Modeling dissolution and volatilization of LNAPL sources migrating on the groundwater table

A vertically averaged two-dimensional model was developed to describe areal spreading and migration of light nonaqueous-phase liquids (LNAPLs) introduced into the subsurface by spills or leaks from underground storage tanks. The NAPL transport model was coupled with two-dimensional contaminant transport models to predict contamination of soil gas and groundwater resulting from a LNAPL migrating on the water table. Numerical solutions were obtained by using the finite-difference method. Simulations and sensitivity analyses were conducted with a LNAPL of pure benzene to study LNAPL migration and groundwater contamination. The model was applied to subsurface contamination by jet fuel. Results indicated that LNAPL migration were affected mostly by volatilization. The generation and movement of the dissolved plume was affected by the geology of the site and the free-product plume. Most of the spilled mass remained as a free LNAPL phase 20 years after the spill. The migration of LNAPL for such a long period resulted in the contamination of both groundwater and a large volume of soil.     

Effect of water-table fluctuation on dissolution and biodegradation of a multi-component, light nonaqueous-phase liquid

Light nonaqueous-phase liquids (LNAPLs) such as gasoline and diesel fuel are among the most common causes of soil and groundwater contamination. Dissolution and subsequent advective transport of LNAPL components can negatively impact water supplies, while biodegradation is thought to be an important sink for this class of contaminants. We present a laboratory investigation of the effect of a water-table fluctuation on dissolution and biodegradation of a multi-component LNAPL (85% hexadecane, 5% toluene, 5% ethylbenzene, and 5% 2-methylnapthalene on a molar basis) in a pair of similar model aquifers (80 cm x 50 cm x 3 cm), one of which was subjected to a water-table fluctuation. Water-table fluctuation resulted in LNAPL and air entrapment below the water table, an increase in the vertical extent of the LNAPL source zone (by factor 6.7), and an increase in the volume of water passing through the source zone (by factor ~18). Effluent concentrations of dissolved LNAPL components were substantially higher and those of dissolved nitrate lower in the model aquifer where a fluctuation had been induced. Thus, water-table fluctuation led to enhanced biodegradation activity (28.3 mmol of nitrate consumed compared to 16.3 mmol in the model without fluctuation) as well as enhanced dissolution of LNAPL components. Despite the increased biodegradation, fluctuation led to increased elution of dissolved LNAPL components from the system (by factors 10-20). Hence, water-table fluctuations in LNAPL-contaminated aquifers might be expected to result in increased exposure of downgradient receptors to LNAPL components. Accordingly, water-table fluctuations in contaminated aquifers are probably undesirable unless the LNAPL is of minimal solubility or the dissolved-phase plume is not expected to reach a receptor due to distance or the presence of some form of containment.     

Simulation of the effect of remediation on EDB and 1,2-DCA plumes at sites contaminated by leaded gasoline

An analytical model is used to simulate the effects of partial source removal and plume remediation on ethylene dibromide (EDB) and 1,2-dichloroethane (1,2-DCA) plumes at contaminated underground storage tank (UST) sites. The risk posed by EDB, 1,2-DCA, and commingled gasoline hydrocarbons varies throughout the plume over time. Dissolution from the light nonaqueous phase liquid (LNAPL) determines the concentration of each contaminant near the source, but biological decay in the plume has a greater influence as distance downgradient from the source increases. For this reason, compounds that exceed regulatory standards near the source may not in downgradient plume zones. At UST sites, partial removal of a residual LNAPL source mass may serve as a stand alone remedial technique if dissolved concentrations in the source zone are within several orders of magnitude of the applicable government or remedial standards. This may be the case with 1,2-DCA; however, EDB is likely to be found at concentrations that are orders of magnitude higher than its low Maximum Contaminant Level (MCL) of 0.05 μg/L (micrograms per liter). For sites with significant EDB contamination, even when plume remediation is combined with source depletion, significant timeframes may be required to mitigate the impact of this compound. Benzene and MTBE are commonly the focus of remedial efforts at UST sites, but simulations presented here suggest that EDB, and to a lesser extent 1,2-DCA, could be the critical contaminants to consider in the remediation design process at many sites.     

