Probabilistic Approach for Locating Organic Liquids in Subsurface Sediments
Project Description and Significance
The Safety and Risk Analysis Department and the Geoscience Assessment and Validation Department at Sandia are developing improved methods for locating dense nonaqueous phase liquids (DNAPLs) in subsurface sediments. Potential customers for this work could include chemical companies, other private industry, and federal agencies tasked with remediation efforts.
Chlorinated organic solvents such as TCE, PCE, TCA, and carbon tetrachloride have been used for many years as de-greasers in many industrial applications across the nation. Initial estimates by the Environmental Protection Agency (EPA) suggest that nearly 50% of all sites on Superfund's national priority list are plagued by subsurface contamination by DNAPLs. These toxic compounds have a low solubility in water and tend to travel through the subsurface as a separate organic phase. These compounds are also more dense than water; therefore they are able to sink deeply into aquifers. Although DNAPLs tend to move through the subsurface as a separate phase, they also slowly dissolve into ground water and contaminate aquifers. Thus, because of their low solubility in water, relatively small quantities of DNAPLs can contaminate extremely large volumes of ground water. Slow dissolution of DNAPLs into ground water can continue for decades. It is because of this slow dissolution process that we must either remove or contain the separate organic phase, the DNAPL, to successfully remediate aquifers.
Recovery of DNAPLs is no easy task for two reasons. First, DNAPLs can become immobilized in the subsurface. As DNAPLs migrate, capillary forces trap some portion, leaving behind an immobile residual saturation that is difficult to recover. A simple pump-and-treat technique is extremely inefficient in recovering this residual since it must rely on the slow dissolution of the DNAPL. Fortunately, promising techniques are being developed as improvements over traditional pump and treat. However, these techniques all require some knowledge about DNAPL location.
The second reason that DNAPLs are difficult to remove from the subsurface and the focus of our work is that DNAPLs can be difficult to locate. DNAPL migration through aquifers is controlled by the competition between buoyancy and capillary forces. Buoyancy forces cause DNAPLs to migrate downward, while capillary forces can redirect that migration. DNAPL migration is sensitive to textural heterogeneities because less capillary pressure is required for the DNAPL to penetrate into coarse-grained material than into fine-grained material. This interplay between capillarity and soil heterogeneities is a mixed blessing for the migration of DNAPLs through aquifers. On one hand, the presence of layered sediments helps to divert the DNAPL from its density-driven destination to the bottom of the aquifer. On the other hand, the presence of lenses and layering can cause DNAPLs to travel in unforeseen directions. Development of improved remediation technologies to clean up DNAPLs cannot be successful unless we also improve our ability to discover where the spilled DNAPLs reside in the subsurface. It is clear that the source of contamination, the DNAPL itself, must be removed in order to achieve aquifer remediation. However, if we can't find the DNAPLs, we can't clean them up. We have shown that probabilistic predictions of DNAPL location can be used to suggest the best sampling locations for delineating the DNAPL to the extent required by applicable remedial technologies.
Our strategy uses probabilistic predictions of DNAPL location to suggest the best sampling locations for delineating DNAPLs. The strategy involves combining multiple representations of possible DNAPL plumes to form a probability map of DNAPL location; this strategy reflects our degree of certainty about where DNAPLs might reside in the subsurface. These realizations are generated using fluid-flow models that capture the physics of DNAPL movement, while including the geologic features controlling DNAPL migration. In this way, our uncertainty about the distribution of geologic features and the disposal history at the site can be propagated through a fluid-flow model to reflect our degree of uncertainty in DNAPL location. After we have produced a probability map, a sampling scheme that applies the map would reduce the uncertainty in DNAPL location with the construction of as few wells as possible. We have shown the efficacy of our approach using a hypothetical problem as an example. In FY96, we will be field testing the approach using a DNAPL spill at DOE's Paducah Gaseous Diffusion Plant. This approach is a new application of the probabilistic risk-assessment methodology, used to quantitatively describe the degree of certainty associated with predicting the future performance of engineered or natural systems such as nuclear reactors or waste repositories. (For more information, see "Risk-Assessment Methodology for Low-Level Radioactive Waste Disposal" and "Compliance Assessment for Greater Confinement Disposal Compliance Assessment Program.")
Future Work
Assisted by partnerships with industry and academia, we believe we can lead the effort for solving this problem of national importance. And we believe we can solve this problem using existing tools within a risk-based framework. This proposed framework is entirely consistent with the decision framework we have developed for our EPA Superfund work (see fact sheet titled "Risk-Based Decision Making"). This approach is being embodied in an automated decision support system.
For further information, contact:
Stephen H. Conrad
Sandia National Laboratories, MS-1345
Albuquerque, NM 87185-1345
Phone: (505) 848-0759
e-mail: shconra@sandia.gov
Submitted October 1995 Layout design by Wanda Mar.
Submitted October 1995