Fluid and Porous Media Property Effects on Dense ... - ACS Publications

The effects of fluid and porous media properties on dense nonaqueous phase liquid (DNAPL) migration and associated contaminant mass flux generation we...
0 downloads 0 Views 140KB Size
Environ. Sci. Technol. 2007, 41, 1622-1627

Fluid and Porous Media Property Effects on Dense Nonaqueous Phase Liquid Migration and Contaminant Mass Flux C . T . T O T T E N , † M . D . A N N A B L E , * ,† J.W. JAWITZ,‡ AND J.J. DELFINO† Department of Environmental Engineering Sciences, and Soil and Water Science Department, University of Florida, Gainesville, Florida, 32611

The effects of fluid and porous media properties on dense nonaqueous phase liquid (DNAPL) migration and associated contaminant mass flux generation were evaluated. Relationships between DNAPL mass and solute mass flux were generated by measuring steady-state mass flux following stepwise injection of perchloroethylene (PCE) into flow chambers packed with homogeneous porous media. The effects of fluid properties including density and interfacial tension (IFT), and media properties including grain size and wettability were evaluated by varying the density contrast and interfacial tension properties between PCE and water, and by varying the porous media mean grain diameter and wettability characteristics. Contaminant mass flux was found to increase as grain size decreased, suggesting enhanced lateral and vertical DNAPL spreading with higher fluid entry pressure. Mass flux showed a slight increase as the DNAPL approached neutral buoyancy, likely due to enhanced vertical spreading above the injection point. DNAPL spatial distribution and contaminant mass flux were only minimally affected by IFT and by intermediatelevel wettability changes, but were dramatically affected by wettability reversal. The relationship between DNAPL loading and flux generation became more linear as grain size decreased and density contrast between fluids decreased. These results imply that capillary flow characteristics of the porous media and fluid properties will control mass flux generation from source zones.

Introduction Understanding of how dense nonaqueous phase liquids (DNAPLs) behave in the subsurface is important to manage contaminated sites and to propose appropriate remedies for source zones. Several field-scale studies have been completed that demonstrate source zone remedial techniques can remove a large portion of the contaminant mass present (1-11). In most of these studies, a common performance metric has been total mass fraction or volume fraction of the initial contaminant removed. However, this metric may only provide partial assessment of remediation effectiveness. Contaminant flux, J [ML-2T-1], is an emerging metric for characterizing contaminant source zones (12, 13), and the * Corresponding author phone: 352-392-3294; fax: 352-392-3076; e-mail: [email protected]. † Department of Environmental Engineering Sciences. ‡ Soil and Water Science Department. 1622