Occurrence and attenuation of specific organic compounds in the groundwater plume at a former gasworks site

The changing contaminant pattern with travelled distance was investigated in the anaerobic groundwater plume downstream from an extended zone containing residual NAPL at a former gas manufacturing plant. With increasing distance, O- and N-heterocyclic aromatic compounds are enriched in the plume relative to the usually assessed coal tar constituents (poly- and monocyclic aromatic compounds). In a first approximation, the overall concentration decrease of the investigated compounds follows a first order overall decay. The half life distance in the plume downgradient from the source varied between 20 m for benzene and up to 167-303 m for alkyl-naphthalenes. Acenaphthene is degraded only within about 50 m downstream from the source area, then its concentration remains constant (ca. 180 microg/l) and far above the legal limit. Dimethyl-benzofurans were the most recalcitrant among all compounds which could be quantified with the analytical method available. The overall groundwater contamination in the plume is seriously underestimated if only BTEX and 16-EPA-PAHs are monitored.     

A pragmatic approach for estimation of source-zone emissions at LNAPL contaminated sites

When considering natural attenuation as a remediation strategy at a site contaminated by a light non-aqueous phase liquid (LNAPL), it is important to consider the emission of contaminants from the source zone. A quantification of source-zone emissions is essential both for comparison with down-gradient mass fluxes to provide an estimate of fractional mass flux reduction, as well as for estimating the source lifetime. Because the spatial distribution of LNAPL at a field site is strongly dependent on both the spill circumstances and the heterogeneity of the geologic materials, which can be problematic for in-situ determination, alternative methods for estimating source-zone emissions are needed. In this work, a three-dimensional multiphase flow and transport modelling approach is used to investigate the relationship between the lateral extent of an LNAPL body and the emission of contaminants to groundwater at a contaminated site. For simulations involving an LNAPL release in an aquifer comprised of heterogeneous porosity and permeability distributions that were generated geostatistically, it is shown that a simple linear relationship exists between the lateral extent of the LNAPL body in the capillary fringe and the emission to the aqueous phase. The parameters describing the relationship are found to be linear functions of the groundwater flow velocity and the vertical infiltration rate. This site-specific relationship provides a simple method to estimate contaminant emissions to groundwater at LNAPL contaminated sites.     

Inverse modeling to estimate LNAPL plume release timing

Methodologies are presented for dating releases of light nonaqueous phase liquids (LNAPLs) using an inverse modeling approach with simple analytical models. Models for LNAPL plume migration are presented to predict LNAPL plume velocity in the unsaturated and saturated zones as a function of basic soil and fluid properties. A relative mobility factor is introduced for LNAPL movement at the water table that depends primarily on the van Genuchten n parameter (related to the breadth of the soil pore size distribution) and the magnitude of water table fluctuations. Estimated LNAPL plume velocities compare reasonably with more rigorous numerical models, which may be used in cases where data availability warrant the greater effort entailed.Two methods of estimating release timing and its uncertainty are investigated. A direct estimation method is described that determines travel time for a single observed travel distance based on estimated soil and fluid properties. Release date uncertainty may be determined using the first order (FO) or Monte Carlo (MC) methods. The second method for estimating release date involves nonlinear parameter estimation utilizing distance vs. time measurements and other data.A case study is presented for a field site where independent estimates of release timing were obtained from a numerical modeling analysis. Release timing estimates based on direct inversion of the analytical timing model agree well with the numerical analysis. Results for a second field site indicate that release date confidence limits estimated by the FO method, assuming log-normally distributed travel times, are close to values determined by the MC method, which makes no assumption regarding the form of the travel time probability distribution.Results for a hypothetical problem indicate that LNAPL velocity and travel time may be accurately estimated if sufficient data on travel distance vs. time are available. Incorporating prior information on relevant soil and fluid properties into the objective function reduces the uncertainty in release date if prior estimates are accurate. However, biased prior estimates may lead to over- or underestimation of release date uncertainty. Simultaneous estimation of soil and fluid properties and release date is possible if prior information is available to condition the parameter estimates.     

