Dense nonaqueous phase liquid (DNAPL)-contaminated sites pose a significant challenge at many Department of Defense facilities. There is a pressing need to realistically assess and reduce overall management costs by jointly considering site characterization, remediation, and reliability of reaching remediation objectives at such sites. This will require an integrated methodology that includes sound but practical process modeling, data analysis, evaluation of uncertainty, and cost analysis.

The objective of this project was to develop a practical tool for optimizing the design and operation of groundwater remediation systems that explicitly considers uncertainty in site and remediation system characteristics, performance and cost model limitations, and measurement uncertainties that affect predictions of remediation performance and cost. The project was specifically focused on chlorinated solvent-contaminated sites with DNAPL sources.

The method is based on a semi-analytical mathematical model to simulate DNAPL source depletion and dissolved phase transport in response to natural and engineered conditions. The performance model is coupled with cost functions for thermal source zone treatment and enhanced bioremediation. Compliance criteria are defined by statistical rules. The performance model also is coupled with an inverse solution to estimate model parameters, parameter covariances, and residual prediction error. A stochastic cost optimization (SCO) algorithm is used to determine values for design variables that minimize expected net present value cost over Monte Carlo realizations. The method was implemented in SCOToolkit software and applied to two well-characterized sites where different remedial technologies were used to evaluate its ability to reduce costs and improve remedial designs.

Stochastic cost optimization of design and monitoring variables was found to reduce expected costs by 50% or more relative to conventional design methods and to substantially increase the probability of meeting compliance targets. Although additional field applications to demonstrate the method are needed, along with development of a user friendly interface, the method was shown to be highly effective for two field test sites.

At the Fort Lewis East Gate Disposal Yard (EGDY), optimization of thermal source treatment indicated a need for a much larger treatment area than was actually employed to avoid a high failure probability associated with source delineation uncertainty based on available source characterization data. Source treatment may be cost effective if additional characterization were undertaken to reduce source zone uncertainty. The method also was used to optimize source and plume bioremediation at the EGDY site using whey injection without additional source reduction. The results indicated that this remedial strategy should achieve the maximum contaminant limit by 2110, with a 93% probability of success when using relatively low whey injection rates.

At Dover Air Force Base Area 5, the method was applied to optimize enhanced bioremediation. The results suggested that electron donor injection may be needed indefinitely in the future to meet remediation criteria, as a result of the uncertainty associated with the source mass estimates and the difficulties in characterizing and treating source zones without disrupting base operations. Optimization analyses to minimize long-term operating costs indicated compliance criteria could be met using only five of the current ten emulsified vegetable oil injection galleries, with operating costs approximately half of current costs. Recalibration and optimization after 50 years, using additional data from this period, was projected to further reduce operating costs.

The results indicate that the SCOToolkit has the potential to significantly improve remediation performance and reduce costs. Use of the method also can identify the most critical data gaps and the uncertainties that will most affect costs and performance projections.