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 41, NO. 5, 2007

relationship between contaminant mass flux and the mass of contaminant present has only recently been explored (1417). To date, investigations exploring this relationship have not systematically assessed the significance of fluid and media properties (18, 19). The spatial distribution of DNAPL in an aquifer can be considered the source zone architecture. The contaminant mass flux that is generated by the groundwater flow field through the DNAPL architecture is a direct function of the transverse area of exposure between DNAPL and mobile groundwater. This interfacial area is a function of media properties including grain size and wettability, and fluid properties including DNAPL/water density contrast and interfacial tension. The effect of these properties on DNAPL architecture and the resulting groundwater contaminant mass flux in a homogeneous sand pack is the focus of this research. Fluid Properties. The distribution of DNAPL following a spill is affected by aqueous phase/DNAPL density differential (Fw - Fo) and interfacial tension (20). As a chlorinated solvent migrates through the aquifer it is held up by capillary retention. The capillary retention is directly proportional to the interfacial tension and inversely proportional to radius of curvature (7). As interfacial tension increases and pore size decreases, a larger head is required to produce downward migration. This may result in a lateral movement of liquid along layers of finer grain sands controlling the DNAPL distribution. The aqueous phase/DNAPL density differential generally causes DNAPL to migrate downward. When encountering a capillary barrier, DNAPL can move laterally, sometimes even counter to the existing hydraulic gradient depending on impedance by lateral capillary forces. Lateral DNAPL spreading will occur along pathways of least capillary and permeability resistance (21). As the density difference between DNAPL and water decreases (i.e., approaches neutral buoyancy), the gravitational driving force promoting vertical migration decreases, potentially leading to increased lateral migration. Common DNAPL density contrasts with water vary up to 60%, with only a 1% difference needed to influence fluid movement (21). Porous Media Properties. Porous media grain size is an important parameter that affects water flow and capillary pressure acting on a DNAPL globule in the subsurface. In fine-grain porous media, higher capillary entry head can inhibit movement. Additionally, water flow is lower due to reduced hydraulic conductivity impacting the mobility of DNAPL. Trapping theory indicates that intrinsic permeability strongly influences migration (22). In horizontal flow systems where groundwater flow is perpendicular to gravitational forces, the pressure gradient is usually small compared to gravitational forces when determining migration behavior in the vertical direction. This pressure is a function of DNAPL spill rate and fluid viscosity. Porous media wettability also can affect DNAPL behavior in porous media (23). Wettability is the relative affinity of the solid for fluids such as air, water, or oil. Hydrophilic soils have an affinity for water while hydrophobic soils have an affinity for oil. Naturally water-wet soils can also become oil-wet through physical, chemical, or biological transformations. Cationic surfactants and additives in gasoline have been shown to cause water-wet material to become oil-wet (23-25). Additionally, Powers et al. (24) found that complex DNAPL mixtures can create a wide range of wetting conditions. Finally, control of media wettability may offer the potential to enhance DNAPL extraction from the subsurface. 10.1021/es061639r CCC: $37.00

 2007 American Chemical Society Published on Web 01/23/2007

Fractional Mass Flux versus Fractional Mass Loading. The significance of DNAPL source zone transverse area to contaminant flux generation can be conceptualized by considering two hypothetical source zones with similar mass releases that migrate differently with respect to the groundwater flow field (a conceptual figure is provided in the Supporting Information (SI)). As mass is added, one source zone expands in the same direction as groundwater flow while the other expands transverse to groundwater flow. For the same mass of DNAPL released, mass flux generation increases at a greater rate for the source zone expanding transverse to the flow direction. This source geometry intercepts a greater proportion of the flow field resulting in a greater interfacial area generating higher contaminant mass flux. The relationship between DNAPL mass loading and resulting mass flux can be used to characterize source zones. Conceptually, this is similar to relationships explored for mass reduction and resulting flux reduction in DNAPL source zones (14, 16, 17). Data measured to characterize this relationship can be fit using the following empirical model:

RJ ) RβM

(1)

where RJ represents a scaled source mass flux, RM is fractional mass increase, and β is a fitting coefficient. This model is similar to that proposed by Rao et al. (13) for mass depletion/ flux reduction relationships in source zones. The model, as applied in this work, is a retrospective look at the flux and mass relationships for a given system since they are normalized to the maximum observed mass flux values.

