Assessing local uncertainty of intrinsic groundwater vulnerability using nonparametric geostatistics: A probabilistic approach

Abstract

Uncertainty is inherent in the evaluation of groundwater vulnerability, and modeling uncertainty is crucial for decision-making related to groundwater resources management. This study investigated local uncertainty of the DRASTIC model established for the Pingtung Plain, Taiwan, using indicator kriging and nonparametric Monte Carlo simulation. Local uncertainty of classifying the ratings of DRASTIC parameters was probabilistically measured using one minus the largest conditional probability and local entropy (LE). Sensitivity analysis was employed to analyze an integrating uncertainty derived from weighted averages of the probabilistic measures. Local uncertainty of variations of DRASTIC vulnerability indices (DVIs) was stochastically measured using interquartile range (IQR) and conditional variance (CV). Spearman’s rank correlations were used to analyze local uncertainty of key DRASTIC parameters influencing that of DVIs. The results revealed that the measures of local uncertainty were considerably consistent with one minus the largest conditional probability and LE for classifying the ratings of DRASTIC parameters and IQR and CV for rating variations of DVIs, demonstrating a reliable result of uncertainty assessment. The local uncertainty of the groundwater depth (i.e., D parameter) considerably influenced that of the DRASTIC model. To reduce uncertainty, increasing wells for monitoring groundwater levels should be considered in determined groundwater protection zones with high DVIs and high uncertainty.