Environ. Sci. Technol. 2008, 42, 5285–5291
Reactive Tracer Tests To Predict Dense Nonaqueous Phase Liquid Dissolution Dynamics in Laboratory Flow Chambers X. CHEN AND J. W. JAWITZ* Soil and Water Science Department, University of Florida, Gainesville, Florida 32611
Received November 27, 2007. Revised manuscript received April 10, 2008. Accepted April 14, 2008.
Reactive tracer tests were conducted to evaluate the relationship between contaminant mass reduction, Rm, and flux reduction, Rj, in laboratory experiments with porous media contaminated with a dense nonaqueous phase liquid (DNAPL). The reduction in groundwater contaminant flux resulting from partial mass removal was obtained from continuous and pulsed cosolvent and surfactant flushing dissolution tests in laboratory flow chambers packed with heterogeneous porous media. Using the streamtubes concept, a Lagrangian analytical solution was applied to study the contaminant dissolution. The analytical solution was independently parametrized using nonreactive and reactive tracer tests and the predicted dissolution was compared to the observed data. Analytical solution parameters related to aquifer hydrodynamic heterogeneities were determined from a nonreactive tracer, while those related to DNAPL spatial distribution heterogeneity were obtained from a reactive tracer. Reactive travel time variance, derived from this combination of tracers, was used to predict the relationship between Rm and Rj. Predictions based on the tracer tests closely matched measured dissolution data, suggesting that tracers can be used to characterize the DNAPL spatial distribution heterogeneity controlling the dissolution behavior. Experimental results demonstrated that increased reactive travel time variance led to greater flux reduction for a given partial mass removal.
Introduction Sites contaminated with dense nonaqueous phase liquids (DNAPLs) present significant challenges for characterization of the contaminant source zone and implementation of riskbased removal technologies. Several field-scale studies have demonstrated that a high percentage (from 70% to 90%) of the DNAPL mass can be removed from source zones by implementing aggressive in situ remediation technologies, such as cosolvent/surfactant flushing or steam flooding (1–3). However, the efficiency of DNAPL mass extraction decreases as the mass is depleted (4, 5). Therefore, it is important to evaluate whether partial mass removal can fulfill the objectives of reductions in risk, source longevity, and site operation and maintenance costs (6). Source zone architecture, which refers to the DNAPL content, geometry, and spatial distribution, has been reported * Corresponding author phone: 352-392-1951 × 203; e-mail:jawitz@ ufl.edu. 10.1021/es7029653 CCC: $40.75
Published on Web 06/18/2008
2008 American Chemical Society
to be an important factor affecting the efficiency of partial mass removal (6). Increased heterogeneity of the velocity field was found to lead to a more favorable relationship between DNAPL mass reduction, Rm, and contaminant flux reduction, Rj (7), where the term flux refers to the total dissolved contaminant mass flux (ML-2 T-1) discharged from the source zone to the dissolved contaminant plume under natural gradient conditions. High-resolution numerical simulations of NAPL dissolution with heterogeneous architectures demonstrated approximately equivalent fractional reductions in contaminant flux for given fractional reductions in DNAPL mass (5). Parker and Park (8) suggested that the Rj(Rm) relationship depends on NAPL distribution, groundwater velocity field, and the correlation between the two and these authors proposed a power function model based on the Damkohler number (9). Ganglia-to-pool (GTP) ratio is another measure of source zone heterogeneity that has been proposed to be correlated to source zone dissolution (10). Suchomel and Pennel (11) found that the GTP ratio decreased with increasing mass removal. Christ et al. (12) established an empirical relationship between the GTP and mass transfer coefficient to replace the upscaled model calibration parameters (8). However, it is unclear how to obtain the GTP or Damkohler number and other such power function model parameters at field sites or in any setting other than a numerical model or laboratory visualization, where extensive or even complete spatial information is available (12). Using the streamtubes concept, Jawitz et al. (13) proposed a Lagrangian analytical solution for the Rj(Rm) relationship, applying the parameters measured from tracer tests. These authors described aquifer hydrodynamic heterogeneities using a distribution of nonreactive travel times obtained from nonpartitioning tracer tests, while DNAPL source zone architecture and spatial distribution heterogeneity were characterized by a distribution of reactive travel times obtained from partitioning tracer tests (14, 15). On the basis of the combination of the statistics of the nonreactive and reactive travel time distributions, a Lagrangian analytical solution for contaminant dissolution was developed to predict the equilibrium dissolution remediation performance in terms of breakthrough curves (BTCs) and Rj(Rm). Only limited data are currently available to validate the Lagrangian analytical streamtube model under heterogeneous conditions. Fure et al. (16) successfully predicted NAPL dissolution behavior using this model in laboratory experiments where the Lagrangian NAPL heterogeneity was estimated from image analysis. However, experimental validation of tracer-based parametrization of the Lagrangian model is needed prior to broader application, such as at field sites. This paper reports on laboratory experiments in twodimensional flow chambers with heterogeneous porous media and NAPL source zones where reactive tracer tests were employed to predict DNAPL dissolution dynamics from one cosolvent (50% ethanol/50% water) and three surfactant (2% Tween-80/water) flushing tests. Both continuous and pulse injection of cosolvent and surfactant solutions were conducted with a range of NAPL architectures to evaluate the ability of tracer-based parametrization of the dissolution prediction model. The pulse flushing tests enabled direct evaluation of the reduction in groundwater contaminant flux resulting from partial mass removal.