Determination of the applicability of CERCLA's petroleum exclusion at contaminated sites – focus on metals

Areas impacted by petroleum refining and handling operations may become subject to CERCLA enforcement. Because of CERCLA's petroleum exclusion clause, determining whether contamination in a CERCLA Site originated from petroleum products or hazardous wastes becomes important. Because certain metals are typically enriched in wastes relative to petroleum products and background soils, knowledge of metal contents in these potential end member metal sources is an important step towards contaminant source identification in soils and sediments. In LNAPL plumes, metal content, particularly lead, may be claimed to be the result of wastes mishandling and not due only to the presence of leaded gasoline in the plume. Analysis of the percent gasoline in the plume and accounting for weathering are steps to determining whether the lead content in an LNAPL plume is within the historical lead concentration ranges in gasolines. In addition to metals analyses, understanding of operational parameters such as the history of petroleum refining and handling operations, leaks, spills, and cleanup activities are needed for successful conclusion of the applicability of the petroleum exclusion.     

地下水中轻质有机污染物(LNAPL)透镜体研究

在二维砂槽模型中模拟了轻质油在均匀多孔介质地下水非饱和区中的运移过程。模拟结果表明,地下水毛细区是轻质油污染的重点区,除了ＬＮＡＰＬ的残留以外,进入地下水饱和的ＬＮＡＰＬ终将被地下水顶托回到毛细区中,毛细区以上的约大多民将在重力作用下进入毛细区中,试验中观察到达稳定状态时ＬＮＡＰＬ透镜体的上边缘略微高出毛细区。利用多孔介质毛细管模型,建立了利用界面张力、接触角、介质特征孔隙直径等物理量估算不同位置     

Numerical experiments and field results on the size of steady state plumes

Contaminated groundwater poses a serious risk for drinking water supplies. Under certain conditions, however, groundwater contamination remains restricted to a tolerable extent because of natural attenuation processes. We present an innovative approach to evaluate the size of these so-called steady-state plumes by 2-D and 1-D modelling in homogeneous aquifers. If longitudinal mixing is negligible, scenarios can be modelled in a simplified way using a 1-D domain vertical to the direction of flow. We analysed the sensitivity of the plume length with respect to biodegradation kinetics, flow velocity, transverse vertical dispersivity alphat, the source and aquifer geometry and reaction stoichiometry. Our findings indicate that for many readily biodegradable compounds transverse-dispersive mixing rather than reaction kinetics is the limiting factor for natural attenuation. Therefore, if alphat, aquifer and source geometry and concentrations of electron acceptors and donors are known, the length of the steady state contaminant plume can be predicted. The approach is validated under field conditions for an ammonium plume at a former landfill site in SW Germany.     

Oxygenated gasoline release in the unsaturated zone, Part 2: Downgradient transport of ethanol and hydrocarbons

In the event of a gasoline spill containing oxygenated compounds such as ethanol and MTBE, it is important to consider the impacts these compounds might have on subsurface contamination. One of the main concerns commonly associated with ethanol is that it might decrease the biodegradation of aromatic hydrocarbon compounds, leading to an increase in the hydrocarbon dissolved plume lengths. The first part of this study (Part 1) showed that when gasoline containing ethanol infiltrates the unsaturated zone, ethanol is likely to partition to and be retained in the unsaturated zone pore water. In this study (Part 2), a controlled field test is combined with a two-dimensional laboratory test and three-dimensional numerical modelling to investigate how ethanol retention in the unsaturated zone affects the downgradient behaviour of ethanol and aromatic hydrocarbon compounds. Ethanol transport downgradient was extremely limited. The appearance of ethanol in downgradient wells was delayed and the concentrations were lower than would be expected based on equilibrium dissolution. Oscillations in the water table resulted in minor flushing of ethanol, but its effect could still be perceived as an increase in the groundwater concentrations downgradient from the source zone. Ethanol partitioning to the unsaturated zone pore water reduced its mass fraction within the NAPL thus reducing its anticipated impact on the fate of the hydrocarbon compounds. A conceptual numerical simulation indicated that the potential ethanol-induced increase in benzene plume length after 20 years could decrease from 136% to 40% when ethanol retention in the unsaturated zone is considered.     

