- Email: [email protected]
Numerical modeling of oxygen exclusion experiments of anaerobic bioventing
Journal of Contaminant Hydrology 58 (2002) 209 – 220 www.elsevier.com/locate/jconhyd
Numerical modeling of oxygen exclusion experiments of anaerobic bioventing Philip G. Mihopoulos a,*,1, Makram T. Suidan a, Gregory D. Sayles b, Sebastien Kaskassian a a
Department of Civil and Environmental Engineering, University of Cincinnati, Cincinnati, OH 45221-0071, USA b U.S. Environmental Protection Agency, National Risk Management Research Laboratory, Cincinnati, OH 45268, USA Received 15 December 2000; received in revised form 8 January 2002; accepted 12 April 2002
Abstract A numerical and experimental study of transport phenomena underlying anaerobic bioventing (ABV) is presented. Understanding oxygen exclusion patterns in vadose zone environments is important in designing an ABV process for bioremediation of soil contaminated with chlorinated solvents. In particular, the establishment of an anaerobic zone of influence by nitrogen injection in the vadose zone is investigated. Oxygen exclusion experiments are performed in a pilot scale flow cell (2 1.1 0.1 m) using different venting flows and two different outflow boundary conditions (open and partially covered). Injection gas velocities are varied from 0.25 10 3 to 1.0 10 3 cm/ s and are correlated with the ABV radius of influence. Numerical simulations are used to predict the collected experimental data. In general, reasonable agreement is found between observed and predicted oxygen concentrations. Use of impervious covers can significantly reduce the volume of forcing gas used, where an increase in oxygen exclusion efficiency is consistent with a decrease in the outflow area above the injection well. D 2002 Elsevier Science B.V. All rights reserved. Keywords: Anaerobic; Bioventing; Modeling; Oxygen exclusion
* Corresponding author. Present address: Tait Environmental Management, Inc., 701 North Parkcenter Drive, Santa Ana, CA 92705, USA. Fax: +1-714-560-8235. E-mail addresses: [email protected] (P.G. Mihopoulos), [email protected] (M.T. Suidan), [email protected] (G.D. Sayles), [email protected] (S. Kaskassian). 1 Fax: + 1-513-556-2599.
0169-7722/02/$ - see front matter D 2002 Elsevier Science B.V. All rights reserved. PII: S 0 1 6 9 - 7 7 2 2 ( 0 2 ) 0 0 0 3 7 - 2
210
P.G. Mihopoulos et al. / Journal of Contaminant Hydrology 58 (2002) 209–220
1. Introduction Bioventing has been used extensively in recent years to biologically remediate unsaturated soils contaminated with petroleum hydrocarbons, which have been introduced in the subsurface from leaking underground storage tanks, surface spills and improper disposal methods. Anaerobic bioventing is a new technology intended to treat highly chlorinated compounds in the vadose zone (Sayles et al., 1997; Mihopoulos et al., 2000; Shah et al., 2001). While it employs the basic principles of aerobic bioventing (EPA Manual, 1995; Van Eyk, 1997; Dupont et al., 1993), it is applied to reduce, under anaerobic environments, compounds that are recalcitrant under aerobic conditions. In experimental studies, a gaseous mixture of nitrogen, hydrogen and carbon dioxide, injected in columns representing unsaturated zone profiles contaminated with PCE, was shown to reduce PCE concentrations of 10 ppmv with a half-life of 7 min. Also, complete remediation of PCE was shown in a series of anaerobic –aerobic columns. (Mihopoulos et al., 2001). Anaerobic bioventing is applied by injecting a mixture of nitrogen containing small concentrations of carbon dioxide and hydrogen (well below the lower flammable limit of 4%). The chlorinated vapors are reduced, utilizing hydrogen as an electron donor, under induced anaerobic conditions, while the partially reduced byproducts can be reduced further, or oxidized in the presence of increased oxygen levels as the plume rises to the surface. In a field application, the goal is to maximize biodegradation, by minimizing advection of the contaminants in the gas phase, thus, allowing sufficient residence times for both the anaerobic and the subsequent aerobic step. ABV can be particularly attractive for a big family of chlorinated contaminants that are recalcitrant under aerobic conditions, where a two-step anaerobic – aerobic remediation scheme appears to be the best alternative. The limiting step in developing an engineered bioprocess for complete destruction of chlorinated compounds appears to be a rapid and reliable anaerobic phase. ABV combines principles of diffusive and convective gas transport, interphase mass transfer and reaction kinetics. Portions of free and dissolved nonaqueous phase liquids can partition in the gas phase when in direct contact with the advected gases (N2, H2 and CO2). Typical injection rates for aerobic bioventing are in the order of 0.1 cm/s (EPA Manual, 1995), while gas velocities for soil vapor extraction (SVE) applications are usually greater than 1.5 cm/s in the vicinity of the injection well. At those velocities, advective and dispersive are the dominant transport and mixing mechanisms. The flows required for ABV should be considerably less than those used for SVE and aerobic bioventing, with diffusive fluxes expected to dominate advective fluxes. A key characteristic of ABV, in contrast to aerobic bioventing, is the requirement of an anaerobic zone in the vicinity of the contaminant plume, since aerobic zones are naturally present in the upper layers of the vadose zone. The purpose of this work is to investigate the establishment of an anaerobic environment by forcing nitrogen gas through the vadose zone. The paper is organized as follows. First, we report on visualization venting experiments conducted in a 2-D flow cell. Using nitrogen as a driving gas, the displacement of the initially present oxygen in the porous medium is studied. We monitor the variation in the concentrations of oxygen with time and space, and with different injection flowrates. Subsequently, numerical simulations of the process are performed. The numerical program used is calibrated using data from the oxygen displacement experi-
P.G. Mihopoulos et al. / Journal of Contaminant Hydrology 58 (2002) 209–220
211
ments. The solutions of the numerical model are then compared to independent experimental data. Sensitivity studies are conducted to analyze the dependence of radius of influence on the injected flows.
