� .Journal of Contaminant Hydrology 37 1999 159–178 Movement and remediation of trichloroethylene in a saturated heterogeneous porous medium 1. Spill behavior and initial dissolution M. Oostrom a,), C. Hofstee b, R.C. Walker c, J.H. Dane d a En˝ironmental Technology Di˝ision, Pacific Northwest National Laboratory, P.O. Box 999, MS K9-33, Richland, WA, 99352, USA b WAREM, Uni˝ersity of Stuttgart, D-70550, Stuttgart, Germany c Department of Ci˝il Engineering, Auburn Uni˝ersity, AL, 36849, USA d Department of Agronomy and Soils, Auburn Uni˝ersity, AL, 36849, USA Received 12 February 1998; accepted 19 October 1998 Abstract An intermediate-scale flow cell experiment was conducted to study the flow of liquid and the � .transport of dissolved trichloroethylene TCE in a saturated, heterogeneous porous medium system. The 1.67-m long by 1.0-m high by 0.05-m wide flow cell was packed with three layers and five lenses consisting of four different sands. All lenses and layers had horizontal interfaces, except the lowest interface, which was pointed down in the middle. Groundwater flow was imposed by manipulating the water levels in two head chambers. Over 500 ml of dyed TCE was allowed to infiltrate at a constant rate into the porous medium from a narrow source located on the surface. A dual-energy gamma radiation system was used to determine TCE saturations at 1059 locations. Fluid samples were collected from 20 sampling ports to determine dissolved TCE � .concentrations. The TCE migrated downwards in the form of several relatively narrow 3–8 mm fingers. Visual observations and measured TCE saturations indicated that the spilled TCE accumulated on top of, but did not penetrate into, fine-grained sand lenses and layers but that some TCE infiltrated into medium-grained sand lenses. This behavior is a result of the different nonwetting-fluid entry and permeability values of the sands. Most of the TCE finally pooled on top of a fine-grained sand layer located in the bottom part of the flow cell. A multifluid code � .STOMP: subsurface transport over multiple phases , accounting for TCE entrapment, was used to simulate the movement of liquid TCE. Using independently obtained hydraulic parameter values, the code was able to qualitatively predict the observed behavior at the interfaces of the lenses and sand layers. Simulation results suggest that most of the liquid TCE at the lowest interface was in ) Corresponding author. Tel.: q1-509-372-6044; fax: q1-509-372-6089; e-mail:
[email protected] 0169-7722r99r$ - see front matter q 1999 Elsevier Science B.V. All rights reserved. � .PII: S0169-7722 98 00153-3 ( )M. Oostrom et al.rJournal of Contaminant Hydrology 37 1999 159–178160 free, continuous form, while most of the other TCE remaining in the flow cell was entrapped and discontinuous. A simple pool dissolution model was used to predict observed dissolved TCE concentrations. Results show that the measured concentrations could only be predicted with unrealistically high transverse dispersivity values. The observed TCE concentrations are a result of a combination of entrapped and pool dissolution. q 1999 Elsevier Science B.V. All rights reserved. Keywords: DNAPL; Remediation; Dissolution; Trichloroethylene; Plume behavior 1. Introduction � .The presence of dense nonaqueous-phase liquids DNAPLs in the subsurface poses a major environmental problem because these liquids might form a source of widespread and long term contamination of groundwater. Over the last decade a large number of experimental studies on the infiltration and redistribution of DNAPLs in water-saturated �porous media have been conducted e.g., Schwille, 1988; Kueper et al., 1989; Illan- .gasekare et al., 1995; Hofstee et al., 1998 . These studies have demonstrated the importance of subsurface heterogeneity on the migration and final distribution of DNAPLs. Important aspects of the redistribution are unstable behavior of downward DNAPL flow, the nonuniform distribution of entrapped DNAPL and the formation of DNAPL pools on top of fine-grained layers. In addition to experimental studies, several �numerical simulators have been used to predict the movement of DNAPLs e.g., Kueper .and Frind, 1991; Mayer and Miller, 1996; White and Oostrom, 1996 . So far, testing of numerical simulators against infiltration data has been virtually nonexistent. Hofstee et � .al. 1998 concluded that existing continuum-based numerical simulators are unable to predict DNAPL infiltration because of the inability of the existing constitutive theory to account for unstable flow behavior in the form of fingering. Besides studies dealing with liquid DNAPL migration, other research has been conducted to investigate the dissolution and subsequent migration of the organic components as constituents of the aqueous phase. Most efforts in this area were directed towards dissolution mechanisms of residual, entrapped NAPL. Less emphasis has been placed on dissolution of NAPL pools on top of fine-grained deposits. A detailed overview of dissolution mechanisms of entrapped NAPL can be found in the work of � .Mayer and Miller 1996 . Several studies based on one-dimensional column experiments have indicated that rapid equilibrium is attained when water flows through zones with �entrapped NAPL e.g., Fried et al., 1979; Miller et al., 1990; Powers et al., 1991, 1992; .Imhoff et al., 1993 unless the aqueous phase velocities are relatively high or the amount of residual NAPL low. In field experiments, organic compounds are rarely found at concentrations approach- � .ing their solubility limit Mackay et al., 1985; Hunt et al., 1988 . The apparent contradiction between field and laboratory data was investigated experimentally and � .theoretically in a series of papers by Anderson et al. 1992a,b and Johnson and Pankow � .1992 . The authors proposed five possible explanations for the observed low concentra- � . � .tions: 1 reduction of advective transport near source areas, 2 rate limited dissolution, � . � .3 dissolution