Steven L. Bryant
Contamination of groundwater by NAPLs is a widespread problem. Typically, however, the location of NAPL contaminant sources is very poorly known, and characterization of a site is expensive. Design of remediation projects is greatly compromised by this uncertainty. For example, a scheme designed to remove uniformly distributed low saturations of NAPL may be very inefficient if the NAPL is concentrated in a few regions of high saturation.
Water soluble tracers that interact with an organic phase, for example dissolving into it (partitioning tracers) or concentrating at the interface between phases (interfacial tracers), have been used to infer the volume of a NAPL contaminant in an aquifer. The fundamental question addressed by this project was how can we extract more information from partitioning interwell tracer tests? In particular, can the shape of the tracer concentration history measured at an extraction well help delineate the spatial location of the NAPL? The answer depends in part on the answer to the following question: Can the influence of spatial distribution of NAPL on partitioning tracer histories be distinguished from the influences of unrelated phenomena such as mass transfer kinetics, dispersion, etc.? We conducted numerical simulations and bench-scale experiments to address these questions.
Preliminary proof of concept simulations were conducted in homogeneous model aquifers (constant permeability throughout) with a single injector/monitor well pair. Distributing NAPL heterogeneously in the model aquifer induced characteristic "humps" in the simulated concentration history at the monitor well, and these features became more apparent as the partitioning coefficient (measure of the solubility of the tracer in the NAPL) increased. One way to understand this effect is to observe that the partitioning causes a wide range of residence times for the tracer in the aquifer. Tracer concentrations migrating along streamlines that do not pass through any NAPL are not retarded at all, whereas streamlines that encounter large local NAPL saturations exhibit significant retardation. By comparison a nonpartitioning tracer experiences a relatively narrow range of residence times, because the streamlines between injector and monitor well do not vary widely in a homogeneous aquifer.
The focus of the numerical simulation work was on partitioning tracer behavior in a typical field situation, modeled closely upon the Hill Air Force Base Operable Unit 2 (OU2). The disposal of spent chlorinated degreasing solvents in unlined trenches at OU2 from 1967 to 1975 introduced a multicomponent DNAPL composed primarily of trichloroethene (TCE), other chlorinated solvents, and oil and grease into the shallow alluvial aquifer. Hence, pooling of the DNAPL beneath the disposal trenches occurred within the alluvium above the Alpine clay aquiclude. A large volume of residual DNAPL remains trapped by capillary forces in the pore spaces of the aquifer. For the current investigation of high-resolution PITTs, a three-dimensional 40x17x20 grid was developed for the forward flow and transport modeling. The simulated source zone is 125 ft long, 50 ft wide, and 14 ft thick including both sand and clay cells. The wellfield consists of one injection well and two extraction wells, one to the north and one to the south of the injection well, and several monitoring wells. Many geostatistical realizations of the permeability field were generated to test the influence of permeability heterogeneity, the primary control on NAPL distribution during contamination.
We simulated partitioning tracer tests in these realizations to determine the limits of this approach for pinpointing spatial distributions of NAPL. The results confirm that bypassing (flow around regions of low permeability, instead of through them) influences the overall tracer history. When examined in more detail, the histories for the most strongly retarded tracer also exemplify the fundamental thesis of this research: the shape of the history shows clear evidence of the heterogeneous distribution of the NAPL. For example, a 'shoulder' in one concentration history was a direct consequence of the relatively small volume of relatively high saturation of NAPL in the bottom center of the domain. However, the results also indicate that unlike the proof-of-concept cases, it was relatively rare for tracers to exhibit obvious manifestations of spatial heterogeneity. More critically, heterogeneity in the distribution of permeability within the aquifer has the same qualitative effect on tracer concentration histories as heterogeneity in the spatial location of NAPL. Thus in order to infer NAPL locations in the field, one would have to independently determine the variation in permeability with location in the aquifer. Some indication of this variation can be obtained from the behavior of a nonpartitioning tracer. The level of detail needed to distinguish the contributions of the two types of heterogeneity is unlikely to be generally available. Moreover, the permeability heterogeneities may be correlated with the heterogeneity in NAPL saturations; in the case of OU2, the NAPL tends to be at the bottom of the aquifer, which is also the lowest permeability zone in the aquifer. Thus it does not appear to be feasible to reliably infer NAPL locations only from partitioning tracer tests. Substantial additional site characterization would generally be needed in order to enable the inference of NAPL locations.