2. Experimental setup Soil in the vadose zone was experimentally simulated with a flow cell, shown schematically in Fig. 1. The cell dimensions were 2 m long by 1.1 m high by 0.01 m wide. The front side (2 1.1 m) consisted of glass, while the backside of the cell was made of stainless steel with 220 monitoring/sampling ports of 1/4-in. Swagelok fittings. The influent nitrogen stream was passed through an oxygen trap containing a solution of 100 mg/l Na2SO3 and 25 mg/l CoCl26H2O. The injection port, designed to simulate a screened injection well, was 10 10 cm in dimension, located 30 cm above the left bottom corner of the flow cell, while a woven stainless steel mesh (opening 10 Am) was placed in direct contact with the sand. The nitrogen flowrate was regulated with a mass flow controller (MKS Instruments, Andover, MA) as shown in Fig. 1. A high-pressure cylinder served as the source of UHP Grade Nitrogen (Matheson Gas Products, Chicago, IL). Gas samples were collected at various sampling ports with Pressure-Lok gas tight syringes (Dynatech Baton Rouge, LA) and analyzed immediately. Oxygen gas analysis was performed on a Model 5890 Series II Hewlett-Packard (Mountain View, CA) gas chromatograph, equipped with a thermal conductivity detector. The carrier gas (argon) flow was 30 mL/min, and retention times for O2 and N2 were 0.61 and 0.72 min, respectively, on a 10-ft 45/60 Molecular Sieve column (Hewlett Packard). The oven temperature was maintained isothermal at 50 jC.
Fig. 1. Experimental setup.
212
P.G. Mihopoulos et al. / Journal of Contaminant Hydrology 58 (2002) 209–220
3. Porous media and gas properties The sand used in this study (ASTM #D-1556) was obtained from ELE International Soiltest (Lake Buff, IL). The sand had a bulk density of 1.49 g/cm3, and a porosity of 0.35. The probability distribution was determined with sieve analysis, where 93% by mass was retained between US Sieve nos. 70 and 50 (geometric mean size (210297)0.5=250 Am). The value of the diffusion coefficient D for the gas pair nitrogen – oxygen is 1.811.10 5 m2 s 1 (Cussler, 1984). The effective diffusion coefficient, De, is usually reduced by an order of magnitude due to tortuosity and constrictivity effects as: De=hD/a2, where h is the porosity and a is the actual pore length per distance in the direction of diffusion (Geankoplis, 1972). Lumping those effects, and using the formula of Millington and Quirk (1961), De=D/s, where the tortuosity for dry porous media was approximated as s=h 2/3, we obtained a value for tortuosity of 2, which is in the typical range for sandy soils. Vadose zone remediation by ABV depends primarily on vapor phase flowrates that influence the transport and removal of gas-phase contaminants. Thus, it is critical to study processes affecting gas transport and removal at gas velocities applicable to the field. During SVE, gas velocities greater than 1.6 cm/s in the vicinity of the extraction well have
Fig. 2. Oxygen concentration contours for 100 ml/min. Subplots [a], [b], [c] and [d] represent experimental results for 0.5, 1, 2 and 5 days of venting, respectively. Subplot [e] represents numerical results after 5 days of venting.
P.G. Mihopoulos et al. / Journal of Contaminant Hydrology 58 (2002) 209–220
213
been reported from field studies (Gibson et al., 1993). In our case, an inlet flowrate of 100 mL/min, corresponded to an interstitial velocity of 0.007 cm/s in the vicinity of the injection well. Popovica and Brusseau (1997), concluded that hydrodynamic dispersion dominates molecular diffusion at gas velocities greater than 1 cm/s, while at gas velocities lower than 0.3 cm/s molecular diffusion is the major source of spreading. At particle Reynolds numbers above 10, the dispersion coefficient varies linearly with flow, whereas at low flowrates dispersion becomes a function of the Schmidt number (kinematic viscosity/diffusion coefficient) because both flow and molecular diffusion are important in this region (Cussler, 1984). In addition to advective and diffusive fluxes of the injected gas, soil gas flow is shown to be induced by cyclic barometric pressure fluctuations when pressure differentials are bigger than 2 –3 kPa (Massmann and Farrier, 1993). Although those effects might be considered at specific field applications (storm events), they were not significant in this study and were neglected.
4. Two-dimensional flow experiments Three different venting flows of 100, 200 and 500 ml/min were investigated. The twenty-five sampling ports are shown in the mesh diagram of Fig. 2a and in Fig. 1. Initial conditions were atmospheric concentrations of oxygen. To obtain uniform initial con-
Fig. 3. Partially covered experimental [a] and numerical [b] values of oxygen concentrations for Q = 100 mL/min at steady state.
214
P.G. Mihopoulos et al. / Journal of Contaminant Hydrology 58 (2002) 209–220
ditions in all experiments, a period of 7 days was allowed between consecutive experiments. The use of impermeable covers was investigated as a mean of increasing in situ oxygen exclusion efficiencies, by partially blocking the top left outflow boundary with a 0.75-m long impermeable plastic material, as shown in Fig. 7. All experiments were conducted at room temperature (21 –23 jC) and pressure, while all chromatographic analysis was completed within an hour of every sampling event.