from pools instead of residual, entrapped NAPL, 4 dispersive mixing, ( )M. Oostrom et al.rJournal of Contaminant Hydrology 37 1999 159–178 161 � .and 5 dilution of thin andror narrow plumes by uncontaminated water in monitoring wells. Their investigations indicated that low concentrations are mainly a result of NAPLs being dissolved from pools on top of fine-grained lenses or layers in combina- tion with plume dilution in monitoring wells. Compared to entrapped NAPLs, pools present very low contact areas to the moving groundwater. � .Schwille 1988 described flow cell experiments in which dissolution of 1,1,1-TCA � . � .tricholoroethane and TCE trichloroethylene pools was studied. After several days of flushing with clean water, only a few percent of the initial amounts of DNAPL present were removed. It was observed that when the flow rate increased, the removal rate � .increased as well. Schwille 1988 acknowledged, however, that the results were not of adequate detail to allow the determination of functional relationships governing pool � . � .dissolution of DNAPLs. Schwille 1988 data were used by Johnson and Pankow 1992 to test an analytical expression for dissolution of idealized rectangular and uniform pools � .proposed by Hunt et al. 1988 . It was concluded that under typical field conditions, � .dissolution from pools is likely to be very slow. Pearce et al. 1994 indicated that dissolved TCE plumes are very sensitive to deviations of the source area from the � .idealized rectangular and uniform shape assumed by Hunt et al. 1988 . Anderson et al. � .1992b used analytical transport models to simulate contaminant plume development from fingers and chlorinated solvent pools. Both fingers and pools were assumed to have regular shapes and the overall contaminant distribution was calculated as the superposi- tion of the contribution of individual dissolution sources. It was concluded that pools are far more persistent sources of contamination than fingers. Other physical model studies � . � .with similar conclusions were reported by Pearce et al. 1994 and Whelan et al. 1994 who used a cooking pan to confine a DNAPL pool at the bottom of their flow container. � .Long term use of TCE at a Department of Energy DOE facility in Paducah, KY, has resulted in the presence of the DNAPL at several locations in the heterogeneous subsurface. Monitoring wells in the vicinity of the spills indicated that significant plumes of dissolved TCE had developed. As part of a clean up effort, site personnel are currently investigating whether surfactants can be used to solubilize and remove the DNAPL. For this effort to be successful, it is critical to determine the possible migration pathways and the final distribution of the spilled TCE. Results of a limited field test, using the surfactant T-MAZ-80, were inconclusive at best, mainly due to engineering � .problems Intera, 1995 and a failure to bring the injected surfactant solution in contact with liquid TCE. Before attempting another remediation demonstration, it was decided that a better understanding of DNAPL infiltration and redistribution in heterogeneous porous media, as well as the subsequent dissolution into moving groundwater, needed to be developed through a detailed, quantitative intermediate-scale laboratory experiment. Although we were not allowed to use actual site materials, the particle sizes and the configuration of the sands used in the study represent the subsurface materials at the Paducah site. In addition, a multifluid code was tested to identify DNAPL flow phenomena that can be successfully simulated using current knowledge of constitutive theory. We also tested the utility of a simple pool dissolution model. The second part of the experiment, which deals with TCE remediation by pump-and-treat and surfactant flushing techniques, is � .described in a companion paper Oostrom et al., 1998b; this issue . ( )M. Oostrom et al.rJournal of Contaminant Hydrology 37 1999 159–178162 2. Materials and methods 2.1. Flow cell experiment � .A flow cell Oostrom et al., 1997; Hofstee et al., 1998 consisting of an inlet head chamber, a porous medium chamber and an outlet head chamber was packed under �water-saturated conditions with layers and lenses of four different sands Table 1; F&S .Abrasives, Birmingham, AL . A schematic of the flow cell with the locations of the layers and lenses is shown in Fig. 1. The internal dimensions of the porous medium chamber were 1.67-m long by 1.0-m high by 0.05-m wide. The water levels in two head chambers, one on either side of the porous medium chamber, were manipulated to establish horizontal discharge rates. In this experiment, flow was from left to right and the discharge rate was set at 2.8 ml miny1, corresponding to an average Darcy velocity y1 � .of approximately 0.15 m day in the coarse-grained sand layer a12; Table 1; Fig. 1 . The imposed Darcy velocity was similar to the one observed in the main aquifer at the Paducah site. The front side of the flow cell consisted of glass to allow flow and transport visualization, while the back side of the cell was made out of Kynar, a translucent, chemically inert and easily machinable plastic. The particle size distributions of the four sands are shown in Fig. 2, indicating that sand a12 is a coarse-grained sand, a14 a medium-grained sand and F75 and F65 are fine-grained sands. The flow cell was filled with three layers. The top, middle and bottom layer consisted of a14, a12, and F65 sand, respectively. The middle layer, with the coarse-grained a12 sand, contained two 0.3-m long and 0.05-m high lenses packed with the fine-grained sand F75 and three 0.2-m long and 0.05-m high lenses packed with Table 1 Porous medium and fluid properties Sand a12 a14 F75 F65 Porous medium properties � .Brooks–Corey TCE entry pressure head, h cm H O 2.3 3.9 19.7 23.6d 2 Brooks–Corey l 4.3 4.7 5.1 5.2 Irreducible water saturation, S 0.11 0.12 0.14 0.15ir maxMaximum residual TCE saturation, S 0.22 0.20 0.19 0.19nr