Reducing Uncertainty in Site Characterization Using Bayes Monte Carlo Methods

Abstract

A Bayesian uncertainty analysis approach is developed as a tool for assessing and reducing uncertainty in ground-water flow and chemical transport predictions. The method is illustrated for a site contaminated with chlorinated hydrocarbons. Uncertainty in source characterization, in chemical transport parameters, and in the assumed hydrogeologic structure was evaluated using engineering judgment and updated using observed field data. The updating approach using observed hydraulic head data was able to differentiate between reasonable and unreasonable hydraulic conductivity fields but could not differentiate between alternative conceptual models for the geological structure of the subsurface at the site. Updating using observed chemical concentration data reduced the uncertainty in most parameters and reduced uncertainty in alternative conceptual models describing the geological structure at the site, source locations, and the chemicals released at these sources. Thirty-year transport projections for no-action and source containment scenarios demonstrate a typical application of the methods.