Materials and Methods General Experimental Procedure. Relationships between DNAPL mass and down-gradient solute mass flux were measured by injecting PCE in 0.5 cm3 increments into a twodimensional flow chamber packed with water-saturated porous media under static conditions (all chemical sources and purities are provided in the SI). The pore volume was approximately 200 cm3. Down-gradient mass flux was measured under steady-state water flow following stabilization of the PCE concentration. This procedure was repeated until about 5 cm3 of PCE had been introduced into the porous media. At each injection step, the red-dyed PCE distribution was traced from the side of the chamber to maintain a qualitative record of the PCE geometry. The experimentally measured contaminant concentrations were compared to predicted values assuming both equilibrium and rate-limited dissolution. A grid was overlain on the flow domain forming horizontal stream tubes, and the amount of DNAPL contacted in a given streamtube (or grid row) was tabulated in a manner similar to that described by Fure et al. (18). The equilibrium solute concentration was estimated as simply the product of the PCE solubility limit and the fraction of the streamtubes exposed to PCE. Rate-limited dissolution was accounted for using the modified Sherwood number (Sh′) of Powers et al. (26) for each streamtube. Experimental Apparatus. Fifteen laboratory experiments were conducted in a 14 cm × 28 cm two-dimensional flow chamber similar to that described by Jawitz et al. (27) (photo provided in SI). For most experiments, the chamber was wetpacked homogeneously with Accusand (28) (Table 1) and covered with a layer of bentonite clay to minimize volatile losses and to allow pressurized flow. For the experiments investigating wettability effects, the polarity of the sand surface was reversed from water-wet to oil-wet by treating the Accusand with octadecyl trichlorosilane (OTS) (23, 29). Accusand was rotated in a solution of methanol and 5% OTS for 5 h, followed by a methanol rinse and air drying. Untreated sand and OTS-treated sand were mixed in various ratios to

TABLE 1. PCE/n-Decane Mixture Densities, Media Sieve and Grain Sizes, and Best Fit β Values for Mass Loading/Flux Generation Relationships Measured in These Media DNAPL density (g/cm3) 1.6 1.6 1.6 1.6 1.6 1.4 1.4 1.1 1.1 1.0

sieve size

mean grain size (mm)

β ( R 2)

20/30 30/40 40/50 40/60 50/70 30/40 40/50 30/40 40/50 30/40

0.68 0.48 0.35 0.32 0.23 0.48 0.35 0.48 0.35 0.48

0.41 (0.97) 0.54 (0.81) 0.93 (0.97) 0.26 (0.91) 0.43 (0.88) 0.66 (0.96) 0.83 (0.99) 0.88 (0.97) 1.0 (0.99) 0.92 (0.98)

produce degrees of oil/water wettability. To eliminate trapped air in the experiments with OTS-modified sand, both the sand and bentonite were first dry-packed. The bentonite was hydrated from above to form a seal over the dry sand, and the sand was purged with CO2 and then water flooded. Steady-state water flow through the chamber (approximately 0.7 cm3/min) was controlled using a constant head reservoir at the inlet and a constant head outlet. Based on a porosity of 0.38, the porewater velocity was maintained at approximately 1.8 m/day. Effluent samples were analyzed by gas chromatography with a flame ionization detector (Perkin-Elmer Autosystems). Fluids and Fluid Properties. PCE colored with Oil-red-O dye (e1 × 10-4 M) was injected into the center of the chamber 6.5 cm above the bottom and approximately 4 cm below the top of the sand layer. In a separate pooling experiment, the injection port was situated 1 cm above the bottom of the chamber. In all experiments, PCE was injected through a 20-gauge stainless steel needle epoxy-sealed into the glass wall. Stepwise injections of 0.5 cm3 PCE were conducted at 0.1 cm3/min for 5 min using a syringe pump (Harvard Apparatus) and 10 cm3 syringe (Hamilton Gastight). The PCE/water interfacial tension was modified by adding 0.0025, 0.005, 0.01, 0.025, 0.05, and 0.1% by volume of the surfactant Span 80 to PCE. Span 80 (sorbitan monooleate) was selected because of its low hydrophile/lipophile balance (HLB) value of 4.3. A low HLB surfactant was required to limit partitioning of the surfactant into the aqueous phase. The effect of surfactant concentration on PCE/water IFT was measured using a tensiometer (Tensiomat model 21, Fisher Scientific: accuracy ( 0.25%). The DNAPL/water density difference was modified by using mixtures of PCE and n-decane (F ) 0.73 g/cm3). The low aqueous solubility of n-decane (0.003 mg/L) ensured limited partitioning from the nonaqueous to the aqueous phase. The IFT of the 50/50 mole fraction of PCE and n-decane mixture (52 dynes/cm) was found to be similar to that of PCE/water (47 dynes/cm), thus density was the primary property being varied in these experiments.

