Groundwater quality assessment using an adaptive GT-FAHP-entropy-weighted index: The Bafgh aquifer, Iran

Groundwater quality degradation in arid and semi-arid regions requires assessment approaches capable of identifying critical parameter exceedances while minimizing the masking effects commonly observed in aggregated water quality indices. Conventional entropy-weighted water quality indices (EWQI) may underestimate groundwater deterioration because they primarily rely on data variability and may not adequately represent persistent guideline exceedances. This study develops an adaptive groundwater-quality assessment framework integrating fuzzy analytic hierarchy process (FAHP), game theory, and exceedance-frequency-based weighting. The conventional EWQI was used as a benchmark, while two improved indices, including GT-FAHP-EWQI and GT-FAHP-AF-EWQI, were developed for the Bafgh aquifer, Iran. The proposed framework integrates subjective and objective weighting through FAHP, Shannon entropy, and game theory, while dynamically adjusting parameter weights according to exceedance frequency. The performance of the indices was evaluated using eclipsing, ambiguity, sensitivity, specificity, balanced accuracy, ROC analysis, and robustness assessment. The results showed that GT-FAHP-AF-EWQI improved groundwater-quality classification compared with EWQI. The optimal adaptive coefficient (λ = 1.75) reduced the eclipsing rate from 27.23% to 7.33%, increased sensitivity to 92.67%, and achieved a balanced accuracy of 96.34%. Leave-one-observation-out analysis confirmed the robustness of the framework (Spearman correlation coefficient = 0.99). Hydrochemical analysis indicated that salinity-related processes, particularly EC and TDS, were dominant factors controlling groundwater deterioration. Health risk assessment revealed that NO 3 - and F - contributed to non-carcinogenic risks, with children showing higher vulnerability. The proposed framework provides a practical tool for groundwater management and monitoring prioritization. Future studies should evaluate its transferability using multi-year datasets, denser monitoring networks, and additional contaminants such as heavy metals and microbiological indicators.