A study of trichloroethylene (TCE) release rates from Borden, Ont. aquifer sediments through numerical modeling
MetadataShow full item record
Remediation strategies at sites contaminated with chlorinated volatile organic compounds (cVOCs), have enlightened how the aquifer, when flushed, shows initial high contaminant release rates followed by a long period of tailing and rebound after cessation of pumping. This results from the interplay between long-term contaminant mass release due to slow desorption and diffusion of contaminant mass from low-permeability zones into high permeability zones. The goal of this research is to quantify the effects of nonlinear sorption on non equilibrium intragranular diffusion and mass release from contaminated grains through numerical modeling of mass release rate experiments conducted in a companion study. The sorption process is explored in order to understand whether extremely different initial concentrations can be both fitted with one desorption model and, accordingly, whether the same set of parameters can be applied for different grain sizes. The laboratory experiments employed two sizes of sieved grains (0.7–2 mm and 2.8–4 mm) of the Gull River Formation. This Formation is a source rock for the well studied Borden Aquifer, Ontario, Canada. The Gull River Formation contains a low fraction organic carbon (f oc = 0.07%) content that is predominantly kerogen, a condensed form of organic matter acting as natural sorbent for nonpolar hydrophobic organic compounds. The grains were pre-equilibrated with dissolved trichloroethylene (TCE), one of the most common cVOCs detected in USA groundwater sites, at high (~10 3 mg/L) and low (~10 0 mg/L) initial concentrations. The sealed rock and water system was purged weekly and analyzed by gas chromatography over a period of up to about 200 days. The retarded intraparticle pore diffusion equation has been solved using a previously developed FORTRAN code that allows selection of nonlinear isotherm formulation. A method for automated calibration has been developed involving R and Ostrich. The input parameters for the models were estimated by independent measurements, while the most uncertain retardation factor, tortuosity and radius values, were estimated through the fitting routine. Three isotherms forms have been employed for the numerical simulations of the desorption experiments: linear, freundlich and Polanyi-Partition. The latter resulted as the best fit, based on the weighted root mean square error between simulated and observed data, and it was possible to identify one parameter set for radius, tortuosity and retardation factor that provided a reasonable fit to all the laboratory sets. Then the independence of the sorption isotherm form and parameterization from the initial concentration and grains' dimension, under non equilibrium conditions, has been demonstrated. Furthermore this study highlights the importance of a calibration methodology employing desorption and sorption at the equilibrium experiments to simultaneously reach a reliable solution for the sorption isotherm parameters.