Strategies for Managing Risk due to Back Diffusion.

Back diffusion of contaminants from secondary sources may hamper site remediation if it is not properly addressed in the remedial design. A review of all reported technologies and strategies that have been or could be applied to address plume persistence due to back diffusion as published in the peer-reviewed literature is provided. We classify these into four major categories. The first category consists of those approaches that do not include active measures to specifically address contamination in the low permeable zones (LPZs) and can therefore be considered passive LPZ management approaches. A disadvantage of these approaches is the long duration that may be required to meet acceptable endpoints; however, this allows degradation to potentially play a significant part even at modest rates. The remaining three categories all use approaches to specifically address contaminants in the LPZ. The second category consists of strategies that promote contaminant destruction through the forward diffusion of amendments into the LPZ. A variety of laboratory tests indicate concentration or flux reductions range from no improvement, to reductions as high as four orders-of-magnitude depending on the evaluation metric. The third category consists of strategies that alter physical characteristics of the secondary source, and includes viscosity modification, fracturing, and soil mixing. Each of these offer unique advantages and are often used to deliver one or more amendments for contaminant treatment. The final category consists of thermal and electrokinetic remediation, both less susceptible to permeability contrast limitations. However, they are not routinely used for secondary-source treatment.

Source strength functions from long-term monitoring data and spatially distributed mass discharge measurements.

Source strength functions (SSF), defined as contaminant mass discharge or flux-averaged concentration from dense nonaqueous phase liquid (DNAPL) source zones as a function of time, provide a quantitative model of DNAPL source-zone behavior. Such information is useful for calibration of screening-level models to assist with site management decisions. We investigate the use of historic data collected during long-term monitoring (LTM) activities at a site in Rhode Island to predict the SSF based on temporal mass discharge measurements at a fixed location, as well as SSF estimation using mass discharge measurements at a fixed time from three spatially distributed control planes. Mass discharge based on LTM data decreased from ~300 g/day in 1996 to ~70 g/day in 2012 at a control plane downgradient of the suspected DNAPL source zone, and indicates an overall decline of ~80% in 16 years. These measurements were compared to current mass discharge measurements across three spatially distributed control planes. Results indicate that mass discharge increased in the downgradient direction, and was ~6 g/day, ~37 g/day, and ~400 g/day at near, intermediate, and far distances from the suspected source zone, respectively. This behavior was expected given the decreasing trend observed in the LTM data at a fixed location. These two data sets were compared using travel time as a means to plot the data sets on a common axis. The similarity between the two data sets gives greater confidence to the use of this combined data set for site-specific SSF estimation relative to either the sole use of LTM or spatially distributed data sets.

Screening-level estimates of mass discharge uncertainty from point measurement methods.

The uncertainty of mass discharge measurements associated with point-scale measurement techniques was investigated by deriving analytical solutions for the mass discharge coefficient of variation for two simplified, conceptual models. In the first case, a depth-averaged domain was assumed, consisting of one-dimensional groundwater flow perpendicular to a one-dimensional control plane of uniformly spaced sampling points. The contaminant flux along the control plane was assumed to be normally distributed. The second case consisted of one-dimensional groundwater flow perpendicular to a two-dimensional control plane of uniformly spaced sampling points. The contaminant flux in this case was assumed to be distributed according to a bivariate normal distribution. The center point for the flux distributions in both cases was allowed to vary in the domain of the control plane as a uniform random variable. Simplified equations for the uncertainty were investigated to facilitate screening-level evaluations of uncertainty as a function of sampling network design. Results were used to express uncertainty as a function of the length of the control plane and number of wells, or alternatively as a function of the sample spacing. Uncertainty was also expressed as a function of a new dimensionless parameter, Ω, defined as the ratio of the maximum local flux to the product of mass discharge and sample density. Expressing uncertainty as a function of Ω provided a convenient means to demonstrate the relationship between uncertainty, the magnitude of a local hot spot, magnitude of mass discharge, distribution of the contaminant across the control plane, and the sampling density.

The uncertainty of mass discharge measurements using pumping methods under simplified conditions.