y12 2� .Permeability 10 m 121 72 2.1 1.1 y1 2 2� .Standard deviation 10 m 2.4 1.7 0.19 0.11 Porosity 0.34 0.35 0.41 0.40 Standard deviation 0.012 0.014 0.019 0.016 y3� .Bulk density kg m 1749 1722 1564 1590 Fluid properties y3� .TCE density kg m 1460 � .TCE viscosity Pa s 0.00057 y3� .TCE solubility in water at 208C kg m 1.1 y1� .TCE–water interfacial tension N m 0.0345 ( )M. Oostrom et al.rJournal of Contaminant Hydrology 37 1999 159–178 163 Fig. 1. Configuration of the porous medium, sampling ports, and TCE source in the flow cell. the medium-grained a14 sand. The interface between the middle and the bottom layer � .was pointed downward Fig. 1 . It started at an elevation of zs0.4 m near the end chambers and decreased linearly to a minimum elevation of zs0.35 m in the middle of the container. Before the start of the TCE spill, bulk density and porosity values were measured at a total of 1059 locations with a dual-energy gamma radiation system. The system was � .calibrated using procedures described by Oostrom and Dane 1990 . The average values Fig. 2. Cumulative particle size distribution of the four sands. ( )M. Oostrom et al.rJournal of Contaminant Hydrology 37 1999 159–178164 and standard deviation are listed in Table 1. The bulk densities were assumed to be constant for the remainder of the experiment. The highest density of measurement locations was near the interfaces between different sands. TCE was injected at a constant rate of 0.5 ml miny1 for 19.1 h from a 1-cm wide � .source, located on top of the porous medium at xs58 cm Fig. 1 . The start of the � .injection is indicated as ts0. The TCE Aldrich, Milwaukee, WI used in the y1 � .experiments was dyed with 0.05 g l of the dye Oil Red Aldrich for visualization purposes. The injection was terminated when movement of liquid TCE into the left head chamber was observed. A total of 68 ml of TCE, captured with traps at the bottom, moved into the inlet head chamber, while a total of 505 ml of TCE remained in the porous medium chamber. Two days after termination of the spill, when movement of TCE had virtually ceased, TCE and water saturation values were obtained at all predetermined locations with the gamma scanner according to new procedures outlined � . �by Oostrom et al. 1998a . Prior to the remediation stage of this experiment Oostrom et .al., 1998b; this issue , at ts14 days, another full gamma scan was obtained. The counting time per location was 60 s. To determine the concentration of TCE in the aqueous phase, water samples were � .extracted from 20 ports Fig. 1 . The ports consisted of stainless steel probes with 10 evenly distributed small holes, extending from the Kynar back plate to the glass wall. The extracted water samples were assumed to yield width-averaged TCE concentrations. The samples were analyzed on a Shimadzu GC14 gas chromatograph with a flame � .ionization detector FID , which was coupled to a Restek RTX-1 capillary column. 2.2. Porous medium and fluid properties Hydraulic properties of the four sands are given in Table 1. The listed parameters were determined to allow numerical simulation of the experiment. Brooks and Corey � .1964 main drainage retention parameters and maximum residual TCE saturations for all four sands were obtained using a modified Teflon pressure cell according to � . � .techniques outlined by Lenhard and Parker 1987b and Lenhard 1992 . To determine the TCE entry pressure head, h , the value of the l parameter, and the irreducible waterd saturation, S , the sands were initially packed under water and TCE was then allowed toir maxslowly displace the water. In a separate procedure the maximum residual TCE, S ,nr � .was determined. First, the sands were packed under the nonwetting compared to water fluid TCE. The TCE was then displaced by water to ensure that the water–TCE system was on a main imbibition path. At a capillary pressure of zero, the remaining TCE was assumed to have attained its maximum residual value. The intrinsic permeability of the sands was determined with the constant head method � .Klute and Dirksen, 1986 using 0.4-m long, 0.05-m internal diameter columns. The values for the density, viscosity and solubility of the dye TCE in water were obtained � .from Schwille 1988 . The water-dyed TCE interfacial tension was determined using the � .ring method described by Du Nouy 1919 . Measurements indicated that addition of the¨ dye caused a slight interfacial tension reduction of 0.0015 N my1 compared to a TCE–water system without dye. ( )M. Oostrom et al.rJournal of Contaminant Hydrology 37 1999 159–178 165 2.3. Numerical simulations �The experiment was simulated with the STOMP simulator White and Oostrom, .1996 . This fully implicit integrated finite difference code has been used to simulate a �variety of multifluid systems e.g., Oostrom et al., 1997; Oostrom and Lenhard, 1998; .Schroth et al., 1998 . The two-dimensional nonuniform computational grid was gradu- ally refined near the location of the source and near the porous medium interfaces until simulation results did not change anymore. The smallest cells near the source and interfaces were 0.25=0.25 cm and the final computational domain consisted of 1508 cells. The smallest cells near the source and interfaces were 0.25=0.25 cm. Zero flux boundary conditions were specified for the aqueous phase at the top and the bottom of the domain, while at the left- and right-hand side hydraulic gradient boundaries were assumed. The specified pressures at the left hand side were 20 Pa larger than at the right hand side to yield an average Darcy velocity of 0.15 m dayy1 in the coarse-grained sand layer and a total discharge of 2.8 ml miny1. The 20-Pa pressure difference in the numerical simulation corresponds well to the 2-mm head difference imposed during the flow cell experiment. The TCE was allowed to infiltrate into the saturated sand at a rate y1 �of 0.5 ml min for 1010 min from a source area at the top Neumann boundary .condition . Zero flux boundary