Fingerprinting of Gasoline and Coal Tar NAPL Volatile Hydrocarbons Dissolved in Groundwater

Accidental spills and chronic leaks of fuel oil or other hydrocarbon material (e.g., coal tar) often result in subsurface accumulation of nonaqueous phase liquid (NAPL), which can be a subsequent source of contamination in groundwater. Linking hydrocarbons in groundwater to a source NAPL has been difficult when using standard target analytes (e.g., BTEX) because of differences in partitioning properties of the analytes between the source NAPL and groundwater. Because aqueous solubility is predicted to be the controlling influence in the partitioning of hydrocarbons from NAPL to groundwater, a solubility-based approach to matching dissolved hydrocarbons in groundwater to their source NAPL has been developed and validated for two sites with commonly encountered types of NAPL contamination. Specifically, a gasoline LNAPL and a coal tar DNAPL from two separate sites (West Virginia and California) and groundwater interfaced with these NAPLs were analyzed for approximately 50 gasoline-range hydrocarbons consisting of paraffin, isoparaffin, (mono-) aromatic, naphthene, and olefin compounds (PIANO). Solubility characteristics of selected alkyl aromatic hydrocarbons from the PIANO analysis were used to identify a set of diagnostic hydrocarbons, expressed as hydrocarbon ratios, which were found to be useful in distinguishing the source(s) of hydrocarbons in groundwater. At the West Virginia site, the diagnostic ratios in a downgradient groundwater sample were similar to those of a gasoline NAPL at that site, indicating the source of hydrocarbons to the groundwater was the upgradient gasoline NAPL. The diagnostic ratios of the groundwater in contact with the gasoline NAPL and the remote groundwater were also similar, providing evidence that the diagnostic ratios were retained during transport in the aquifer. At the California site, diagnostic ratios in a cross-gradient groundwater sample differed from those of the coal tar NAPL at that site, indicating that the remote groundwater hydrocarbons did not originate from the coal tar contamination. Environmental factors such as selective degradation of specific isomers and various geological conditions (e.g., soil mineralogy, and organic content) may confound the application of this solubility-based fingerprinting approach. Thus, it is recommended that multiple diagnostic pairs be simultaneously evaluated when considering this fingerprinting approach for specific sites and product types.     

Evaluating LNAPL contamination using GPR signal attenuation analysis and dielectric property measurements: practical implications for hydrological studies

Groundwater and sub-surface contamination by Light Non-Aqueous Phase Liquids (LNAPLs) is one of the industrial world's most pressing environmental issues and a thorough understanding of the hydrological, physical and bio-chemical properties of the sub-surface is key to determining the spatial and temporal development of any particular contamination event. Non-invasive geophysical techniques (such as electrical resistivity, electromagnetic conductivity, Ground-Penetrating Radar, etc.) have proved to be successful sub-surface investigation and characterisation tools with Ground-Penetrating Radar (GPR) being particularly popular. Recent studies have shown that the spatial/temporal variation in GPR signal attenuation can provide important information on the electrical properties of the sub-surface materials that, in turn, can be used to assess the physical and hydrological nature of the pore fluids and associated contaminants. Unfortunately, a high percentage of current LNAPL-related GPR studies focus on contaminant mapping only, with little emphasis being placed on characterising the hydrological properties (e.g., determining contaminant saturation index, etc.). By comparing laboratory-based, dielectric measurements of LNAPL contaminated materials with the GPR signal attenuation observed in both contaminated and 'clean' areas of an LNAPL contaminated site, new insights have been gained into the nature of contaminant distribution/saturation and the likely signal attenuation mechanisms. The results show that, despite some practical limitations of the analysis technique, meaningful hydrological interpretations can be obtained on the contaminant properties, saturation index and bio-degradation processes. A generalised attenuation/saturation model has been developed that describes the physical and attenuation enhancement characteristics of the contaminated areas and reveals that the most significant attenuation is related to smeared zone surrounding the seasonally changing water table interface. It is envisaged that the model will provide a basis for the interpretation of GPR data from analogous LNAPL contaminated sites and provide investigators with an appreciation of the merits and limitations of GPR-based, attenuation analysis techniques for hydrological applications.     