5. Numerical simulations The computational domain was two-dimensional (200 110 cm) and consisted of an injection area placed in the bottom left side of the flow cell to simulate the experimental setup as shown in Fig. 1. The numerical code FEHM (Zyvoloski et al., 1997) was used to simulate gas movement in the unsaturated zone. This finite element 3-D code can simulate multiphase and multicomponent transport of air, water and oil in porous media, and has been widely used and tested in a variety of studies. The two-dimensional grid used was uniformly refined from 200 to 1800 quadrilateral elements (spacing 10.16 and 3.4 cm, respectively) and was then continuously refined near the injection source until uniform results were obtained from consecutive grid refinements. The nonuniform grid divided the y-axis in three sections. The midsection, corresponding to the injection screening height, had a uniform spacing of 0.7 cm after the final refinement. The top and bottom sections had elements that increased proportionally to a weighing factor with increasing distance from the injection point. The x-axis was also refined using cells that increased in size with increasing distance from the injection point. In the current study, only vapor transport was considered while the liquid phase was passive. Mass transport of multiple nonreactive solutes through an unsaturated porous medium in 2-D was solved in FEHM using the Galerkin finite element formulation. The discretized nonlinear set of equations were solved using the Newton– Raphson iterative procedure. Semiautomatic step control was used in macros ctrl and trac, with a tolerance of 10 6 and a time step multiplier of 1.5. Zero flux boundary conditions were specified at the lateral and bottom boundaries of the domain. At the top outflow boundary, the concentrations of oxygen and nitrogen were fixed at atmospheric values. At the nitrogen injection point, the injection flowrates were equally distributed in 15 lateral nodes, using a Neumann boundary, while a zero flux boundary was imposed at the top impermeable section. A time-step increment of 1.5 was used after convergence. The maximum number of Newton iterations was 20, with a convergence factor of 10 6. The upwind averaging factor was set at 1.2. Values of porosity, permeability and diffusion coefficients were assumed uniform throughout the domain. An initial pressure field of 101.33 kPa was assumed, and a temperature of 23 jC. The porous medium permeability to air flow was approximated by the Carman – Kozeny equation (due to the homogeneity of the used silica sand particle size, and the assumption that the average particle shape does not deviate strongly from a spherical shape), as 100 Da (10 10 m2). Practically, gas permeabilities higher than 0.1 Da are adequate for applications of aerobic bioventing (EPA Manual, 1995), where for permeabilities below 0.01 Da gas flow is primarily through the secondary porosity and field testing is required to establish feasibility. The small flowrates required for an ABV
P.G. Mihopoulos et al. / Journal of Contaminant Hydrology 58 (2002) 209–220
215
application, make the process attractive for sites with small permeability, which are typically encountered in clay formations. For permeabilities lower than 0.001 Da (Sleep, 1998), contributions to molecular diffusion from gas – wall interactions (Knudsen diffusion) become significant, and a correction factor should be introduced to account for this effect or the dusty gas model (Mason and Malinauskas, 1983) should be used instead.
6. Results and discussion 6.1. Open boundary experiments Figs. 2 –4 summarize transient experimental and steady state numerical results, for flows of 100, 200 and 500 ml/min, respectively. Oxygen contours from experimental values are plotted for t = 12 h, 1, 2 and 5 days of venting operation as shown in subplots [a] through [d]. Subplots [e] of Figs. 2 – 4 show the numerical results obtained for 5 days of nitrogen injection for the corresponding flow. From comparison of the experimental contours [a] –[d], it was concluded that steady state operation was achieved after one day of venting for every flow examined. The numerical results obtained at 5 days (Figs. 2e –
Fig. 4. Oxygen concentration contours for 200 mL/min. Subplots [a], [b], [c] and [d] represent experimental results for 0.5, 1, 2 and 5 days of venting, respectively. Subplot [e] represents numerical results after 5 days of venting.
216
P.G. Mihopoulos et al. / Journal of Contaminant Hydrology 58 (2002) 209–220
Fig. 5. Oxygen concentration contours for 500 mL/min. Subplots [a], [b], [c] and [d] represent experimental results for 0.5, 1, 2 and 5 days of venting, respectively. Subplot [e] represents numerical results after 5 days of venting.
4e) were in close agreement with the corresponding experimental values (Figs. 2d– 4d), and could qualitative predict the shape and extent of the anaerobic radius of influence. For the injection rate of 200 mL/min, the transient behavior of the pumping and recovery phases were examined in a different experiment. Starting from atmospheric conditions, a flow of 200 mL/min was injected, and oxygen levels were monitored continuously until steady state. Subsequently, the nitrogen flow was shut off and oxygen was allowed to diffuse back from the open boundary (recovery phase). The oxygen levels during these two phases were monitored at four specific locations (A through D as shown in Fig. 1) in the flow cell, and the experimental results are shown with discrete points in Fig. 5. The corresponding predicted values from the numerical model are shown with continuous lines in Fig. 5, where close agreement between numerical and experimental values is observed. Effects of longitudinal dispersion could be safely neglected since the injection velocities were sufficiently small. For the experiments under consideration, the diffusive time scaled with L2/De f 5 10 3 s, where L is a characteristic length, here equal with the mean grain diameter, and convective time scaled with L/U f 10 s, where U is the average gas velocity, making molecular diffusion the predominant transport mechanism. With an effective diffusivity of 0.9 10 6 m2/s, and assuming a dispersivity value in the
P.G. Mihopoulos et al. / Journal of Contaminant Hydrology 58 (2002) 209–220
217
Fig. 6. Predicted and experimental values of transient oxygen concentrations for a flow of 200 mL/min.