conditions were imposed for liquid TCE on all other domain boundaries. A time-step increment factor of 1.25 was used after convergence. The maximum number of Newton iterations was eight, with a convergence factor of 10y6. Upwind interfacial averaging was used for TCE and water relative permeabilities. Harmonic averages were used for all other flux components. The sands used in the experiment were fairly uniform and distinct nonwetting-fluid � .entry pressures were observed during the retention measurements Table 1 . As a result, � . � .saturation S –capillary pressure P equations based on relations described by Brooks � .and Corey 1964 were employed in this simulation. Previous implementations of the STOMP simulator have incorporated fully hysteretic � . � .correlations according to Lenhard and Parker 1987a and Parker and Lenhard 1987 to � .express the relative permeability–saturation–capillary pressure k–S–P relations for � .multifluid systems White et al., 1995 . Experience has shown that this approach requires a considerable experimental effort to obtain all the required parameter values. In addition, this approach limits simulations to computational grids of one or two dimen- sions. The current STOMP formulation uses simplified hysteretic k–S–P relations which account for the effects of NAPL entrapment following procedures outlined by � .Kaluarachchi and Parker 1992 . Their procedures were extended to include modified � . � .Brooks and Corey 1964 relations as listed by White and Oostrom 1998 . The method computes entrapped NAPL saturations as a function of capillary pressure and the minimum water saturation since the occurrence of NAPL at a particular location � .Kaluarachchi and Parker, 1992 : s min1yS 1ySl lS smin y ,S 1� .snt nmin� 5 � 51qR 1yS 1qR 1yS� . � .l l minwhere S is the effective entrapped NAPL saturation, S is the minimum effectivent ls s � .aqueous saturation, S is the apparent aqueous phase saturation S sS qS , and S isl l l nt n ( )M. Oostrom et al.rJournal of Contaminant Hydrology 37 1999 159–178166 � . � .the effective NAPL saturation. The effective saturation S s SyS r 1yS , where Sl ir ir is the actual saturation and S is the irreducible saturation. The Land’s parameter Rir � .Land, 1968 is given by 1 Rs y1 2� . maxSnr maxwhere S is the maximum effective residual NAPL saturation, obtained on the mainnr imbibition branch. Compared to nonhysteretic simulations, this saturation is the only additional parameter to be determined. An overview of experimentally determined maxS -values for the four sands used in the experiment is given in Table 1. The relativenr permeability–saturation relations in the model are based on the on the Burdine pore size � .distribution model Burdine, 1953 , analogous to relations based the Mualem model � . � .Mualem, 1976 , as used by Kaluarachchi and Parker 1992 . ( )2.4. Pool dissolution model Hunt et al., 1988 � .Hunt et al. 1988 proposed a simple dissolution model for pools of NAPLs along aquifer bottoms or floating on the water table. During steady-state water flow, the advection–dispersion equation describing vertical mixing into a semi-infinite medium can be written as E C E 2 C ˝ sD x , z)0 3a� .x z 2E x E z � y1 .where ˝ is the horizontal pore water velocity LT , C is the NAPL concentrationx � y3 . � 2 y1.ML , and D the vertical dispersion coefficient L T . The dispersion coefficientz D st D qa ˝ where t is the tortuosity coefficient, D is the diffusion coefficientz o t x o � 2 y1. � .L T , and a the vertical dispersivity L . Given the boundary conditionst C x , zs‘ s0� . C x , zs0 sC 0FxFL� . 3b� .s C xs0, z s0� . � y3 .where C is the water solubility of the NAPL ML and L is the horizontal pools � .length, the solution of Eq. 3a for 0sxsL is z C x , z sC erfc 4� . � .s 1r2� /2 D xr˝� .z x where erfc is the complimentary error function. 3. Results and discussion 3.1. Liquid TCE mo˝ement Injection of the denser and less viscous TCE, compared to water, caused unstable displacement resulting in the development of fingers. A picture showing the observed ( )M. Oostrom et al.rJournal of Contaminant Hydrology 37 1999 159–178 167 TCE distribution at ts5 h is presented in Fig. 3a. The situation shortly after the termination of the spill at ts24 h is shown in Fig. 3b. Above the left F75 lens, the fingers appeared on both the glass and Kynar side of the flow cell. Although the Kynar is slightly more wetting to TCE than the glass, no distinct preference for this material by the TCE was observed. Between the left F75 lens and the a14 lenses below it, no fingers were observed on either side of the flow cell. Below the a14 lenses, fingers � .appeared again on both sides of the flow cell Fig. 3a,b . The observed finger width varied from approximately 3 to 8 mm. Fingering behavior, occurring when DNAPLs � .move into saturated regions, has also been observed by Schwille 1988 , Oostrom et al. � . � . � . � .1995 , Held and Illangasekare 1995 , and Hofstee et al. 1998 . Schwille 1988, p. 35 showed, through excavation of several glass columns, that preferential vertical move- ment along the walls was not a problem. The TCE initially pooled on top of the left F75 lens. As the spill progressed, TCE moved downward at both sides of the lens. No apparent penetration into the F75 lens was observed. Migrating TCE also accumulated on the two a14 lenses below the F75 lens. However, visual inspection indicated that some of the TCE penetrated into these medium sand lenses. Most of the TCE finally ended up on top of the F65 sand layer � .Fig. 3b . No penetration into this fine sand was observed. During the first 3 days after initiation of the spill, the discharge rate of water slowly decreased to 2.3 ml miny1, while maintaining constant head values in the end chambers. Apparently, the infiltrated TCE had occupied sufficient pore space to alter the rate considerably. After three days, TCE movement had come to a virtual stop and static equilibrium was assumed. The dual-energy gamma radiation