Theory A previously described streamtube-based analytical model of contaminant dissolution (13) is summarized below. Also VOL. 42, NO. 14, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
5285
summarized below are how the key parameters that drive this model may be determined from tracer tests using the method of moments (13, 15) and how these parameters can be used to directly predict contaminant BTCs and contaminant flux reduction resulting from a given mass reduction or Rj(Rm). Streamtube Model Summary. The streamtube model represents a porous medium as a collection of noninteracting stream tubes (14, 15). The assumptions are that flow is stable and steady, advection is dominant over other transport processes such as transverse dispersion, and dissolution dynamics is governed primarily by the NAPL architecture, with negligible effects from relative permeability changes during dissolution (13, 16). The contaminant dissolution analytical solution of Jawitz et al. (13) is formulated in terms of the moments of the distribution of reactive travel time, τ, which is defined as the dissolution/flushing duration required to deplete DNAPL from a stream tube, i Si τi ) tiRi ) ti + tiKfk
(1)
where t is the nonreactive travel time, Kf ) FN/Cf, FN is the NAPL density (ML-3), Cf is the contaminant concentration in the flushing solution, and R is the contaminant retardation factor. The trajectory-average saturation along a streamline, k SN (L3 L-3), is defined distinctly from the NAPL saturation at point spatial locations, SN, and the average saturation for the entire domain, Sj . The spatial variability of the trajectoryaverage DNAPL content, k S)k S Nη/θ [where η is the porosity (L3 L-3), and θ is the water content (L3 L-3)], is employed in place of k SN, because it exhibits a range of (0, ∞) rather than (0, l) and is consistent with log-normal and γ probability distributions (15). In the reactive travel time formulation of eq 1, the time required for dissolution increases with NAPL content and decreases as flushing solution solubility increases.Usingthisframeworkandassumptions,theLagrangianbased NAPL architecture statistics can be evaluated from tracer tests using higher moment expressions (15), as discussed in the following section. Temporal Moment Expressions. The first two moments of the reactive travel time distribution, mτ1 and mτ2, can be expressed in terms of the moments of the t and S k distributions (15): k
k
mτ1 ) mt1 + Kfmt1m1S γ m2S k
(2)
k
k
mτ2 ) mt2(1 + 2Kfm1S γ2 + Kf2m2S γ4) m1S
[
t0 k k np S np S mp1 ) (1 - φ)mnp 1 + φ m1 + KNm1 m1 γ - m1 KN 2 k
]
(4)
k
np S 2 2 S 4 mp2 ) (1 - φ)mnp 2 + φ (1 + 2KNm1 γ + KN m2 γ )m2 -
t
4
k S 2 0 (KNm1kSt0γ + KN2m2kSt0γ2)mnp 1 - m1 KN 2
]
(5)
where t0 is the tracer pulse injection duration, KN is the DNAPL-water partitioning coefficient, and φ is the fraction of contaminated streamtubes. When the NAPL architecture is the primary factor controlling dissolution behavior and 5286
9
µln X ) 2 ln mX1 - 0.5 ln mX2
(6)
mX1 0.5
(7)
σln X ) (ln
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 42, NO. 14, 2008
mX2 - 2
ln
)
Equilibrium Dissolution. Under equilibrium conditions, the flux-averaged contaminant concentration exiting the source zone can be expressed as function of flushing duration T (13): Cf(T) ) Cw(1 - fQ,t