A field trial to assess the performance of CO2-supersaturated water injection for residual volatile LNAPL recovery

A pilot scale field trial was conducted to evaluate the recovery of volatile, light non-aqueous phase liquids (LNAPLs) using a novel remediation method termed supersaturated water injection (SWI). SWI uses a patented technology to efficiently dissolve high concentrations of CO₂ into water at elevated pressures. This water is injected into the subsurface resulting in the nucleation of CO₂ bubbles at and away from the injection point. The nucleating bubbles coalesce, rise and volatilize residual LNAPL ganglia. In this study, an LNAPL composed of 103 kg of volatile pentane and hexane, and 30 kg of non-volatile Soltrol was emplaced below the water table at residual saturation. The SWI technology removed 78% of the pentane and 50% of the less volatile hexane. Contaminant mass was still being removed when the system was shut down for practical reasons. The mass removed is comparable to that expected for air sparging but a much smaller volume of gas was injected using the SWI system.     

Groundwater-surface water interaction and its role on TCE groundwater plume attenuation

A field investigation of a TCE plume in a surficial sand aquifer shows that groundwater-surface water interactions strongly influence apparent plume attenuation. At the site, a former industrial facility in Connecticut, depth-discrete monitoring along three cross-sections (transects) perpendicular to groundwater flow shows a persistent VOC plume extending 700 m from the DNAPL source zone to a mid-size river. Maximum TCE concentrations along a transect 280 m from the source were in the 1000s of microg/L with minimal degradation products. Beyond this, the land surface drops abruptly to a lower terrace where a shallow pond and small streams occur. Two transects along the lower terrace, one midway between the facility and river just downgradient of the pond and one along the edge of the river, give the appearance that the plume has strongly attenuated. At the river, maximum TCE concentrations in the 10s of microg/L and similar levels of its degradation product cis-DCE show direct plume discharge from groundwater to the river is negligible. Although degradation plays a role in the strong plume attenuation, the major attenuation factor is partial groundwater plume discharge to surface water (i.e. the pond and small streams), where some mass loss occurs via water-air exchange. Groundwater and stream mass discharge estimates show that more than half of the plume mass discharge crossing the first transect, before surface water interactions occur, reaches the river directly via streamflow, although river concentrations were below detection due to dilution. This study shows that groundwater and surface water concentration measurements together provide greater confidence in identifying and quantifying natural attenuation processes at this site, rather than groundwater measurements alone.     

Two-dimensional laboratory simulation of LNAPL infiltration and redistribution in the vadose zone

A quantitative two-dimensional laboratory experiment was conducted to investigate the immiscible flow of a light non-aqueous phase liquid (LNAPL) in the vadose zone. An image analysis technique was used to determine the two-dimensional saturation distribution of LNAPL, water and air during LNAPL infiltration and redistribution. Vertical water saturation variations were also continuously monitored with miniature resistivity probes. LNAPL and water pressures were measured using hydrophobic and hydrophilic tensiometers. This study is limited to homogeneous geological conditions, but the unique experimental methods developed will be used to examine more complex systems. The pressure measurements and the quantification of the saturation distribution of all the fluids in the entire flow domain under transient conditions provide quantitative data essential for testing the predictive capability of numerical models. The data are used to examine the adequacy of the constitutive pressure-saturation relations that are used in multiphase flow models. The results indicate that refinement of these commonly used hydraulic relations is needed for accurate model prediction. It is noted in particular that, in three-fluid phase systems, models should account for the existence of a residual NAPL saturation occurring after NAPL drainage. This is of notable importance because residual NAPL can act as a non negligible persistent source of contamination.     

