Predicting Abiotic TCE Transformation Rate Constants—A Bayesian Hierarchical Approach

Abstract

Fe(II) minerals can mediate the abiotic reduction of trichloroethylene (TCE), a widespread groundwater contaminant. If reaction rates are sufficiently fast for natural attenuation, the process holds potential for mitigating TCE pollution in groundwater. To assess the variability of abiotic TCE reduction rate constants, we collected pseudo-first-order rate constants for natural sediments and rocks from the literature, as well as intrinsic (surface-area-normalized) rate constants of individual minerals. Using a Bayesian hierarchical modeling approach, we were able to differentiate the contributions of natural variability and experimental error to the total variance. Applying the model, we also predicted rate constants at new sites, revealing a considerable uncertainty of several orders of magnitude. We investigated whether incorporating additional information about sediment composition could reduce this uncertainty. We tested two sets of predictors: reactive mineral content (measured by X-ray diffraction) combined with surface areas and intrinsic rate constants, or the extractable Fe(II) content. Knowledge of the mineral composition only marginally reduced the uncertainty of predicted rate constants. We attribute the low information gain to the inability to measure the (reactive) surface areas of individual minerals in sediments or rocks, which are subject to environmental factors like aqueous geochemistry and redox potential. In contrast, knowing the Fe(II) content reduced the uncertainty about the first-order rate constant by nearly two orders of magnitude, because the relationship between Fe(II) content and rate constants is approximately log–log-linear. We demonstrate how our approach provides estimates for the range of cleanup times for a simple example of diffusion-controlled transport in a contaminated aquitard.