Linear and Monte Carlo uncertainty analysis for subsurface contaminant transport simulation

Abstract

Transport simulation can be used to predict subsurface contaminant distributions and the effectiveness of site remediation technologies. Because of the uncertainty inherent in subsurface transport prediction, an integral part of predictive modeling is uncertainty analysis. We have investigated the uncertainty of simulated trichloroethylene (TCE) concentrations due to parameter uncertainty and variation in conceptual model. The transport simulation is performed with T2VOC, a three‐dimensional integral finite difference code for three‐phase (gas, aqueous, non‐aqueous phase liquid), three‐component (air, water, volatile organic compound), nonisothermal subsurface flow. Our modeling is based on an actual site and considers a three‐dimensional system with a thick vadose zone (25 m) into which 35,000 kg of TCE was disposed of in surface trenches over 10 years. Subsurface transport involves TCE moving in the vadose and saturated zones from the source trench toward a nearby residence and includes the processes of advection, diffusion, and adsorption over extended distances and timescales. Uncertainties in the calculated concentrations due to variance in the major transport parameters are quantified using the inverse modeling code ITOUGH2. ITOUGH2 calculates uncertainty in the T2VOC simulation by both first‐order, second‐moment (FOSM) and Monte Carlo analyses. Significant uncertainty in simulated TCE concentrations at a site of potential human exposure is observed owing to uncertainty in permeability, porosity, diffusivity, chemical solubility, and adsorption within a single conceptual model. For the case study considered the linear FOSM analysis generally captures the uncertainty range calculated by the more accurate Monte Carlo method. Calculations also show that significant output uncertainty is introduced by conceptual model variation. Because human exposure and health risk depend strongly on contaminant concentrations, risk assessment and remediation selection based on transport simulation is meaningful only if the analysis includes quantitative estimates of transport simulation uncertainty.