Advancing groundwater vulnerability assessment to nitrate contamination: a comprehensive evaluation of index-based, statistical and machine learning approaches with sensitivity analysis

Jing Yang, Heng Dai, Honghua Liu, Ming Ye, Tian Jiao, Ze Liu, Tongju Xing, Jie Dong,
Advancing groundwater vulnerability assessment to nitrate contamination: a comprehensive evaluation of index-based, statistical, and machine learning approaches with sensitivity analysis,
Journal of Hydrology,
Volume 663, Part A,
2025,
134189,
ISSN 0022-1694,
https://doi.org/10.1016/j.jhydrol.2025.134189.
(https://www.sciencedirect.com/science/article/pii/S0022169425015276)
Abstract: Methods for assessing groundwater vulnerability vary widely, with distinct objectives, data requirements, and results across units, scales, and sensitivities. Despite their widespread use, the comparative performance of these methods in real-world case studies remains underexplored. This study provides the first comprehensive evaluation of nine methods across three groups: (i) index-based methods (DRASTIC and DRASTIC-L); (ii) statistical methods (weights of evidence, logistic regression, and fuzzy logic); and (iii) machine learning models (multi-layer perceptron, support vector regression, random forest regression, and adaptive boosting). Using the phreatic aquifer of Dagu River Basin as a case study, we created vulnerability maps for each method and conducted an in-depth comparative analysis including visual inspection of vulnerability maps, histograms of normalized vulnerability values, and cumulative area percentages. To assess their performances, Spearman’s rank correlation and the area under the receiver operating characteristic curve (AUC-ROC) were employed. We also explored how factor weights influence index-based methods, and examined the sensitivity of statistical and machine learning methods to measurement errors and incompleteness of nitrate data. Our findings show that machine learning methods exhibit the best predictive performance. Index-based methods are sensitive to factor weights, while statistical methods are robust against nitrate measurement errors. In contrast, machine learning methods are more sensitive to incompleteness of nitrate data. This study provides new insights into the practical application and comparison of these methods, offering guidance on their selection based on data quality and model sensitivity. The integration of sensitivity analysis offers a new perspective to groundwater vulnerability assessment, improving decision-making for groundwater management.
Keywords: Groundwater quality; DRASTIC; Risk assessment; Sensitivity analysis; Dagu River