset across various domains, it is suggested for future research to consider extended spatio-temporal resolution's of the domain including various anthropogenic stressors under different climate attributes. The research findings could be helpful to support the decision makers/policy makers/national and international stakeholders in order to promote sustainable groundwater quality management with less assessment errors using the GWQI model.