Simulation of solute transport of tetrachloroethylene in ground water of the glacial-drift aquifer at the Savage Municipal Well Superfund Site, Milford, New Hampshire, 1960-2000

The Savage Municipal Well Superfund site, named after the former municipal water-supply well for the town of Milford, is underlain by a 0.5-square mile plume of volatile organic compounds (VOCs), primarily tetrachloroethylene (PCE). The plume occurs mostly within a highly transmissive sand-and-gravel unit, but also extends to an underlying till and bedrock unit. The plume logistically is divided into two areas termed Operable Unit No. 1 (OU1), which contains the primary source area, and Operable Unit No. 2 (OU2), which is the extended plume area.

PCE concentrations in excess of 100,000 parts per billion (ppb) had been detected in the OU1 area in 1995, indicating a likely Dense Non-Aqueous Phase Liquid (DNAPL) source. In the fall of 1998, the New Hampshire Department of Environmental Services (NHDES) and the U.S. Environmental Protection Agency (USEPA) installed a remedial system in OU1. The OU1 remedial system includes a low-permeability barrier that encircles the highest detected concentrations of PCE, and a series of injection and extraction wells. The barrier primarily sits atop bedrock and penetrates the full thickness of the sand and gravel; and in some places, the full thickness of the underlying basal till. The sand and gravel unit and the till comprise the aquifer termed the Milford-Souhegan glacial-drift aquifer (MSGD).

Two-dimensional and three-dimensional finite-difference solute-transport models of the unconsolidated sediments (MSGD aquifer) were constructed to help evaluate solute-transport processes, assess the effectiveness of remedial activities in OU1, and to help design remedial strategies in OU2. The solute-transport models simulate PCE concentrations, and model results were compared to observed concentrations of PCE. Simulations were grouped into the following three time periods: an historical calibration of the distribution of PCE from the initial input (circa 1960) of PCE into the subsurface to the 1990s, a pre-remedial calibration from 1995 to 1998, and a remedial (post-barrier wall) calibration from 1998 to 1999. Model results also were checked against observed PCE concentrations from May and June 2000 as a post-audit of model performance.

Results of the simulations of the two-dimensional model for the historical calibration indicate that the model-computed length of the plume is affected by the retardation factor (retardation). Values of retardation greater than 3 caused the longitudinal length of the computed plume to be too short compared to the observed plume. A retardation of 2-2.5 produced a reasonable comparison between computed and observed PCE concentrations. Testing of different starting times and rates of mass input of PCE indicated that the plume reaches a quasi steady-state distribution in about 20 years regardless of the rate of mass input or values of the solute-transport parameters (retardation, dispersion, and irreversible reaction) assigned the model.

Results of the simulations of the three-dimensional model for the pre-remedial (1995-98) calibration of PCE for the OU2 area identified some spatial biases in computed concentrations that generally were unaffected by changes in retardation. The computed PCE concentrations exceeded observed concentrations along the northern part of the plume in OU2, where PCE increases were observed in a bedrock well. These results indicate that some PCE in this area may be entering the bedrock, which is not simulated in the model. Conversely, computed PCE concentrations were less than observed concentrations along the southern part of the plume in OU2. Because testing of high (above 4) values of retardation did little to reduce residuals, it is concluded that the low computed PCE concentrations along the southern flank are likely the result of an underestimation of the initial PCE mass in this area or an unaccounted source of PCE.

Results of the simulations of the three-dimensional model for the remedial calibration period (1998-99) and po