Results and Discussion Fluid Property Contrasts. Interfacial Tension. Experiments were conducted with DNAPL/water IFT values of 3, 13, and 47 dynes/cm (Span 80 concentrations of 0.05, 0.025, and 0 volume percent) in 30/40 sand packs (IFT-concentration plot provided in (SI)).In this range of Span 80 concentrations, PCE solubility increased approximately 6% (30). The general pattern of DNAPL migration after initial release at the injection point was rapid migration to the base of the flow chamber followed by lateral spreading, producing a zone of residually trapped DNAPL between the injection point and the flow chamber base, and a pool across the base (Figure VOL. 41, NO. 5, 2007 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

1623

FIGURE 2. Fractional flux increase versus DNAPL loading for each interfacial tension value in untreated 30/40 sand.

FIGURE 1. Tracings of observed distributions of DNAPL injected at H ) 6.5 cm above the flow chamber bottom in 30/40 sand with (a) unmodified PCE and untreated sand, (b) PCE and 50% OTS treated hydrophobic sand mixture, (c) PCE and 100% hydrophobic sand, and (d) untreated sand and DNAPL mixture of density 1.1 g/cm3 (symbol + represents injection location). 1a). After each incremental release, the DNAPL distribution extent was traced and the numbers provided on the figure panels represent the cumulative number of 0.5 cm3 PCE injections. Based on observations of the PCE distribution in the flow chamber, the extent of DNAPL lateral spreading increased slightly as IFT decreased (30). Pooling began by the second injection in each experiment, indicating that DNAPL vertical migration in the 30/40 sand was not significantly altered by reducing the IFT, and was therefore controlled by fluid density contrast. The mass flux generation with PCE loading is compared for these experiments in Figure 2. As the IFT decreased, the effluent relative concentrations increased slightly. The maximum relative concentration appeared to be bounded by the injection point location since DNAPL did not migrate above the injection point. Because the media above the injection point (located at 0.6 of the aquifer height) was uncontaminated, the effluent relative concentration would not be expected to exceed this value (assuming DNAPL did not significantly alter the aqueous flow field). The mass flux generation data were fit with eq 1, resulting in β ) {0.54,0.50,0.67} with R 2 ) {0.80,0.95,0.98} for the 47, 13, and 3 dyn/cm systems, respectively. The entire data set was fit in Figure 2 with an average β ) 0.5 (R 2 ) 0.82). These results indicate that over the range of IFTs tested here, relatively minor changes (β ) 0.5-0.67) were observed in the characteristics of the mass flux generation with PCE 1624

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 41, NO. 5, 2007

FIGURE 3. Fractional flux increase versus PCE loading for varying density and media. [β ) 0.54 is lowest observed and β ) 0.91 is average of 40/50 sands]. loading. Interfacial tensions lower than those tested here may be required to affect migration or prevent entrapment, which could significantly reduce mass flux. Fluid Density. Fluid density, modified using different ratios of PCE and n-decane, indicated that as density difference decreased, vertical migration was reduced and lateral migration was enhanced. This trend is illustrated in Figure 1a and d for pure PCE and the F ) 1.1 g/cm3 system. In experiments with PCE, pooling usually began by the second injection (1.0 cm3), indicating complete vertical migration from the injection port to the bottom of the chamber. The lower density experiments required as many as nine injections (4.5 cm3) to cause full migration from the injection port to the bottom of the experimental chamber and in the case of the F ) 1.0 g/cm3 experiment, no pooling occurred, indicating no full downward migration. The PCE relative concentrations at final PCE loading were similar for all experiments, with the lower density tests generally exhibiting slightly higher values (Figure 3). These results are likely due to the observed migration of DNAPL above and below the injection port (Figure 1d), providing a larger cross sectional area exposure to flow. The β values fit for each DNAPL density experiment are provided in Table 1. The mass loading/flux generation relationship became more linear as the density contrast decreased. The trend was particularly evident in the 30/40 sand. The shape of these curves, coupled with the DNAPL tracings, indicates that reduced vertical migration following each 0.5 cm3 injection altered the mass flux generation relationship. Rapid downward migration, typically for higher density contrasts (i.e., pure PCE) resulted in rapid mass flux

FIGURE 4. Fractional mass flux increase versus PCE loading for various media grain sizes.