Mass discharge measurements at contaminated sites have been used to assist with site management decisions, and can be divided into two broad categories: point-scale measurement techniques and pumping methods. Pumping methods can be sub-divided based on the pumping procedures used into sequential, concurrent, and tandem circulating well categories. Recent work has investigated the uncertainty of point measurement methods, and to a lesser extent, pumping methods. However, the focus of this study was a direct comparison of uncertainty between the various pumping method approaches that have been used, as well as a comparison of uncertainty between pumping and point measurement methods. Mass discharge measurement error was investigated using a Monte Carlo modeling analysis as a function of the contaminant plume position and width, and as a function of the pumping conditions used in the different pumping tests. Results indicated that for the conditions investigated, uncertainty in mass discharge estimates based on pumping methods was 1.3 to 16 times less than point measurement method uncertainty, and that a sequential pumping approach resulted in 5 to 12 times less uncertainty than the concurrent pumping or tandem circulating well approaches. Uncertainty was also investigated as a function of the plume width relative to well spacing. For a given well spacing, uncertainty decreased for all methods as the plume width increased, and comparable levels of uncertainty between point measurement and pumping methods were obtained when three wells were distributed across the plume. A hybrid pumping technique in which alternate wells were pumped concurrently in two separate campaigns yielded similar uncertainty to the sequential pumping approach. This suggests that the hybrid approach can be used to capitalize on the advantages of sequential pumping yet minimize the overall test duration.

Changes in contaminant mass discharge from DNAPL source mass depletion: evaluation at two field sites.

Changes in contaminant fluxes resulting from aggressive remediation of dense nonaqueous phase liquid (DNAPL) source zone were investigated at two sites, one at Hill Air Force Base (AFB), Utah, and the other at Ft. Lewis Military Reservation, Washington. Passive Flux Meters (PFM) and a variation of the Integral Pumping Test (IPT) were used to measure fluxes in ten wells installed along a transect down-gradient of the trichloroethylene (TCE) source zone, and perpendicular to the mean groundwater flow direction. At both sites, groundwater and contaminant fluxes were measured before and after the source-zone treatment. The measured contaminant fluxes (J; ML(-2)T(-1)) were integrated across the well transect to estimate contaminant mass discharge (M(D); MT(-1)) from the source zone. Estimated M(D) before source treatment, based on both PFM and IPT methods, were approximately 76 g/day for TCE at the Hill AFB site; and approximately 640 g/day for TCE, and approximately 206 g/day for cis-dichloroethylene (DCE) at the Ft. Lewis site. TCE flux measurements made 1 year after source treatment at the Hill AFB site decreased to approximately 5 g/day. On the other hand, increased fluxes of DCE, a degradation byproduct of TCE, in tests subsequent to remediation at the Hill AFB site suggest enhanced microbial degradation after surfactant flooding. At the Ft. Lewis site, TCE mass discharge rates subsequent to remediation decreased to approximately 3 g/day for TCE and approximately 3 g/day for DCE approximately 1.8 years after remediation. At both field sites, PFM and IPT approaches provided comparable results for contaminant mass discharge rates, and show significant reductions (>90%) in TCE mass discharge as a result of DNAPL mass depletion from the source zone.

Continuous determination of high-vapor-phase concentrations of tetrachloroethylene using on-line mass spectrometry.

A method was developed to determine the vapor concentration of tetrachloroethylene (PCE) at and below its equilibrium vapor-phase concentration, 168 000 microg/L (25 degrees C). Vapor samples were drawn by vacuum into a six-port sampling valve and injected through a jet separator into an ion trap mass spectrometer (MS). This on-line MS can continuously sample a vapor stream and provide vapor concentrations every 30 s. Calibration of the instrument was done by creating a saturated stream of PCE vapor, sampling the vapor with the on-line MS and with thermal desorption tubes, and correlating the peak area response from the MS with the vapor concentration determined by automated thermal desorption gas chromatography mass spectrometry. Dilution of the saturated stream provided lower concentrations of PCE vapor. The method was developed to monitor the vapor concentration of PCE that was sparged from a two-dimensional flow chamber and for determination of the total PCE mass removed during each sparge event. The method has potential application for analysis of gas-phase tracers.

Field demonstration of surfactant-enhanced solubilization of DNAPL at Dover Air Force Base, Delaware.