order of the grain size, the ratio of De/au scales as O (104 –105) at flowrates ranging from 100 to 500 mL/min. The differences between simulations accounting for, and neglecting, longitudinal dispersion were negligible and supported the above assumption. Oostrom et al. (1999), using a similar experimental setup, estimated as reasonable dispersivity values less than 0.001 m. A final experiment was conducted with the open boundary, where the steady state profile (after 5 days of operation) was obtained for four additional nitrogen injection rates of 1, 10, 50 and 1000 mL/min, respectively. For every additional injection rate, the 25 sampling ports were analyzed for oxygen after 5 days of venting. The final oxygen concentrations (normalized by the ambient atmospheric oxygen concentration) where averaged and plotted as a function of the specific flowrate in Fig. 6. 6.2. Partially covered vadose zone Venting experiments and simulations have been conducted with the top outflow boundary partially (75%) covered. An injection rate of 100 mL/min was applied and experimental steady state results are shown in Fig. 7a. The corresponding numerical results were in reasonable agreement with the experimental values as shown in Fig. 7b. Those results indicate that impermeable covers can be used to improve the oxygen exclusion efficiency approximately six times, compared with an open outflow boundary, while this number is expected to increase with the use of lower venting flows. By comparing Figs. 7 and 2e, injecting a flow of 100 mL/min with a partially covered outflow boundary, can yield similar results with that of a 500 mL/min flow with an open
218
P.G. Mihopoulos et al. / Journal of Contaminant Hydrology 58 (2002) 209–220
Fig. 7. Residual oxygen concentration as a function of the injection flowrate.
outflow boundary. The numerical model was used to examine the effect of this specific boundary condition under different injection flowrates. This is shown in Fig. 8, where the area percentage with oxygen values less than 1% is plotted as a function of venting flow for both totally and partially open outflow boundaries. Increasing the nitrogen venting
Fig. 8. Comparison of anaerobic (oxygen level less than 1%) area percentage between open and partially covered outflow boundaries.
P.G. Mihopoulos et al. / Journal of Contaminant Hydrology 58 (2002) 209–220
219
Fig. 9. Predicted and experimental values of transient oxygen concentrations for a flow of 200 mL/min and a partially covered outflow.
flow, beyond 1000 mL/min (open) and 300 mL/min (covered) does not correspond to an increase in the anaerobic zone, since back-diffusion of atmospheric oxygen at the top layers dominates convective flow that originates from the injection section. Also, a pumping and recovery phase experiment was conducted, similar with the one described for the open boundary. The results are shown in Fig. 9, where port A reached anaerobic conditions (0.1% oxygen) after 1 day of injection, while under open boundary conditions oxygen levels in port A equilibrated at 2.15% (Fig. 5). Also, the recovery time for the partially covered configuration was delayed by a factor of 3.5 (it took 7 days instead of 2 to equilibrate with atmospheric conditions). Close agreement was also observed between numerical simulations and experimental values as shown in Fig. 9.
7. Conclusions Simulations and visualization experiments were performed to investigate the establishment of the zone of influence in anaerobic bioventing. The results indicate that injection of anaerobic gas mixture can effectively establish a reducing zone, while impermeable barriers may dramatically improve the oxygen exclusion efficiency. Numerical data using the FEHM code have been validated with experimental data. Although the experiments were conducted under a number of simplifying assumptions, (homogeneous and dry sand
220
P.G. Mihopoulos et al. / Journal of Contaminant Hydrology 58 (2002) 209–220
formation, nonreactive gases), the proposed method and the numerical model presented in this paper can be extended to account for field heterogeneities and assist in the design of an effective ABV soil remediation system. Also, optimization of the anaerobic and aerobic residence times, can make ABV a particularly attractive in situ bioremediation process, by minimizing ex situ treatment costs.
Acknowledgements The authors would like to thank Gaius J. Roemer, and George A. Zyvoloski from Lanl EES-5 for their assistance in FEHM simulations. This research was supported by U.S. EPA under COE-UC/NMRREL Cooperative Agreement CR-821029, and Battelle Memorial Institute, Columbus, OH.
References Cussler, E.L., 1984. Diffusion: Mass Transfer in Fluid Systems. Cambridge Univ. Press, UK. Dupont, R.R., 1993. Environ. Prog. 12, 45 – 53. EPA Manual, 1995. Bioventing Principle and Practice. Volume I: Bioventing Principles, Volume II: Bioventing Design EPA/540/R-95/534a, Sept. 1995. Geankoplis, C., 1972. Mass Transfer Phenomena. Holt, Rinehart, and Winston, New York, p. 149ff. Gibson, T.L., Abdul, A.S., Glasson, W.A., Ang, C.C., Gatlin, D.W., 1993. Vapor extraction of volatile organic compounds from clay soil: a long-term pilot study. Groundwater 31 (4), 616 – 626. Mason, E.A., Malinauskas, A.P., 1983. Gas Transport in Porous Media: The Dusty Gas Model. Chem. Eng. Monogr., vol. 17. Elsevier, New York. Massmann, J., Farrier, D.F., 1993. Effects of atmospheric pressures on gas transport in the vadose zone. Water Resour. Res. 28 (3), 777 – 791. Mihopoulos, P.G., Suidan, M.T., Sayles, G.D., 2000. Vapor phase treatment of PCE by lab-scale anaerobic bioventing. Water Res. 34 (12), 3231 – 3237. Mihopoulos, P.G., Suidan, M.T., Sayles, G.D., 2001. Complete remediation of PCE contaminated unsaturated soils by sequential anaerobic – aerobic bioventing. Water Sci. Technol. 43 (5), 365 – 372. Millington, R.J., Quirk, J.P., 1961. Permeability of porous solids. Trans. Faraday Soc. 57, 1200 – 1207. Oostrom, M., Hofstee, C., Walker, R.C., Dane, J.H., 1999. Movement and remediation of trichloroethylene it a saturated heterogeneous medium: 1. Spill behavior and initial dissolution. J. Contam. Hydrol. 37, 159 – 178. Popovica, J., Brusseau, M.L., 1997. Dispersion and transport of gas-phase contaminants in dry porous media: effect of heterogeneity and gas velocity. J. Contam. Hydrol. 28, 157 – 169. Sayles, G.D., Mihopoulos, P.G., Suidan, M.T., 1997. In: Alleman, B.C., Leeson, A. (Eds.), In Situ and On-Site Bioremediation, vol. 1. Battelle Press, Columbus, pp. 353 – 359. Shah, J.K., Sayles, G.D., Suidan, M.T., Mihopoulos, P.G., Kaskassian, S.R., 2001. Anaerobic bioventing of unsaturated zone contaminated with DDT and DNT. Water Sci. Technol. 43 (2), 35 – 42. Sleep, B., 1998. Modeling transient organic vapor transport in porous media with the dusty gas model. Adv. Water Resour. 22 (3), 247 – 256. Van Eyk, J., 1997. Petroleum Bioventing. Balkema, Netherlands. Zyvoloski, G.A., Robinson, B.A., Dash, Z.V., Trease, L.L., 1997. User’s manual for the FEHM application—a finite element mass and heat transfer code. Los Alamos National Laboratory report LA-13306-M.