A more subtle indication of heterogeneity in NAPL distribution was detected, however. The indication was a consistent deviation from straight-line semilog behavior in the "tail" of the tracer concentration history at the monitoring well. This tail contributes to the overall mass of NAPL inferred from a tracer test. The customary approach of extrapolating the tail as exponential decay, once concentrations fall below the detection limit, can result in errors of a few percent NAPL volume. This finding has obvious practical implications since NAPL volumes will be underestimated. In order to carry out the extrapolation correctly in practice, one would need an a priori basis for determining the deviation from straight-line semilog behavior. Although no such basis could be ascertained from this study, the size of the error in estimated NAPL volume was correlated with the degree of heterogeneity of the permeability field.
A standard assumption in modeling tracer tests is that the aqueous and NAPL phases are in local chemical equilibrium. Depending on the flow rate, the NAPL saturation and the ratio of surface area to volume of the NAPL phase, this assumption may not be valid. The variation of interfacial area with NAPL saturation is of particular concern, but very little data is available. By leveraging other sources of support, we also conducted a series of experiments with interfacial tracers (compounds that adsorb to the NAPL/water interface) in order to better define the possible range of interfacial areas during contamination events (NAPL spill) and remediation (injection of aqueous phase). The measurements were compared with recent theoretical advances in predicting fluid configurations within porous media obtained within our group in other projects.
Flow experiments were conducted in glass columns packed with cleaned glass beads. The glass beads were cleaned using 0.1 M HCl and 15% H2O2 in order to remove surface impurities such as metal ions and to oxidize the organic matter. The glass beads were then thoroughly washed with water to remove the residual acid and peroxide, and then oven-dried before storing them in a clean beaker for use in column experiments. The glass beads were made oil-wet using silane and chloroform.
Water was used as the aqueous phase and decane was the organic phase. Experiments were conducted with decane loaded in different ways (e.g. water loaded first vs decane loaded first) in columns packed with water-wet or oil-wet glass beads. Potassium iodide (KI), and pure 3-phenyl decyl benzene sulfonate (C10 3-φ LAS), an anionic surfactant were used as conservative and interfacial tracers, respectively. C10 3-φ LAS is a 10 carbon, linear alkyl benzenesulphonate, with the benzene ring at the third carbon position. It adsorbs at the water-decane interface but does not dissolve into the decane. The sorbed concentration is proportional to the area of the water-decane interface. Observing the degree of retardation of the surfactant concentration front in the column effluent (relative to a passive tracer) therefore gives a measure of the interfacial area.
The interfacial area tracer experiments were conducted at variable flow rates, including stop-and-restart tests, and in all cases there was no evidence of non-equilibrium sorption.
The most remarkable result from the experiments was the new evidence of how thin films of wetting phase on soil/rock grains influence measurements of interfacial area. Related theoretical work led by the PI under separate EPA funding recently showed that simple fluid/soil systems should exhibit a maximum in interfacial area between macroscopic volumes of the wetting and nonwetting phases as the saturation of nonwetting phase increases. Measurements reported in the literature show monotonic increases in interfacial area, however. We obtained similar results in water-wet systems (glass beads). When the system wettability is reversed, i.e. when we conduct the same experiment in a column of oil-wet beads, so that the nonwetting phase is water, the experiments showed qualitatively different behavior: the interfacial area exhibited a maximum when plotted as a function of saturation.
