Predicting areal extent of groundwater contamination through geostatistical methods exploration in a data-limited rural basin

Abstract

The areal extent of regulatory standard exceedance for groundwater arsenic, fluoride, and nitrate was estimated and compared for multiple methods to evaluate groundwater contamination in a rural basin of New Mexico, USA. First, three geostatistical kriging techniques – Ordinary Kriging (OK), Universal Kriging (UK), and Empirical Bayesian Kriging (EBK) – were explored. Next, the exploration was extended to multivariate analysis (Co-OK, Co-UK, and EBK-regression) using groundwater pH as the other variable, compared to interpolation only considering the contaminant as a primary variable. Finally, non-parametric Indicator Kriging (IK) was utilized for probability mapping and assessing the likelihood of surpassing critical thresholds for contaminant concentrations. The contamination area changes with different kriging methods deployed, and the performances of kriging variants vary for the three chemical components. The models that best fit the empirical data were determined. The study examined the influence of the explanatory variable on the multivariate kriging prediction accuracy. The multivariate kriging approaches enhanced surface prediction accuracy and tended to cover the data limitation of the primary variable interpolations. The multivariate kriging improved spatial prediction of arsenic and fluoride by increasing the Empirical Coverage Probability (ECP) up to 0.05–0.25 from case to case; however, no improvement was observed in ECP for nitrate contamination distribution, as the explanatory variable pH was only weakly correlated with nitrate, suggesting that a stronger correlation between the primary and secondary variables can increase the probability of obtaining prediction improvements through multivariate kriging. This study developed a methodology for characterizing and identifying locations where groundwater had elevated concentrations of contaminants, threatening human health through drinking water exposure. By improving the spatial characterization of groundwater contaminants within the data-limited rural basin, policymakers planning improvements will support mitigation strategies to manage contamination sources of groundwater resources.