Numerical modeling of oxygen exclusion experiments of anaerobic bioventing Philip G. Mihopoulos a,*,1, Makram T. Suidan a, Gregory D. Sayles b, Sebastien Kaskassian a a
Department of Civil and Environmental Engineering, University of Cincinnati, Cincinnati, OH 45221-0071, USA b U.S. Environmental Protection Agency, National Risk Management Research Laboratory, Cincinnati, OH 45268, USA Received 15 December 2000; received in revised form 8 January 2002; accepted 12 April 2002
Abstract A numerical and experimental study of transport phenomena underlying anaerobic bioventing (ABV) is presented. Understanding oxygen exclusion patterns in vadose zone environments is important in designing an ABV process for bioremediation of soil contaminated with chlorinated solvents. In particular, the establishment of an anaerobic zone of influence by nitrogen injection in the vadose zone is investigated. Oxygen exclusion experiments are performed in a pilot scale flow cell (2 1.1 0.1 m) using different venting flows and two different outflow boundary conditions (open and partially covered). Injection gas velocities are varied from 0.25 10 3 to 1.0 10 3 cm/ s and are correlated with the ABV radius of influence. Numerical simulations are used to predict the collected experimental data. In general, reasonable agreement is found between observed and predicted oxygen concentrations. Use of impervious covers can significantly reduce the volume of forcing gas used, where an increase in oxygen exclusion efficiency is consistent with a decrease in the outflow area above the injection well. D 2002 Elsevier Science B.V. All rights reserved. Keywords: Anaerobic; Bioventing; Modeling; Oxygen exclusion
* Corresponding author. Present address: Tait Environmental Management, Inc., 701 North Parkcenter Drive, Santa Ana, CA 92705, USA. Fax: +1-714-560-8235. E-mail addresses: [email protected] (P.G. Mihopoulos), [email protected] (M.T. Suidan), [email protected] (G.D. Sayles), [email protected] (S. Kaskassian). 1 Fax: + 1-513-556-2599.
0169-7722/02/$ - see front matter D 2002 Elsevier Science B.V. All rights reserved. PII: S 0 1 6 9 - 7 7 2 2 ( 0 2 ) 0 0 0 3 7 - 2
210
P.G. Mihopoulos et al. / Journal of Contaminant Hydrology 58 (2002) 209–220
1. Introduction Bioventing has been used extensively in recent years to biologically remediate unsaturated soils contaminated with petroleum hydrocarbons, which have been introduced in the subsurface from leaking underground storage tanks, surface spills and improper disposal methods. Anaerobic bioventing is a new technology intended to treat highly chlorinated compounds in the vadose zone (Sayles et al., 1997; Mihopoulos et al., 2000; Shah et al., 2001). While it employs the basic principles of aerobic bioventing (EPA Manual, 1995; Van Eyk, 1997; Dupont et al., 1993), it is applied to reduce, under anaerobic environments, compounds that are recalcitrant under aerobic conditions. In experimental studies, a gaseous mixture of nitrogen, hydrogen and carbon dioxide, injected in columns representing unsaturated zone profiles contaminated with PCE, was shown to reduce PCE concentrations of 10 ppmv with a half-life of 7 min. Also, complete remediation of PCE was shown in a series of anaerobic –aerobic columns. (Mihopoulos et al., 2001). Anaerobic bioventing is applied by injecting a mixture of nitrogen containing small concentrations of carbon dioxide and hydrogen (well below the lower flammable limit of 4%). The chlorinated vapors are reduced, utilizing hydrogen as an electron donor, under induced anaerobic conditions, while the partially reduced byproducts can be reduced further, or oxidized in the presence of increased oxygen levels as the plume rises to the surface. In a field application, the goal is to maximize biodegradation, by minimizing advection of the contaminants in the gas phase, thus, allowing sufficient residence times for both the anaerobic and the subsequent aerobic step. ABV can be particularly attractive for a big family of chlorinated contaminants that are recalcitrant under aerobic conditions, where a two-step anaerobic – aerobic remediation scheme appears to be the best alternative. The limiting step in developing an engineered bioprocess for complete destruction of chlorinated compounds appears to be a rapid and reliable anaerobic phase. ABV combines principles of diffusive and convective gas transport, interphase mass transfer and reaction kinetics. Portions of free and dissolved nonaqueous phase liquids can partition in the gas phase when in direct contact with the advected gases (N2, H2 and CO2). Typical injection rates for aerobic bioventing are in the order of 0.1 cm/s (EPA Manual, 1995), while gas velocities for soil vapor extraction (SVE) applications are usually greater than 1.5 cm/s in the vicinity of the injection well. At those velocities, advective and dispersive are the dominant transport and mixing mechanisms. The flows required for ABV should be considerably less than those used for SVE and aerobic bioventing, with diffusive fluxes expected to dominate advective fluxes. A key characteristic of ABV, in contrast to aerobic bioventing, is the requirement of an anaerobic zone in the vicinity of the contaminant plume, since aerobic zones are naturally present in the upper layers of the vadose zone. The purpose of this work is to investigate the establishment of an anaerobic environment by forcing nitrogen gas through the vadose zone. The paper is organized as follows. First, we report on visualization venting experiments conducted in a 2-D flow cell. Using nitrogen as a driving gas, the displacement of the initially present oxygen in the porous medium is studied. We monitor the variation in the concentrations of oxygen with time and space, and with different injection flowrates. Subsequently, numerical simulations of the process are performed. The numerical program used is calibrated using data from the oxygen displacement experi-
P.G. Mihopoulos et al. / Journal of Contaminant Hydrology 58 (2002) 209–220
211
ments. The solutions of the numerical model are then compared to independent experimental data. Sensitivity studies are conducted to analyze the dependence of radius of influence on the injected flows.