Semi-analytical solution of two-dimensional sharp interface LNAPL transport models

In this paper, we present semi-analytical solutions for two-dimensional equations governing transport of Light Non-Aqueous Phase Liquids (LNAPL) in unconfined aquifers. The proposed model is based on sharp interface displacement and steady groundwater flow assumptions, where both the water–LNAPL interface and the LNAPL–air interface are represented as sharp interfaces. In the case of steady groundwater flow, these equations can be reduced to a two-dimensional nonlinear solute transport equation, with the LNAPL thickness in the free product lens being the primary unknown variable. The linearized form of this solute transport equation falls into the category of two-dimensional transport equation with time-dependent dispersion coefficients. This equation can be solved analytically for an infinite domain region. In this paper, the general form of the analytical solution for the transport equation, as well as the solutions for some specific cases are presented. To demonstrate the utility of the proposed solution, numerical results obtained for two example problems are discussed and presented comparatively with a finite-element solution and other more restrictive solutions available in the literature. Although the solutions discussed in this paper have some simplifying assumptions, such as sharp-interfaces between fluid phases, steady groundwater flow and homogeneous aquifer properties, the semi-analytical solutions presented in this study may be used effectively as bench mark solutions in evaluating LNAPL migration in the subsurface. These solutions are simple and cost effective to implement and may be used in the calibration of other more complex numerical solutions that can be found in the literature.     

Progression of natural attenuation processes at a crude-oil spill site: I. Geochemical evolution of the plume

A 16-year study of a hydrocarbon plume shows that the extent of contaminant migration and compound-specific behavior have changed as redox reactions, most notably iron reduction, have progressed over time. Concentration changes at a small scale, determined from analysis of pore-water samples drained from aquifer cores, are compared with concentration changes at the plume scale, determined from analysis of water samples from an observation well network. The small-scale data show clearly that the hydrocarbon plume is growing slowly as sediment iron oxides are depleted. Contaminants, such as ortho-xylene that appeared not to be moving downgradient from the oil on the basis of observation well data, are migrating in thin layers as the aquifer evolves to methanogenic conditions. However, the plume-scale observation well data show that the downgradient extent of the Fe2+ and BTEX plume did not change between 1992 and 1995. Instead, depletion of the unstable Fe (III) oxides near the subsurface crude-oil source has caused the maximum dissolved iron concentration zone within the plume to spread at a rate of approximately 3 m/year. The zone of maximum concentrations of benzene, toluene, ethylbenzene and xylene (BTEX) has also spread within the anoxic plume. In monitoring the remediation of hydrocarbon-contaminated ground water by natural attenuation, subtle concentration changes in observation well data from the anoxic zone may be diagnostic of depletion of the intrinsic electron-accepting capacity of the aquifer. Recognition of these subtle patterns may allow early prediction of growth of the hydrocarbon plume.     

Estimation of Methyl tert -Butyl Ether Plume Length Using the Domenico Analytical Model