system was used to � .measure water and TCE saturations at all measurement locations Fig. 4 . The gamma system measurements indicate that some TCE resided on top of the left F75 lens and the two a14-sand lenses below it, but that most liquid TCE accumulated on top of the F65 sand layer, where measured TCE saturations were as high as 0.84. The measurements support the visual observations that no TCE penetrated the F75 sand lens, but that TCE infiltration occurred in both a14-sand lenses below it. The main reason for this behavior � .is that the a14 sand has a considerable lower TCE entry pressure 3.9 cm H O than the2 � .F75 sand 19.7 cm H O making it easier for the TCE to infiltrate into the a14 sand. In2 addition, the F75 sand has a lower permeability than the a14 sand. Even if the TCE entry pressure at the F75 sand lens interface had been temporarily exceeded, the amount of TCE moving into the lens would have been limited as a result of the low permeability of the lens. In that case, most of the TCE would have been transported laterally on top of the interface because of the considerable larger permeability of the a12 sand. � .The measured saturations Fig. 4 suggest that no significant amounts of liquid TCE were present other than those near the lenses and on top of the F65 sand layer. However, visual observations showed that TCE was also present between the source and the F75 lens, between the F75 lens and the a14-sand lenses below it, and below the a14 sand lenses and the TCE pool on top of the F65 sand. The TCE present in these areas moved in the form of relatively narrow fingers. Due to their random nature, the limited amount of measurement locations, and the fact that gamma radiation measurements represent average values over the width of the flow cell, only minute amounts of TCE were measured. ( )M. Oostrom et al.rJournal of Contaminant Hydrology 37 1999 159–178168 � . � .Fig. 3. TCE distribution at a ts5 h, and b ts24 h. Liquid TCE is dyed red. The interfaces between the a14-sand lenses and the a12 sand are accentuated with a blue marker. ( )M. Oostrom et al.rJournal of Contaminant Hydrology 37 1999 159–178 169 Fig. 4. TCE saturations obtained with the dual-energy gamma radiation system at ts3.5 days. Numerical simulations were conducted to investigate whether the behavior of the TCE near the various interfaces could be predicted. No attempts were made to simulate the observed fingering. It was assumed that the TCE moved downward as a continuous phase. Simulated TCE saturations at ts3.5 days are shown in Fig. 5a. This plot can be � .compared with the measured TCE saturations at approximately the same time Fig. 4 . The simulated saturations indicate that TCE did not penetrate the F75 lens but infiltrated the a14 lenses. No penetration into the F65 sand was predicted. The predicted results qualitatively agree with the experimental observations. The computed maximum satura- tions on top of the lenses agree reasonably well with the experimentally obtained saturations. Both computed and measured maximum saturations were about 0.15 at this � .point in time. In contrast to the measured data, the model predicted small -0.02 but distinct saturations between the source and the left F75 lens, between the left F75 lens and the a14-sand lenses below it, and between the a15-sand lenses and the TCE pool on top of the F65 sand. The code assumes TCE infiltration in the form of a continuous phase, while in reality TCE moved in the form of fingers that were not measured by the dual-energy gamma radiation system. The computed entrapped saturations are shown in Fig. 5b. It is interesting to see that most of the TCE on top of the horizontal lenses is predicted to be in entrapped form. In contrast, all the TCE on top of the F65 sand, having the pointed down interface with the a12 sand layer above, is in free form. To help explain the trapped TCE saturation �distribution in the flow cell, plots of capillary pressure sdifference in phase pressure .between TCE and water , total and entrapped TCE saturation vs. time are shown in Fig. 6 for two locations. In Fig. 6a, the selected location, labeled A in Fig. 1, is 1 cm above the left F75 lens at xs0.58 m, while in Fig. 6b the location, labeled B in Fig. 1, ( )M. Oostrom et al.rJournal of Contaminant Hydrology 37 1999 159–178170 � . � .Fig. 5. Computed a total, and b entrapped TCE saturations at ts3.5 days. � .is 1 cm above the F65 sand layer at xs0.835 m the middle of the flow cell . At � .location A Fig. 6a , the capillary pressure initially increased rapidly when the infiltrat- ing TCE displaced the water. After about 5 h, both the capillary pressure and the total TCE saturation reached a plateau. The S–P relation has only been on the main drainage path and, as a result, there is no entrapped TCE present. Shortly after the influx of TCE ( )M. Oostrom et al.rJournal of Contaminant Hydrology 37 1999 159–178 171 � .Fig. 6. Total TCE saturation, entrapped TCE saturation, and capillary pressure as a function of time at a 1 cm � .above the left F75 lense at xs0.58 m, and b 1 cm above the F65 sand layer at xs0.835 m. was terminated, both the TCE pressure and the capillary pressure decreased and the pool above the interface slowly drained. The S–P relation at this location was now on a primary imbibition path and some of the TCE was trapped, the amounts of which were ( )M. Oostrom et al.rJournal of Contaminant Hydrology 37 1999 159–178172 � .calculated according to Eq. 1 . In approximately 6 days, the total TCE saturation decreased to about 0.15, and more than 75% of this amount was entrapped. For location � .B, the situation was different Fig. 6b . The capillary pressure initially increased rapidly after arrival of the liquid TCE before asymptotically growing to a value of 447 Pa. The system remained on a main drainage path and, as a result, there was no TCE entrapment. The total TCE saturation gradually increased to a value of approximately 0.77. Like for location B, a monotone increase in capillary pressure and TCE saturation is predicted for all nodes in the pool area on top of the F65 layer. For all other nodes where TCE had appeared during the simulation, phenomena similar to those as described for location A occurred, resulting in entrapment of part of the TCE. In general, the simulation results suggest that after a spill, a DNAPL may end up in entrapped andror free form. On top of a pointed down interface, most DNAPL is likely to be free and continuous. On horizontal interfaces, TCE will exist in both free and entrapped form. The form in which the DNAPL exists has a large effect on the distribution of the DNAPL in the pore space � .and therefore on the dissolution mechanisms e.g., Schwille, 1988 . 