A unified framework for assessing water quality indices under data and model weight uncertainty

We focus on the assessment of Water Quality Indices (WQIs) upon explicitly quantifying the contribution and uncertainty of each parameter involved in their formulation. Our strategy entails development of a global sensitivity framework that integrates uncertainties in (i) data associated with physiochemical parameters and (ii) their corresponding weight in the evaluation of WQIs across diverse water sources. The approach is exemplified through application to three diverse water systems, formed by lakes, springs, and groundwater wells across the large scale field setting associated with the province of Lecco (Italy). We rest on a probability-based index to quantify parameter-specific influence on the likelihood that WQI values exceed given regulatory thresholds, associated with quality status of a target water source. Our results suggest that major ions predominantly drive probability of exceedance at lower thresholds (corresponding to good water quality), whereas redox-sensitive metals and particulate-associated parameters gain importance at higher thresholds (corresponding to poor water quality). To further evaluate the effect of data uncertainty, we implement a moment-based Global Sensitivity Analysis. The latter enables us to highlight distinct system-specific patterns, including the observation that turbidity dominates in lake water, copper and magnesium in spring water, and magnesium and hardness in groundwater well water. The proposed framework providing critical insights into reliability and interpretation of WQI estimates under uncertainty upon establishing a rigorous basis for uncertainty-aware water quality assessment as a support to design of optimized, source-specific monitoring strategies.