Methyl tert -butyl ether (MTBE) plume is controlled by many factors, primarily by groundwater flow velocity, dispersion, natural attenuation. This study employed an analytical model introduced by Domemico (1987, J. Hydrol 91 , 49-58.) to describe the MTBE concentration distribution horizontal pattern and estimated the MTBE plume length. The model was applied to 90 leaking underground storage tank cases in Los Angeles, CA, U.S.A. The analytical model was calibrated with field data for each ease using a Microsoft Excel spreadsheet program. Methyl tert -butyl ether concentrations in one source monitoring well and one to two downgradient centerline monitoring wells were used for each case study. When the centerline well is not available, the closest off-centerline wells were projected to the centerline using an ellipse trigonometry method. The model parameter values for longitudinal dispersivity, groundwater velocity, and degradation rate constant were calibrated using the field data and then used to estimate the maximum distance between source well and the plume edge. This study demonstrates that the Domenico model can be applied to MTBE plume investigation when adequate field data are available. The correlation coefficients calculated based on the results of the 90 case studies indicate that MTBE plume length has a poor correlation with MTBE concentration at the source well, and a moderate negative correlation with the degradation rate constant ( m 0.65) and u / v ratio (0.64). Furthermore, MTBE plume length has a poor correlation with the longitudinal dispersivity ( m 0.4), hydraulic gradient ( m 0.1), and groundwater velocity (0.17).     

Migration and natural fate of a coal tar creosote plume: 1. Overview and plume development

A volume of sand containing coal tar creosote was emplaced below the water table at CFB Borden to investigate natural attenuation processes for complex biodegradable mixtures. Coal tar creosote is a mixture of more than 200 polycyclic aromatic hydrocarbons, heterocyclic compounds and phenolic compounds. A representative group of seven compounds was selected for detailed study: phenol, m-xylene, naphthalene, phenanthrene, 1-methylnaphthalene, dibenzofuran and carbazole. Movement of groundwater through the source led to the development of a dissolved organic plume, which was studied over a 4-year period. Qualitative plume observations and mass balance calculations indicated two key conclusions: (1) compounds from the same source can display distinctly different patterns of plume development and (2) mass transformation was a major influence on plume behaviour for all observed compounds.     

Plume persistence caused by back diffusion from thin clay layers in a sand aquifer following TCE source-zone hydraulic isolation

This paper concludes that back diffusion from one or a few thin clayey beds in a sand aquifer can cause contaminant persistence above MCLs in a sand aquifer long after the source zone initially causing the plume is isolated or removed. This conclusion is based on an intensive case study of a TCE contaminated site in Florida, with the processes evaluated using numerical modeling. At this site, the TCE DNAPL zone formed decades ago, and was hydraulically isolated by means of an innovative system performing groundwater extraction, treatment and re-injection. Treated water is re-injected in a row of injection wells situated a short distance downgradient of the extraction wells, creating a clean-water displacement front to efficiently flush the downgradient plume. This scheme avoids the creation of stagnation zones typical of most groundwater pump-and-treat systems, thereby minimizing the time for aquifer flushing and therefore downgradient cleanup. The system began operation in August 2002 and although the performance monitoring shows substantial declines in concentrations, detectable levels of TCE and degradation products persist downgradient of the re-injection wells, long after the TCE should have disappeared based on calculations assuming a nearly homogenous sand aquifer. Three hypotheses were assessed for this plume persistence: 1) incomplete source-zone capture, 2) DNAPL occurrence downgradient of the re-injection wells, and 3) back diffusion from one or more thin clay beds in the aquifer. After careful consideration, the first two hypotheses were eliminated, leaving back diffusion as the only plausible hypothesis, supported by detailed measurements of VOC concentrations within and near the clay beds and also by numerical model simulations that closely represent the field site hydrogeologic conditions. The model was also used to simulate a more generalized, hypothetical situation where more thin clayey beds occur in a sand aquifer with an underlying aquitard. While there is no doubt that DNAPL source mass reduction can eventually improve downgradient groundwater quality, the magnitude and time scale over which the improvement occurs is the major uncertainty given current characterization approaches. This study shows that even one thin clay bed, less than 0.2 m thick, can cause plume persistence due to back diffusion for several years or even decades after the flux from the source is completely isolated. Thin clay beds, which have a large storage capacity for dissolved and sorbed contaminant mass, are common in many types of sandy aquifers. However, without careful inspection of continuous cores and sampling, such thin clay beds, and their potential for causing long-term back-diffusion effects, can easily go unnoticed during site characterization.