3.2. TCE dissolution After completion of the spill, a substantial part of the flow cell contained trapped andror free liquid TCE, which caused relatively high TCE concentrations in and downstream of the spill area. Dissolved TCE concentrations were determined for 20 � . � .locations Fig. 1 and the contaminated water exiting the flow cell Fig. 7 . The plot indicates that the outflow concentrations varied between 160 and 340 ppm. These concentrations are well below the chemical equilibrium concentration of around 1100 ppm, which is an indication of considerable movement of fresh water through the TCE � .contaminated areas dilution and, possibly, rate limited mass transfer. In combination with the measured discharge rate, the concentration values in Fig. 7 were used to compute the amount of TCE removed from the flow cell during the first 14 days of the Fig. 7. Measured outflow TCE concentrations. ( )M. Oostrom et al.rJournal of Contaminant Hydrology 37 1999 159–178 173 experiment. These computations yielded the removal of 10.1 g of TCE, which corre- sponds to 6.9 ml or 1.37% of the total volume initially present in the flow cell. Measured concentrations at the extraction ports never reached the chemical equilib- rium concentration, not even at the ports located just downstream from TCE contami- nated areas such as ports 7 and 8. The largest concentrations were observed at ports 13, � .17 and 20 Fig. 8 . These three ports are located just above the pool on top of the F65 � .sand layer Fig. 1 . The concentrations at these ports ranged from about 600 to 800 ppm. To see whether concentrations in this range could be the result of pool dissolution, � � ..computations were conducted using the solution Eq. 4 of the model proposed by � .Hunt et al. 1988 . Examples of these computations for four different transverse dispersivity values are shown in Fig. 9 for port 17. In the computations it was assumed y6 y1 y10 2 y1 � .that xs1.2 m, ˝ s5.2=10 m s , D s7=10 m s Johnson et al., 1989x o � .and ts0.69 Millington and Quirk, 1961 . For a s0.0, molecular diffusion is the onlyt vertical transport process. For the other three cases depicted in Fig. 9, transverse mechanical dispersion dominates molecular diffusion. The approximate vertical distance from port 17 to the top of the pool, estimated from visual observations and dual-energy gamma data, was between 0.06 and 0.08 m. According to Fig. 9, in order to explain the observed concentrations in the range between 600 to 800 ppm from simple pool dissolution, vertical dispersivity values larger than 0.005 m are necessary. It is not likely that the a12 sand has such a large vertical dispersivity. Measurements in similar � .homogeneous sands have yielded values less than 0.001 m Oostrom et al., 1992 . For � .more reasonable dispersivity values -0.001 m , the expected TCE concentration values at port 17 are less than 200 ppm. Computations were also completed for ports 13 and 20, yielding similar results. To explain the observed concentration range, it is obvious that dissolved TCE not only originated from the pool on top of the F65 sand, but also from the contaminated area between the pool and the a14-sand lenses. The � .Fig. 8. Dissolved TCE concentrations ppm vs. time at ports 13, 17 and 20. ( )M. Oostrom et al.rJournal of Contaminant Hydrology 37 1999 159–178174 Fig. 9. Predicted dissolved TCE concentrations for different transverse dispersivity values as a function of � . � .vertical distance from the pool using the Hunt et al. 1988 pool dissolution model Eq. 4 . In the computations, xs1.2 m, ˝ s5.2=10y6 m sy1 , D s7=10y10 m2 sy1 , and t s0.69.x o entrapped TCE in this area apparently was readily available for aqueous dissolution. In a � .companion paper Oostrom et al., 1998b; this issue , we show that the simple pool dissolution model yields good results once the entrapped TCE between the pool and the a14-sand lenses had disappeared. TCE concentration values at ports 1 and 3 in the a12 sand upstream of the TCE contaminated area, remained zero throughout the experiment. Most other ports in this sand showed the behavior demonstrated in Fig. 10 for ports 15 and 16. During the initial breakthrough of TCE, the concentration increased rapidly to values around 600 ppm. At � .Fig. 10. Dissolved TCE concentrations ppm vs. time at ports 15, 16 and 19. ( )M. Oostrom et al.rJournal of Contaminant Hydrology 37 1999 159–178 175 later times, however, the concentration tailed off to about 200 ppm. Other ports, like � .port 19, showed a more irregular behavior Fig. 10 . The large temporal differences in TCE concentrations between ports 15 and 16 on one hand and port 19 on the other, indicate that transverse mixing is small and that the upstream location of entrapped or pooled TCE, in combination with advective patterns, determine the observed TCE concentrations. � .The observed concentration values at the ports in the F65 sand ports 6, 10, 14, 18 gradually increased from zero to between 10 and 15 ppm. The well-known analytic � .solution to the one-dimensional diffusive transport equation Crank, 1956 , employing y10 2 y1 � .diffusion coefficients ranging between 5–9=10 m s Johnson et al., 1989 , yielded final values between 9 and 14 ppm. The agreement between the computed and measured