FIGURE 5. Fraction mass flux increase versus PCE loading for varying hydrophobic sand percentages.

increases at low DNAPL loading when compared to the slower downward migration associated with lower fluid density contrasts. How mass becomes distributed as ganglia and pools, and how uniformly distributed those domains are, define the initial system dissolution characteristics. Media Contrasts. Grain Size. The effect of porous media grain size (range 0.23 to 0.68 mm) on PCE distribution was evaluated. In each experiment, pooling began after the second injection (1 cm3 total volume injected), with the exception of the 40/50 (0.35 mm) sieve experiment, in which pooling appeared after the fourth injection (2 cm3). The general character of the DNAPL distribution at the end of each experiment consisted of a residual trail of DNAPL ganglia located in the migration path between the injection point and the pool at the bottom of the chamber. The pool consisted of higher saturation DNAPL in a shallow zone (typically less than 1 cm high) that extended laterally toward the injection and extraction wells. The quantitative comparison of DNAPL volume loaded versus scaled mass flux for each grain size experiment along with the pooling experiment is shown in Figure 4. The mass flux response to DNAPL loading indicates an initial rapid increase followed by a plateau. The first several injections provided the largest contribution to mass flux with the later injections having a reduced impact. The earliest injections likely generated the residual or ganglia portion of the source zone with the later injections primarily contributing mass to the pool. This is supported by the results of the pooling experiment, which showed that pooling contributes significantly less mass flux per unit volume at full loading (C/Cmax) 0.05, vs 0.37 for residual). This is consistent with the gangliato-pool concept proposed to characterize DNAPL systems (15). The 20/30 (0.68 mm) sieve sand had the lowest final C/Cmax value of 0.27 and the 40/50 (0.35 mm) sieve sand had the highest final C/Cmax value of 0.38. Under ideal conditions, the final C/Cmax would be 0.60 if the DNAPL extended across the flow chamber height (to the injection point) and depth. The observed values were likely lower than this ideal maximum because migration did not produce homogeneously distributed PCE across the thickness of the chamber between the injection point and the pool. Nonequilibrium mass transfer could also have contributed to the lower observed concentrations. There was no clear pattern between grain size and flux values, but there was some indication that as grain size decreased, flux increased. No overall trend was evident in the fit β values with mean grain size (Table 1). For the more uniform sands, 20/30, 30/ 40, and 40/50, β increased (becoming more linear) as grain size decreased. Similar to density contrast changes, this may be a result of increased lateral spreading as mean grain diameter decreases. This trend changed significantly for the

