Open Collections

Abstract

The accuracy and reliability of water quality models and environmental risk assessments are often hindered by missing or incomplete water quality data (historical and/or predictive). Specifically, data gaps can compromise the ability of water quality models to accurately represent complex geochemical interactions and predict future water quality trends while also undermining the confidence in risk assessments that require detailed contaminant and environmental characterization. Conventional approaches to infilling or estimating water quality data records, such as averaging, percentage-based, and interpolation, often fail to accurately represent the complex geochemical interactions that govern water quality dynamics. To address these limitations, this study proposes a four-stage, data-driven approach for infilling and estimating water quality data. First, a high-dimensional database is constructed, integrating water quality and geospatial data. Second, water quality parameters are identified as independent variables for statistical analysis, while machine learning is used to cluster data with similar characteristics (e.g., geochemical). Thirdly, a cluster-specific regression analysis is applied to capture localized interactions. Finally, the simulated data were used as input for the regression models to resolve the data gaps. This data-driven framework effectively addresses water quality data gaps, reducing model uncertainty and enhancing the reliability of both water quality model development and environmental risk assessments. A critical component of water quality modelling and environmental risk assessments is the accurate quantification of environmental toxicity-modifying factors (ETMFs), including alkalinity, pH, and hardness. ETMFs are often used for calculating concentration limits and contaminant benchmarks, and insufficient data can increase uncertainties in environmental risk assessments. Therefore, a key focus of this research is to extend the data-driven approach to specifically address data gaps in ETMF datasets.