concentrations indicate that TCE transport in the fine F65 sand is governed by molecular diffusion. At ts14 days, another gamma scan was completed. The saturation distribution was similar to Fig. 4, supporting the outflow data that no significant amounts of TCE had been removed from the flow cell. 4. Summary and conclusions An intermediate-scale flow cell experiment was conducted to investigate the migra- tion of liquid TCE and the transport of dissolved TCE in a saturated heterogeneous porous medium. The flow cell was packed with a medium-grained sand layer at the top, a coarse-grained sand layer in the middle, and a fine-grained sand layer at the bottom. A total of five lenses, two fine-grained and three medium-grained, were packed into the coarse-grained layer. Groundwater flow was imposed by manipulating the water levels in the inlet and outlet head chambers. More than 500 ml of dyed TCE was injected into the flow cell at a constant rate of 0.5 ml miny1 from a 1-cm wide source located on the top surface. Visual inspection showed that the TCE initially migrated downwards in the form of � .relatively narrow 3–8 mm fingers. Upon encountering a fine-grained, horizontal, sand lens, accumulation occurred followed by continued downward movement along the edges of the lens. Deeper in the coarse-grained sand layer, the migrating TCE accumu- lated on top of two medium-grained, horizontal, sand lenses before continuing down- wards. Visual observations and measured TCE saturations indicated that some of the spilled TCE infiltrated into these medium-grained sand lenses. This behavior is a result of the different nonwetting-fluid entry and permeability values of the sands. Most of the TCE finally collected on the lowest, pointed down, interface. The STOMP code was used to simulate the movement of liquid TCE in the flow cell. The code accounts for nonwetting-fluid entrapment, in this case entrapment of TCE by water. Using independently obtained values for the hydraulic parameters, the code was able to qualitatively simulate the observed behavior at the interfaces of the layers and lenses. Since the spill was of limited duration, TCE entrapment likely occurred. During the spill, locations where TCE was present were on a main drainage path. However, after termination of the spill, TCE–water capillary pressures at several locations ( )M. Oostrom et al.rJournal of Contaminant Hydrology 37 1999 159–178176 decreased and some of the TCE became entrapped. Simulation results suggest that most of the liquid TCE accumulated on top of the fine-grained, pointed down, sand layer � .F65 was in free, continuous form. Most of the other TCE remaining in the flow cell was entrapped and discontinuous. The form in which the DNAPL resides in the subsurface has a large impact on the distribution in the pore space and, therefore, on the subsequent dissolution. Fluid samples were collected from 20 sampling ports and the outflowing water to � .determine dissolved TCE concentrations. A pool dissolution model by Hunt et al. 1988 was used to predict measured TCE concentrations at ports near the pool on top of the fine-grained sand layer. The observed concentration values were in the 600–800 ppm range for the duration of this part of the experiment. These measured concentration values could only be computed using unrealistically high transverse dispersivity values. Using realistic dispersivity values, predicted concentrations were less than 200 ppm. The observed concentrations can only be explained by a combination of dissolution of entrapped TCE, located upstream, and pool dissolution. TCE concentrations measured at the four ports in the fine-grained sand layer could be explained by molecular diffusion as the dominant transport process. After 14 days, only 1.37% of the TCE spill was removed from the flow cell. Using pump-and-treat and surfactant flushing techniques, we then attempted to remove the � .remainder of the TCE Oostrom et al., 1998b; this issue . Acknowledgements This research was supported by the Subsurface Science Program, Office of Health � .and Environmental Research, US Department of Energy DOE . Pacific Northwest National Laboratory is operated by Battelle Memorial Institute for DOE under Contract DE-AC06-76RLO 1830. References Anderson, M.R., Johnson, R.L., Pankow, J.F., 1992a. Dissolution of dense chlorinated solvents into ground water: 1. Dissolution from a well defined residual source. Ground Water 30, 250–256. Anderson, M.R., Johnson, R.L., Pankow, J.F., 1992b. Dissolution of dense chlorinated solvents into groundwa- ter: 3. Modeling contaminant plumes from fingers and pools of solvent. Environ. Sci. Technol. 26, 901–908. Brooks, R.H., Corey, A.T., 1964. Hydraulic properties of porous media. Hydrol. Pap. 3. Civ. Eng. Dept., Colorado State Univ., Fort Collins. Burdine, N.T., 1953. Relative permeability calculations from pore-size distribution data. Petrol. Trans. Am. Inst. Mining Met. Eng. 198, 71–77. Crank, J., 1956. The Mathematics of Diffusion. Oxford Univ. Press, New York. Du Nouy, P.L., 1919. A new apparatus for measuring surface tension. J. Gen. Physiol. 1, 521–524.¨ Fried, J.J., Muntzer, P., Zilliox, L., 1979. Ground-water pollution by transfer of oil hydrocarbons. Ground Water 17, 586–594. Held, R.J., Illangasekare, T.H., 1995. Fingering of dense nonaqueous phase liquids in porous media: 1. Experimental investigation . Water Resour. Res. 31, 1213–1222. ( )M. Oostrom et al.rJournal of Contaminant Hydrology 37 1999 159–178 177 Hofstee, C., Walker, R.C., Dane, J.H., 1998. Infiltration and redistribution of perchloroethylene in stratified water-saturated porous media. Soil Sci. Soc. Am. J. 62, 15–23. Hunt, J.R., Sitar, N., Udell, K.S., 1988. Nonaqueous phase liquid transport and clean up: 1. Analysis of mechanisms. Water Resour. Res. 24, 1247–1258. Illangasekare, T.H., Ramsey, J.L., Jensen, K.H., Butts, M.B., 1995. Experimental study of movement and distribution of dense organic contaminants in heterogeneous aquifers. J. Contam. Hydrol. 