This study reports on a surfactant-based flood for tetrachloroethylene (PCE) removal from a control test cell at the Dover National Test Site. The surfactant formulation (sodium dihexyl sulfosuccinate (Aerosol-MA or AMA), isopropanol and calcium chloride) was able to achieve a high concentration of PCE in swollen micelles (supersolubilization) without vertical PCE migration. The hydraulic system included eight screened wells that were operated in both vertical circulation and line drive configurations. After 10 pore volumes of flushing, the overall PCE removal was 68% (65% of which corresponded to the surfactant flooding alone). In addition, the residual PCE saturation was reduced from 0.7% to 0.2%, and the concentration of PCE in the groundwater was reduced from 37-190 mg/L before the flushing to 7.3 mg/L after flooding. Recycling the surfactant solution reduced the required surfactant mass (and thus cost, and waste) by 90%. Close to 80% of the total PCE removal was obtained during the first five pore volumes which were operated in an upward vertical circulation flow scheme. No free oil phase was observed during the test. Further analysis of multilevel sampler data suggests that most of the trapped oil remaining in the cell was likely localized in secluded regions of the aquifer, which helps explain the lower PCE groundwater concentration after remedial activities. In summary, this field study demonstrated the feasibility of surfactant-enhanced remediation to reduce the mass in the source zone and significantly reduce the PCE aqueous concentration and therefore the risk associated with the contaminant plume.

Errors in NAPL volume estimates due to systematic measurement errors during partitioning tracer tests.

During moment-based analyses of partitioning tracer tests, systematic errors in volume and concentration measurements propagate to yield errors in the saturation and volume estimates for nonaqueous phase liquid (NAPL). Derived expressions could be applied to help practitioners bracket their estimates of NAPL saturation and volume obtained from such tests. In practice, many of these effects may be overshadowed by other complications experienced in the field. Errors are propagated for systematic constant (offset) volume, proportional volume, and constant (offset) concentration errors. Previous efforts to quantify the impact of these errors were predicated upon the specific assumption that nonpartitioning and partitioning masses were equal. The current work relaxes that assumption and is therefore more general in scope. Through the use of nondimensional concentration, systematic proportional concentration errors do not affect the accuracy of the method. Specific consideration needs to be given to accurate flow measurements and minimizing baseline concentration errors when performing partitioning tracer tests in order to prevent the propagation of systematic errors.

Uncertainty in NAPL volume estimates due to random measurement errors during partitioning tracer tests.

The uncertainty in NAPL volume estimates obtained through partitioning tracers can be quantified as a function of random errors in volume and concentration measurements when moments are calculated from experimentally measured breakthrough curves using the trapezoidal rule for numerical integration. The methodology is based upon standard stochastic methods for random error propagation. Monte Carlo simulations using a synthetic data set derived from the one-dimensional solution of the advective-dispersive equation serve to verify the process. It is shown that the uncertainty in NAPL volume predictions nonlinearly increases as the retardation factor decreases. An important result of this observation is that there is a large degree of uncertainty in using partitioning tracers to conclude NAPL is absent from the swept zone. Under the conditions investigated, random errors in concentration measurements are shown to have a greater impact on NAPL volume uncertainty than random errors in volume measurements, and it is also shown that uncertainty in NAPL volume decreases as the resolution of the breakthrough curves increases. The impact of uncertainty in background retardation (i.e., sorption of partitioning tracers in the absence of NAPL) was also investigated, and it likewise indicated that the relative uncertainty in NAPL volume estimates increases as the retardation factor decreases.

Design of aquifer remediation systems: (1) describing hydraulic structure and NAPL architecture using tracers.

Aquifer heterogeneity (structure) and NAPL distribution (architecture) are described based on tracer data. An inverse modelling approach that estimates the hydraulic structure and NAPL architecture based on a Lagrangian stochastic model where the hydraulic structure is described by one or more populations of lognormally distributed travel times and the NAPL architecture is selected from eight possible assumed distributions. Optimization of the model parameters for each tested realization is based on the minimization of the sum of the square residuals between the log of measured tracer data and model predictions for the same temporal observation. For a given NAPL architecture the error is reduced with each added population. Model selection was based on a fitness which penalized models for increasing complexity. The technique is demonstrated under a range of hydrologic and contaminant settings using data from three small field-scale tracer tests: the first implementation at an LNAPL site using a line-drive flow pattern, the second at a DNAPL site with an inverted five-spot flow pattern, and the third at the same DNAPL site using a vertical circulation flow pattern. The Lagrangian model was capable of accurately duplicating experimentally derived tracer breakthrough curves, with a correlation coefficient of 0.97 or better. Furthermore, the model estimate of the NAPL volume is similar to the estimates based on moment analysis of field data.