media with a wider distribution of grain sizes (i.e., 40/60 and 50/70 sands). The wider range of differential grain sizes present in these mixtures may have led to slight layering during the wet packing process used in these experiments. While not visibly evident, very subtle layering can have significant effects on DNAPL behavior. Wettability. Mixtures of 0, 50, and 75% hydrophobic sand displayed a fingered DNAPL distribution showing little effect of media hydrophobicity. In contrast, the 100% hydrophobic mixture was dominated by DNAPL capillary “wicking”. Based on these observations, it was concluded that a major fraction of the media (g75%) must be hydrophobic to allow capillary forces to overcome gravitational forces and cause uniform lateral and vertical spreading. A quantitative comparison assessing mass flux generation is provided in Figure 5. The 100% hydrophobic mixture produced a significantly higher final maximum mass flux value when compared to the intermediate and 0% mixtures. When combined with the traced DNAPL distributions (Figure 1b and c), it appears that the vertical movement of PCE above the injection point was a factor in flux generation in these experiments. There was little difference in flux values between the 0%, 50%, and 75% mixtures. This may be due to the intermediate mixtures providing overall surface area contact between the PCE and aqueous phase similar to that of the untreated sand. Porous media of intermediate wettability have been shown to have lower water imbibing tendency than strongly water wet systems (31). In general, the shape of the flux generation curves for all the media were similar (Figure 5), with β ) {0.27,0.31,0.44}, R 2){0.92,0.93,0.97} for the 50, 75, and 100% hydrophobic fractions, respectively. All of the media with hydrophobic sands had slightly lower β values than the untreated sand. This may indicate that the presence of some fraction of hydrophobic grains in the mixture influenced how the DNAPL initially distributed during loading, but the final mass fluxes generated were very similar (approximately C/Cmax ) 0.4). For 100% hydrophobic sand, the mass flux generation was quite different, producing a very different DNAPL distribution (Figure 1c) and a much higher mass flux C/Cmax ) 0.75. The observed behavior for the hydrophobic sand was due to capillary wetting forces countering gravitational forces, causing slower vertical downward migration and thus reducing the rapid mass flux increase observed in water wet systems. These results suggest that altering media wettability prior to introducing DNAPL can dramatically modify the distribution of DNAPL within an aquifer and the resulting mass flux generation. Flow Bypassing and Rate-Limited Mass Transfer. The mass load and mass flux relationships measured in the laboratory experiments were compared to predictions from VOL. 41, NO. 5, 2007 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

1625

Acknowledgments This study was funded by SERDP, which is a collaborative effort involving the U.S. EPA, U.S. DOE, and U.S. DOD. This document has not been subjected to peer review within these agencies and the conclusions stated here do not necessarily reflect official views of these agencies, nor does the document constitute an official endorsement by these agencies.

Supporting Information Available Additional experimental information. This material is available free of charge via the Internet at http://pubs.acs.org.

Literature Cited

FIGURE 6. Comparison between experimental data, equilibrium and nonequilibrium dissolution flux curves for (a) untreated and (b) 100% hydrophobic 20/30 sand. equilibrium and nonequilibrium dissolution models based on DNAPL contact lengths and Sherwood number correlations applied within 0.5 cm horizontal streamtubes through the flow domain (Figure 6). For the 20/30 sand (Figure 6a), substantial differences were observed between the equilibrium and nonequilibrium predictions. The short length of the source zone in the groundwater flow direction may not have allowed sufficient contact time for equilibrium to have been achieved. In contrast, the longer source zones generated in the 100% hydrophobic sand (Figure 6b) were likely sufficient to achieve equilibrium. However, in both cases the measured relative concentrations were lower than those predicted, even under nonequilibrium conditions. The model observations suggest that flow bypassing likely occurred in both experiments with the hydrophobic sand results comparing more favorably than the untreated sand. The hydrophobic media likely had less bypassing due to the more-uniform distribution of the PCE resulting from capillary wicking. The simple model applied here does capture the general shape of the flux generation mass loading relationships, suggesting that the distribution observed on the chamber wall reflects the general distribution of the DNAPL in the chamber. Application of the mass transfer correlations developed for homogeneously contaminated systems to more complex flow and contaminant distributions may be problematic given the lack of knowledge of water flow paths and DNAPL distributions within those flow paths. The predicted mass flux generation characteristics could be improved with laboratory characterization techniques that account for the DNAPL distribution across the thickness of the flow chamber. While this may provide greater understanding in laboratory systems, field scale assessment may have to rely on tracer approaches to characterize these relationships (32). 1626