20, 1–25. Imhoff, P.T., Jaffe, P.R., Pinder, G.F., 1993. An experimental study of complete dissolution of a nonaqueous´ phase liquid in saturated porous media. Water Resour. Res. 30, 307–320. Intera, 1995. The in situ decontamination of sand and gravel aquifers by chemically enhanced solubilization of multiple-component DNAPLs with surfactant solutions. Phase 1 Final Topical Report. Intera, Austin, TX. Johnson, R.L., Pankow, J.F., 1992. Dissolution of dense chlorinated solvents into ground water: 2. Source functions for pools of solvent. Environ. Sci. Technol. 26, 896–908. Johnson, R.L., Cherry, J.A., Pankow, J.F., 1989. Diffusive contaminant transport in natural clay: a field example and implications for clay-lined disposal sites. Environ. Sci. Technol. 23, 340–349. Kaluarachchi, J.J., Parker, J.C., 1992. Multiple flow with a simplified model for oil entrapment. Transp. Porous Media 7, 1–14. Klute, A., Dirksen, C., 1986. Hydraulic conductivity and diffusivity: laboratory methods. Methods of Soil Analysis: Part 1. Physical and Mineralogical Methods. Soil Science Society of America, Madison, WI, pp. 687–734. Kueper, B.H., Frind, E.O., 1991. Two-phase flow in heterogeneous porous media: 1. Model development. Water Resour. Res. 27, 1049–1057. Kueper, B.H., Abbott, W., Farquhar, G., 1989. Experimental observations of multiphase flow in heterogeneous porous media. J. Contam. Hydrol. 5, 83–95. Land, C.S., 1968. Calculations of imbibition relative permeability for two- and three-phase flow from rock properties. Trans. Am. Inst. Min. Metall. Petrol. Eng. 207, 149–156. Lenhard, R.J., 1992. Measurement and modeling of three-phase saturation pressure hysteresis. J. Contam. Hydrol. 9, 243–269. Lenhard, R.J., Parker, J.C., 1987a. A model for hysteretic constitutive relations governing multiphase flow: 2. Permeability–saturation relations. Water Resour. Res. 23, 2197–2206. Lenhard, R.J., Parker, J.C., 1987b. Measurement and prediction of saturation pressure relationships in three-phase porous media systems. J. Contam. Hydrol. 1, 407–424. Mackay, D.M., Roberts, P.V., Cherry, J.A., 1985. Transport of organic contaminants in groundwater. Environ. Sci. Technol. 19, 385–392. Mayer, A.S., Miller, C.T., 1996. The influence of mass transfer characteristics and porous media heterogeneity on nonaqueous phase dissolution. Water Resour. Res. 32, 1551–1567. Miller, C.T., Poirier-McNeill, M.M., Mayer, A.S., 1990. Dissolution of trapped nonaqueous phase liquids: mass transfer characteristics. Water Resour. Res. 26, 2783–2796. Millington, R.J., Quirk, J.P., 1961. Permeability of porous solids. Trans. Faraday Soc. 57, 1200–1207. Mualem, Y., 1976. A new model for predicting the hydraulic conductivity of unsaturated porous media. Water Resour. Res. 12, 513–522. Oostrom, M., Dane, J.H., 1990. Calibration and automation of a dual-energy gamma system for applications in soil science. Agronomy and Soils Dept. Ser. 145. Alabama Agric. Exp. Station, Auburn, AL. Oostrom, M., Lenhard, R.J., 1998. Comparison of relative permeability–saturation pressure parametric models for infiltration and redistribution of a light nonaqueous phase liquid in sandy porous media. Adv. Water Resour. 21, 145–157. Oostrom, M., Dane, J.H., Guven, O., Hayworth, J.S., 1992. Experimental investigation of dense solute plumes in an unconfined aquifer model. Water Resour. Res. 28, 2315–2326. Oostrom, M., Hofstee, C., Dane, J.H., 1997. Light nonaqueous-phase liquid movement in a variably saturated sand. Soil Sci. Soc. Am. J. 61, 1547–1554. Oostrom, M., Hofstee, C., Dane, J.H., Lenhard, R.J., 1998a. Single-source gamma radiation procedures for improved calibration and measurements in porous medium systems. Soil Science 163, 646–656. Oostrom, M., Hofstee, C., Walker, R.C., Dane, J.H., 1998b. Movement and remediation of TCE in a heterogeneous porous medium: 2. Pump-and-treat and surfactant flushing techniques, this issue. ( )M. Oostrom et al.rJournal of Contaminant Hydrology 37 1999 159–178178 Oostrom, M., Lenhard, R.J., White, M.D., 1995. Infiltration of dense and light nonaqueous phase liquids in � .partly saturated sand columns. In: Morel-Seytoux H. Ed . Proc. of the 15th American Geophysical Union Hydrology Days. Hydrology Days Publications, Atherton, CA, pp. 215–226. Parker, J.C., Lenhard, R.J., 1987. A model for hysteretic constitutive relations governing multiphase flow: 1. Saturation–pressure relations. Water Resour. Res. 23, 2187–2196. Pearce, A.E., Voudrias, E.A., Whelan, M.P., 1994. Dissolution of TCE and TCA pools in saturated subsurface systems. J. Environ. Eng. 120, 1191–1206. Powers, S.E., Loureiro, C.O., Abriola, L.M., Weber, W.J., 1991. Theoretical study of the significance of nonequilibrium dissolution of nonaqueous phase liquids in subsurface systems. Water Resour. Res. 27, 463–477. Powers, S.E., Abriola, L.M., Weber, W.J., 1992. An experimental investigation of NAPL dissolution in saturated subsurface systems: steady-state mass transfer rates. Water Resour. Res. 28, 2691–2710. Schroth, M.H., Istok, J.D., Selker, J.S., Oostrom, M., White, M.D., 1998. Multifluid flow in bedded porous media: laboratory experiments and numerical simulations. Adv. Water Resour. 22, 169–183. Schwille, F., 1988. Dense Chlorinated Solvents in Porous and Fractured Media. Lewis, Chelsia, MI. Whelan, M.P., Voudrias, E.A., Pearce, A., 1994. DNAPL pool dissolution in saturated porous media—proce- dure development and preliminary results. J. Contam. Hydrol. 15, 223–237. White, M.D., Oostrom, M., 1996. STOMP: Subsurface Transport Over Multiple Phases. User’s Guide. PNNL-11218. Pacific Northwest National Laboratory, Richland, WA. White, M.D., Oostrom, M., 1998. Modeling surfactant-enhanced nonaqueous-phase liquid remediation of porous media. Soil Science 163, 931–940. White, M.D., Oostrom, M., Lenhard, R.J., 1995. Modeling fluid flow and transport in variably saturated porous media with the STOMP simulator: 1. Nonvolatile three-phase model description. Adv. Water Resour. 18, 353–364.