Design of aquifer remediation systems: (2) estimating site-specific performance and benefits of partial source removal.

A Lagrangian stochastic model is proposed as a tool that can be utilized in forecasting remedial performance and estimating the benefits (in terms of flux and mass reduction) derived from a source zone remedial effort. The stochastic functional relationships that describe the hydraulic "structure" and non-aqueous phase liquid (NAPL) "architecture" have been described in a companion paper (Enfield, C.G., Wood, A.L., Espinoza, F.P., Brooks, M.C., Annable, M., Rao, P.S.C., this issue. Design of aquifer remediation systems: (1) describing hydraulic structure and NAPL architecture using tracers. J. Contam. Hydrol.). The previously defined functions were used along with the properties of the remedial fluids to describe remedial performance. There are two objectives for this paper. First, is to show that a simple analytic element model can be used to give a reasonable estimate of system performance. This is accomplished by comparing forecast performance to observed performance. The second objective is to display the model output in terms of change in mass flux and mass removal as a function of pore volumes of remedial fluid injected. The modelling results suggest that short term benefits are obtained and related to mass reduction at the sites where the model was tested.

Controlled release, blind test of DNAPL remediation by ethanol flushing.

A dense nonaqueous phase liquid (DNAPL) source zone was established within a sheet-pile isolated cell through a controlled release of perchloroethylene (PCE) to evaluate DNAPL remediation by in-situ cosolvent flushing. Ethanol was used as the cosolvent, and the main remedial mechanism was enhanced dissolution based on the phase behavior of the water-ethanol-PCE system. Based on the knowledge of the actual PCE volume introduced into the cell, it was estimated that 83 L of PCE were present at the start of the test. Over a 40-day period, 64% of the PCE was removed by flushing the cell with an alcohol solution of approximately 70% ethanol and 30% water. High removal efficiencies at the end of the test indicated that more PCE could have been removed had it been possible to continue the demonstration. The ethanol solution extracted from the cell was recycled during the test using activated carbon and air stripping treatment. Both of these treatment processes were successful in removing PCE for recycling purposes, with minimal impact on the ethanol content in the treated fluids. Results from pre- and post-flushing partitioning tracer tests overestimated the treatment performance. However, both of these tracer tests missed significant amounts of the PCE present, likely due to inaccessibility of the PCE. The tracer results suggest that some PCE was inaccessible to the ethanol solution which led to the inefficient PCE removal rates observed. The flux-averaged aqueous PCE concentrations measured in the post-flushing tracer test were reduced by a factor of 3 to 4 in the extraction wells that showed the highest PCE removal compared to those concentrations in the pre-flushing tracer test.

Controlled release, blind tests of DNAPL characterization using partitioning tracers.

The partitioning tracer technique for dense nonaqueous phase liquid (DNAPL) characterization was evaluated in an isolated test cell, in which controlled releases of perchloroethylene (PCE) had occurred. Four partitioning tracer tests were conducted, two using an inverted, double five-spot pumping pattern, and two using vertical circulation wells. Two of the four tests were conducted prior to remedial activities, and two were conducted after. Each test was conducted as a "blind test" where researchers conducting the partitioning tracer tests had no knowledge of the volume, method of release, nor resulting spatial distribution of DNAPL. Multiple partitioning tracers were used in each test, and the DNAPL volume estimates varied significantly within each test based on the different partitioning tracers. The tracers with large partitioning coefficients generally predicted a smaller volume of PCE than that expected based on the actual release volume. However, these predictions were made for low DNAPL saturations (average saturation was approximately 0.003), under conditions near the limits of the method's application. Furthermore, there were several factors that may have hindered prediction accuracy, including tracer degradation and remedial fluid interference.