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 41, NO. 5, 2007

(1) Brooks, M. C.; Annable, M. D.; Rao, P. S. C.; Hatfield, K.; Jawitz, J. W.; Wise, W. R.; Wood, A. L.; Enfield, C. G. Controlled release, blind test of DNAPL remediation by ethanol flushing. J. Contam. Hydrol. 2004, 69 (3-4), 281-297. (2) Falta, R. W.; Lee, C. M.; Brame, S. E.; Roeder, E.; Coates, J. T.; Wright, C.; Wood, A. L.; Enfield, C. G. Field test of high molecular weight alcohol flushing for subsurface nonaqueous phase liquid remediation. Water Resour. Res. 1999, 35 (7), 2095-2108. (3) Fiorenza, S. DNAPL Removal: Surfactants, Foams, and Microemulsions; Lewis Publishers: Boca Raton, FL, 2000. (4) Fountain, J. C.; Starr, R. C.; Middleton, T.; Beikirch, M.; Taylor, C.; Hodge, D. A controlled field test of surfactant-enhanced aquifer remediation. Ground Water 1996, 34 (5), 910-916. (5) Jawitz, J. W.; Annable, M. D.; Rao, P. S. C.; Rhue, R. D. Field implementation of a Winsor type I surfactant/alcohol mixture for in situ solubilization of a complex LNAPL as a single phase microemulsion. Environ. Sci. Technol. 1998, 32 (4), 523-530. (6) Jawitz, J. W.; Sillan, R. K.; Annable, M. D.; Rao, P. S. C.; Warner, K. In situ alcohol flushing of a DNAPL source zone at a dry cleaner site. Environ. Sci. Technol. 2000, 34 (17), 3722-3729. (7) Lowe, D. F.; Oubre, C. L.; Ward, C. H. Surfactants and Cosolvents for DNAPL Remediation; CRC Press: Boca Raton, FL, 1999. (8) Martel, R.; Gelinas, P. J.; Saumure, L. Aquifer washing by micellar solutions: 3 field tests at the Thouin Sand Pit (L’Assomption, Quebec, Canada). J. Contam. Hydrol. 1998, 30 (1-2), 33-48. (9) Meinardus, H. W.; Dwarakanath, V.; Ewing, J.; Hirasaki, G. J.; Jackson, R. E.; Jin, M.; Ginn, J. S.; Londergan, J. T.; Miller, C. A.; Pope, G. A. Performance assessment of NAPL remediation in heterogeneous alluvium. J. Contam. Hydrol. 2002, 54 (3-4), 173-193. (10) Rao, P. S. C.; Annable, M. D.; Sillan, R. K.; Dai, D. P.; Hatfield, K.; Graham, W. D.; Wood, A. L.; Enfield, C. G. Field-scale evaluation of in situ cosolvent flushing for enhanced aquifer remediation. Water Resour. Res. 1997, 33 (12), 2673-2686. (11) McCray, J. E.; Brusseau, M. L. Cyclodextrin-enhanced in situ flushing of multiple-component immiscible organic liquid contamination at the field scale: Mass removal effectiveness. Environ. Sci. Technol. 1998, 32 (9), 1285-1293. (12) Einarson, M. D.; Mackay, D. M. Predicting impacts of groundwater contamination. Environ. Sci. Technol. 2001, 35 (3), 66A73A. (13) Rao, P. S. C.; Jawitz, J. W.; Enfield, C. G.; Falta, R. W.; Annable, M. D.; Wood., A. L. Technology integration for contaminated site remediation: Cleanup goals and performance criteria. In Groundwater Quality 2001: Natural and Enhanced Restoration of Groundwater Pollution; Thornton, S., Oswald, S., Eds.; International Association of Hydrological Sciences Publication No. 273: Paris, 2001; pp 571-578. (14) Jawitz, J. W.; Fure, A. D.; Demmy, G. G.; Berglund, S.; Rao, P. S. C. Groundwater contaminant flux reduction resulting from nonaqueous phase liquid mass reduction. Water Resour. Res. 2005, 41 (10), Art. No. W10408. (15) Lemke, L. D.; Abriola, L. M.; Lang, J. R. Influence of hydraulic property correlation on predicted dense nonaqueous phase liquid source zone architecture, mass recovery and contaminant flux. Water Resour. Res. 2004, 40 (12), Art. No. W12417. (16) Parker, J. C.; Park, E. Modeling field-scale dense nonaqueous phase liquid dissolution kinetics in heterogeneous aquifers. Water Resour. Res. 2004, 40 (5), Art. No. W05109. (17) Rao, P. S. C.; Jawitz, J. W. Comment on “Steady state mass transfer from single-component dense nonaqueous phase liquids in uniform flow fields” by T.C. Sale and D.B. McWhorter. Water Resour. Res. 2003, 39 (3), Art. No. 1068. (18) Fure, A. D.; Jawitz, J. W.; Annable, M. D. DNAPL source depletion: Linking architecture and flux response. J. Contam. Hydrol. 2006, 85 (3-4), 118-140.