Insights obtained in our group in other projects proved crucial to understanding this result. We recently obtained quantitative theoretical estimates of the contribution to interfacial area from films of wetting phase on grains of the porous medium. The film contribution varies monotonically with fluid saturation (phase volume fraction), while the contribution from discrete, bulk volumes of fluids exhibits a maximum. The latter contribution is much smaller than the film contribution. The measurements in the water-wet system are consistent with thin films providing the dominant contribution to interfacial area. In contrast, we expect no contribution from films in the oil-wet system, because decane does not spread on the oil-wet beads. Therefore we expected to see only the contribution from bulk volumes of water and decane in the oil-wet experiments, and the measurements were consistent with this expectation. This was the first experimental confirmation of the theoretical prediction.
The implications of these findings are important. For example, interphase mass transfer is the mechanism for groundwater contamination by nonaqueous phase liquids and for several methods of remediation. Depending on the process of interest, however, this mass transfer may occur only through part of the area measured by the interfacial tracer technique. Such measurements should therefore be used with caution in interpreting field and laboratory observations of mass transfer rates.
Gladkikh, M., V. Jain, S. Bryant and M. Sharma. "Experimental and Theoretical Basis for a Wettability-Interfacial Area-Relative Permeability Relationship," SPE 84544 to be presented at 2003 Ann. Tech. Conf. Exhib. Soc. Pet. Eng., Denver, 8 Oct. 2003.
Bryant, S. and M. Gladkikh. "Predicting Realistic Fluid Configurations in Porous Media and Their Influence on Petrophysical Properties," invited presentation at Baker Atlas Research, Houston, 1 Oct. 2003
Bryant, S. and M. Gladkikh. "Predicting Realistic Fluid Configurations in Porous Media and Their Influence on Petrophysical Properties," presented at Annual Research Review of the Center for Excellence in Formation Evaluation, The University of Texas at Austin, 22 Aug. 2003.
Gladkikh, M., V. Jain, S. Bryant and M. Sharma. "Thermodynamic Analysis of Interfacial Tracer Measurements," submitted to Adv. Wat. Res., July 2003.
Gladkikh, M. and S. Bryant. "Prediction of interfacial areas during imbibition in simple porous media," Adv. Water Res. 26:609-622, 2003.
Bryant, S. and A. Johnson. "Wetting phase connectivity and irreducible saturation in simple granular media," J. Coll. Interfac. Sci. 263(2):572-579, 2003.
Bryant, S. and A. Johnson. "Bulk and Film Contributions to Fluid/Fluid Interfacial Area in Granular Media," in press, Chem. Eng. Comm.
Jain, V., S. Bryant, M. Sharma and A. Johnson. "Sensitivity of Interfacial Tracers to Liquid/Liquid/Solid Interfaces in Granular Media," ACS Colloids and Surface Science Symp., 2002.
Bryant, S. "Contributions to Fluid/Fluid Interfacial Area in Granular Media," Developments in Water Science, 47: Proc. XIV Intl. Conf. on Computational Methods in Water Resources, 177-184, 2002.
Jain, V., S. Bryant, and M. Sharma. "Liquid-liquid interfacial area: influence of wettability and fluid saturation," presented at AGU 2002 Spring Meeting, Washington DC, 28-31 May 2002.
Jain, V., S. Bryant and M. Sharma. "Influence of Wettability and Saturation on Liquid-Liquid Interfacial Area in Porous Media," Environ. Sci. Technol. 2003; 37(3) pp 584-591.
Jayanti, S. "Modeling tracers and contaminant flux in heterogeneous aquifers," Ph.D. dissertation, The University of Texas at Austin, 2003.
Steven L. Bryant
Center for Petroleum and Geosystems Engineering
1 University Station C0304
The University of Texas at Austin
Austin, Texas 78712-0228
Phone: (512) 471-3250 FAX: (512) 471-9605