Quantitative Estimates of the Uncertainty in the Evaluation of Ground Water Remediation Schemes

Abstract

Due to the natural heterogeneity of subsurface formations and to the often limited number of data, the definition of the physical and chemical parameters that govern flow and contaminant transport processes is normally characterized by a high level of uncertainty. Uncertainty in the relevant input parameters leads to uncertainty in modeling predictions, which may influence the selection of the optimal remedial alternative. This paper describes a statistical procedure for quantifying the uncertainty in modeling predictions made in support of the evaluation of ground water remediation designs. The hydraulic conductivity field is defined as a random spatial variable whose statistical structure is inferred from the available hydraulic conductivity data. Monte Carlo simulations are used to generate equiprobable multiple realizations of the hydraulic conductivity conditioned on the available hydraulic conductivity measurements. Using Bayes’theorem, probability-based weights for each of the hydraulic conductivity realizations are estimated from “soft” data consisting of historical concentration data and contamination source information. The probability weights are expressed in terms of the simulated concentration distributions and the observed concentration data. Ground water flow and contaminant transport simulation models are then used to probabilistically evaluate and optimize ground water remediation schemes based on the statistical analysis of key model predictions, such as the probability of a particular part of the aquifer being within the capture zone of an extraction well, or the recovered mass of contaminants as a function of time.