(19) Newman, M.; Hatfield, K.; Hayworth, J.; Rao, P. S. C.; Stauffer, T. A hybrid method for inverse characterization of subsurface contaminant flux. J. Contam. Hydrol. 2005, 81 (1-4), 34-62. (20) Pankow, J. F.; Cherry, J. A. Dense Chlorinated Solvents and other DNAPLs in Groundwater; Waterloo Press: Portland, OR, 1996. (21) Mercer, J. W.; Cohen, R. M. A review of immiscible fluids in the subsurface: Properties, models, characterization, and remediation. J. Contam. Hydrol. 1990, 6, 107-163. (22) Pennell, K. D.; Pope, G. A.; Abriola, L. M. Influence of viscous and buoyancy forces on the mobilization of residual tetrachloroethylene during surfactant flushing. Environ. Sci. Technol. 1996, 30 (4), 1328-1335. (23) Bradford, S. A.; Leij, F. J. Wettability Effects on Scaling 2-Fluid and 3-Fluid Capillary Pressure-Saturation Relations. Environ. Sci. Technol. 1995, 29 (6), 1446-1455. (24) Powers, S. E.; Anckner, W. H.; Seacord, T. F. Wettability of NAPLcontaminated sands. J. Environ. Eng.-ASCE 1996, 122 (10), 889896. (25) Powers, S. E.; Tamblin, M. E. Wettability of Porous-Media after Exposure to Synthetic Gasolines. J. Contam. Hydrol. 1995, 19 (2), 105-125. (26) Powers, S. E.; Abriola, L. M.; Weber, W. J. An Experimental Investigation of Nonaqueous Phase Liquid Dissolution in Saturated Subsurface Systems - Steady-State Mass-Transfer Rates. Water Resour. Res. 1992, 28 (10), 2691-2705.

(27) Jawitz, J. W.; Annable, M. D.; Rao, P. S. C. Miscible fluid displacement stability in unconfined porous media: Twodimensional flow experiments and simulations. J. Contam. Hydrol. 1998, 31 (3-4), 211-230. (28) Schroth, M. H.; Ahearn, S. J.; Selker, J. S.; Istok, J. D. Characterization of miller-similar silica sands for laboratory hydrologic studies. Soil Sci. Soc. Am. J. 1996, 60 (5), 1331-1339. (29) Legrange, J. D.; Markham, J. L.; Kurkjian, C. R. Effects of Surface Hydration on the Deposition of Silane Monolayers on Silica. Langmuir 1993, 9 (7), 1749-1753. (30) Totten, C. T. Effect of porous media and fluid properties on dense non-aqueous phase liquid migration and dissolution mass flux. Ph.D. dissertation, University of Florida, 2005. (31) Moore, T. F.; Slobod, R. L. The effect of viscosity and capillarity on the displacement of oil by water. Producers Monthly 1956 (Aug), 20-30. (32) Jawitz, J. W.; Annable, M. D.; Demmy, G. G.; Rao, P. S. C. Estimating nonaqueous phase liquid spatial variability using partitioning tracer higher temporal moments. Water Resour. Res. 2003, 39 (7).

Received for review July 10, 2006. Revised manuscript received November 28, 2006. Accepted December 4, 2006. ES061639R

VOL. 41, NO. 5, 2007 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

1627