2. Experimental setup Soil in the vadose zone was experimentally simulated with a flow cell, shown schematically in Fig. 1. The cell dimensions were 2 m long by 1.1 m high by 0.01 m wide. The front side (2 1.1 m) consisted of glass, while the backside of the cell was made of stainless steel with 220 monitoring/sampling ports of 1/4-in. Swagelok fittings. The influent nitrogen stream was passed through an oxygen trap containing a solution of 100 mg/l Na2SO3 and 25 mg/l CoCl26H2O. The injection port, designed to simulate a screened injection well, was 10 10 cm in dimension, located 30 cm above the left bottom corner of the flow cell, while a woven stainless steel mesh (opening 10 Am) was placed in direct contact with the sand. The nitrogen flowrate was regulated with a mass flow controller (MKS Instruments, Andover, MA) as shown in Fig. 1. A high-pressure cylinder served as the source of UHP Grade Nitrogen (Matheson Gas Products, Chicago, IL). Gas samples were collected at various sampling ports with Pressure-Lok gas tight syringes (Dynatech Baton Rouge, LA) and analyzed immediately. Oxygen gas analysis was performed on a Model 5890 Series II Hewlett-Packard (Mountain View, CA) gas chromatograph, equipped with a thermal conductivity detector. The carrier gas (argon) flow was 30 mL/min, and retention times for O2 and N2 were 0.61 and 0.72 min, respectively, on a 10-ft 45/60 Molecular Sieve column (Hewlett Packard). The oven temperature was maintained isothermal at 50 jC.
Fig. 1. Experimental setup.
212
P.G. Mihopoulos et al. / Journal of Contaminant Hydrology 58 (2002) 209–220
3. Porous media and gas properties The sand used in this study (ASTM #D-1556) was obtained from ELE International Soiltest (Lake Buff, IL). The sand had a bulk density of 1.49 g/cm3, and a porosity of 0.35. The probability distribution was determined with sieve analysis, where 93% by mass was retained between US Sieve nos. 70 and 50 (geometric mean size (210297)0.5=250 Am). The value of the diffusion coefficient D for the gas pair nitrogen – oxygen is 1.811.10 5 m2 s 1 (Cussler, 1984). The effective diffusion coefficient, De, is usually reduced by an order of magnitude due to tortuosity and constrictivity effects as: De=hD/a2, where h is the porosity and a is the actual pore length per distance in the direction of diffusion (Geankoplis, 1972). Lumping those effects, and using the formula of Millington and Quirk (1961), De=D/s, where the tortuosity for dry porous media was approximated as s=h 2/3, we obtained a value for tortuosity of 2, which is in the typical range for sandy soils. Vadose zone remediation by ABV depends primarily on vapor phase flowrates that influence the transport and removal of gas-phase contaminants. Thus, it is critical to study processes affecting gas transport and removal at gas velocities applicable to the field. During SVE, gas velocities greater than 1.6 cm/s in the vicinity of the extraction well have
Fig. 2. Oxygen concentration contours for 100 ml/min. Subplots [a], [b], [c] and [d] represent experimental results for 0.5, 1, 2 and 5 days of venting, respectively. Subplot [e] represents numerical results after 5 days of venting.
P.G. Mihopoulos et al. / Journal of Contaminant Hydrology 58 (2002) 209–220
213
been reported from field studies (Gibson et al., 1993). In our case, an inlet flowrate of 100 mL/min, corresponded to an interstitial velocity of 0.007 cm/s in the vicinity of the injection well. Popovica and Brusseau (1997), concluded that hydrodynamic dispersion dominates molecular diffusion at gas velocities greater than 1 cm/s, while at gas velocities lower than 0.3 cm/s molecular diffusion is the major source of spreading. At particle Reynolds numbers above 10, the dispersion coefficient varies linearly with flow, whereas at low flowrates dispersion becomes a function of the Schmidt number (kinematic viscosity/diffusion coefficient) because both flow and molecular diffusion are important in this region (Cussler, 1984). In addition to advective and diffusive fluxes of the injected gas, soil gas flow is shown to be induced by cyclic barometric pressure fluctuations when pressure differentials are bigger than 2 –3 kPa (Massmann and Farrier, 1993). Although those effects might be considered at specific field applications (storm events), they were not significant in this study and were neglected.
4. Two-dimensional flow experiments Three different venting flows of 100, 200 and 500 ml/min were investigated. The twenty-five sampling ports are shown in the mesh diagram of Fig. 2a and in Fig. 1. Initial conditions were atmospheric concentrations of oxygen. To obtain uniform initial con-
Fig. 3. Partially covered experimental [a] and numerical [b] values of oxygen concentrations for Q = 100 mL/min at steady state.
214
P.G. Mihopoulos et al. / Journal of Contaminant Hydrology 58 (2002) 209–220
ditions in all experiments, a period of 7 days was allowed between consecutive experiments. The use of impermeable covers was investigated as a mean of increasing in situ oxygen exclusion efficiencies, by partially blocking the top left outflow boundary with a 0.75-m long impermeable plastic material, as shown in Fig. 7. All experiments were conducted at room temperature (21 –23 jC) and pressure, while all chromatographic analysis was completed within an hour of every sampling event.
5. Numerical simulations The computational domain was two-dimensional (200 110 cm) and consisted of an injection area placed in the bottom left side of the flow cell to simulate the experimental setup as shown in Fig. 1. The numerical code FEHM (Zyvoloski et al., 1997) was used to simulate gas movement in the unsaturated zone. This finite element 3-D code can simulate multiphase and multicomponent transport of air, water and oil in porous media, and has been widely used and tested in a variety of studies. The two-dimensional grid used was uniformly refined from 200 to 1800 quadrilateral elements (spacing 10.16 and 3.4 cm, respectively) and was then continuously refined near the injection source until uniform results were obtained from consecutive grid refinements. The nonuniform grid divided the y-axis in three sections. The midsection, corresponding to the injection screening height, had a uniform spacing of 0.7 cm after the final refinement. The top and bottom sections had elements that increased proportionally to a weighing factor with increasing distance from the injection point. The x-axis was also refined using cells that increased in size with increasing distance from the injection point. In the current study, only vapor transport was considered while the liquid phase was passive. Mass transport of multiple nonreactive solutes through an unsaturated porous medium in 2-D was solved in FEHM using the Galerkin finite element formulation. The discretized nonlinear set of equations were solved using the Newton– Raphson iterative procedure. Semiautomatic step control was used in macros ctrl and trac, with a tolerance of 10 6 and a time step multiplier of 1.5. Zero flux boundary conditions were specified at the lateral and bottom boundaries of the domain. At the top outflow boundary, the concentrations of oxygen and nitrogen were fixed at atmospheric values. At the nitrogen injection point, the injection flowrates were equally distributed in 15 lateral nodes, using a Neumann boundary, while a zero flux boundary was imposed at the top impermeable section. A time-step increment of 1.5 was used after convergence. The maximum number of Newton iterations was 20, with a convergence factor of 10 6. The upwind averaging factor was set at 1.2. Values of porosity, permeability and diffusion coefficients were assumed uniform throughout the domain. An initial pressure field of 101.33 kPa was assumed, and a temperature of 23 jC. The porous medium permeability to air flow was approximated by the Carman – Kozeny equation (due to the homogeneity of the used silica sand particle size, and the assumption that the average particle shape does not deviate strongly from a spherical shape), as 100 Da (10 10 m2). Practically, gas permeabilities higher than 0.1 Da are adequate for applications of aerobic bioventing (EPA Manual, 1995), where for permeabilities below 0.01 Da gas flow is primarily through the secondary porosity and field testing is required to establish feasibility. The small flowrates required for an ABV
P.G. Mihopoulos et al. / Journal of Contaminant Hydrology 58 (2002) 209–220
215
application, make the process attractive for sites with small permeability, which are typically encountered in clay formations. For permeabilities lower than 0.001 Da (Sleep, 1998), contributions to molecular diffusion from gas – wall interactions (Knudsen diffusion) become significant, and a correction factor should be introduced to account for this effect or the dusty gas model (Mason and Malinauskas, 1983) should be used instead.
6. Results and discussion 6.1. Open boundary experiments Figs. 2 –4 summarize transient experimental and steady state numerical results, for flows of 100, 200 and 500 ml/min, respectively. Oxygen contours from experimental values are plotted for t = 12 h, 1, 2 and 5 days of venting operation as shown in subplots [a] through [d]. Subplots [e] of Figs. 2 – 4 show the numerical results obtained for 5 days of nitrogen injection for the corresponding flow. From comparison of the experimental contours [a] –[d], it was concluded that steady state operation was achieved after one day of venting for every flow examined. The numerical results obtained at 5 days (Figs. 2e –
Fig. 4. Oxygen concentration contours for 200 mL/min. Subplots [a], [b], [c] and [d] represent experimental results for 0.5, 1, 2 and 5 days of venting, respectively. Subplot [e] represents numerical results after 5 days of venting.
216
P.G. Mihopoulos et al. / Journal of Contaminant Hydrology 58 (2002) 209–220
Fig. 5. Oxygen concentration contours for 500 mL/min. Subplots [a], [b], [c] and [d] represent experimental results for 0.5, 1, 2 and 5 days of venting, respectively. Subplot [e] represents numerical results after 5 days of venting.
4e) were in close agreement with the corresponding experimental values (Figs. 2d– 4d), and could qualitative predict the shape and extent of the anaerobic radius of influence. For the injection rate of 200 mL/min, the transient behavior of the pumping and recovery phases were examined in a different experiment. Starting from atmospheric conditions, a flow of 200 mL/min was injected, and oxygen levels were monitored continuously until steady state. Subsequently, the nitrogen flow was shut off and oxygen was allowed to diffuse back from the open boundary (recovery phase). The oxygen levels during these two phases were monitored at four specific locations (A through D as shown in Fig. 1) in the flow cell, and the experimental results are shown with discrete points in Fig. 5. The corresponding predicted values from the numerical model are shown with continuous lines in Fig. 5, where close agreement between numerical and experimental values is observed. Effects of longitudinal dispersion could be safely neglected since the injection velocities were sufficiently small. For the experiments under consideration, the diffusive time scaled with L2/De f 5 10 3 s, where L is a characteristic length, here equal with the mean grain diameter, and convective time scaled with L/U f 10 s, where U is the average gas velocity, making molecular diffusion the predominant transport mechanism. With an effective diffusivity of 0.9 10 6 m2/s, and assuming a dispersivity value in the
P.G. Mihopoulos et al. / Journal of Contaminant Hydrology 58 (2002) 209–220
217
Fig. 6. Predicted and experimental values of transient oxygen concentrations for a flow of 200 mL/min.
order of the grain size, the ratio of De/au scales as O (104 –105) at flowrates ranging from 100 to 500 mL/min. The differences between simulations accounting for, and neglecting, longitudinal dispersion were negligible and supported the above assumption. Oostrom et al. (1999), using a similar experimental setup, estimated as reasonable dispersivity values less than 0.001 m. A final experiment was conducted with the open boundary, where the steady state profile (after 5 days of operation) was obtained for four additional nitrogen injection rates of 1, 10, 50 and 1000 mL/min, respectively. For every additional injection rate, the 25 sampling ports were analyzed for oxygen after 5 days of venting. The final oxygen concentrations (normalized by the ambient atmospheric oxygen concentration) where averaged and plotted as a function of the specific flowrate in Fig. 6. 6.2. Partially covered vadose zone Venting experiments and simulations have been conducted with the top outflow boundary partially (75%) covered. An injection rate of 100 mL/min was applied and experimental steady state results are shown in Fig. 7a. The corresponding numerical results were in reasonable agreement with the experimental values as shown in Fig. 7b. Those results indicate that impermeable covers can be used to improve the oxygen exclusion efficiency approximately six times, compared with an open outflow boundary, while this number is expected to increase with the use of lower venting flows. By comparing Figs. 7 and 2e, injecting a flow of 100 mL/min with a partially covered outflow boundary, can yield similar results with that of a 500 mL/min flow with an open
218
P.G. Mihopoulos et al. / Journal of Contaminant Hydrology 58 (2002) 209–220
Fig. 7. Residual oxygen concentration as a function of the injection flowrate.
outflow boundary. The numerical model was used to examine the effect of this specific boundary condition under different injection flowrates. This is shown in Fig. 8, where the area percentage with oxygen values less than 1% is plotted as a function of venting flow for both totally and partially open outflow boundaries. Increasing the nitrogen venting
Fig. 8. Comparison of anaerobic (oxygen level less than 1%) area percentage between open and partially covered outflow boundaries.
P.G. Mihopoulos et al. / Journal of Contaminant Hydrology 58 (2002) 209–220
219
Fig. 9. Predicted and experimental values of transient oxygen concentrations for a flow of 200 mL/min and a partially covered outflow.
flow, beyond 1000 mL/min (open) and 300 mL/min (covered) does not correspond to an increase in the anaerobic zone, since back-diffusion of atmospheric oxygen at the top layers dominates convective flow that originates from the injection section. Also, a pumping and recovery phase experiment was conducted, similar with the one described for the open boundary. The results are shown in Fig. 9, where port A reached anaerobic conditions (0.1% oxygen) after 1 day of injection, while under open boundary conditions oxygen levels in port A equilibrated at 2.15% (Fig. 5). Also, the recovery time for the partially covered configuration was delayed by a factor of 3.5 (it took 7 days instead of 2 to equilibrate with atmospheric conditions). Close agreement was also observed between numerical simulations and experimental values as shown in Fig. 9.
7. Conclusions Simulations and visualization experiments were performed to investigate the establishment of the zone of influence in anaerobic bioventing. The results indicate that injection of anaerobic gas mixture can effectively establish a reducing zone, while impermeable barriers may dramatically improve the oxygen exclusion efficiency. Numerical data using the FEHM code have been validated with experimental data. Although the experiments were conducted under a number of simplifying assumptions, (homogeneous and dry sand
220
P.G. Mihopoulos et al. / Journal of Contaminant Hydrology 58 (2002) 209–220
formation, nonreactive gases), the proposed method and the numerical model presented in this paper can be extended to account for field heterogeneities and assist in the design of an effective ABV soil remediation system. Also, optimization of the anaerobic and aerobic residence times, can make ABV a particularly attractive in situ bioremediation process, by minimizing ex situ treatment costs.
Acknowledgements The authors would like to thank Gaius J. Roemer, and George A. Zyvoloski from Lanl EES-5 for their assistance in FEHM simulations. This research was supported by U.S. EPA under COE-UC/NMRREL Cooperative Agreement CR-821029, and Battelle Memorial Institute, Columbus, OH.
References Cussler, E.L., 1984. Diffusion: Mass Transfer in Fluid Systems. Cambridge Univ. Press, UK. Dupont, R.R., 1993. Environ. Prog. 12, 45 – 53. EPA Manual, 1995. Bioventing Principle and Practice. Volume I: Bioventing Principles, Volume II: Bioventing Design EPA/540/R-95/534a, Sept. 1995. Geankoplis, C., 1972. Mass Transfer Phenomena. Holt, Rinehart, and Winston, New York, p. 149ff. Gibson, T.L., Abdul, A.S., Glasson, W.A., Ang, C.C., Gatlin, D.W., 1993. Vapor extraction of volatile organic compounds from clay soil: a long-term pilot study. Groundwater 31 (4), 616 – 626. Mason, E.A., Malinauskas, A.P., 1983. Gas Transport in Porous Media: The Dusty Gas Model. Chem. Eng. Monogr., vol. 17. Elsevier, New York. Massmann, J., Farrier, D.F., 1993. Effects of atmospheric pressures on gas transport in the vadose zone. Water Resour. Res. 28 (3), 777 – 791. Mihopoulos, P.G., Suidan, M.T., Sayles, G.D., 2000. Vapor phase treatment of PCE by lab-scale anaerobic bioventing. Water Res. 34 (12), 3231 – 3237. Mihopoulos, P.G., Suidan, M.T., Sayles, G.D., 2001. Complete remediation of PCE contaminated unsaturated soils by sequential anaerobic – aerobic bioventing. Water Sci. Technol. 43 (5), 365 – 372. Millington, R.J., Quirk, J.P., 1961. Permeability of porous solids. Trans. Faraday Soc. 57, 1200 – 1207. Oostrom, M., Hofstee, C., Walker, R.C., Dane, J.H., 1999. Movement and remediation of trichloroethylene it a saturated heterogeneous medium: 1. Spill behavior and initial dissolution. J. Contam. Hydrol. 37, 159 – 178. Popovica, J., Brusseau, M.L., 1997. Dispersion and transport of gas-phase contaminants in dry porous media: effect of heterogeneity and gas velocity. J. Contam. Hydrol. 28, 157 – 169. Sayles, G.D., Mihopoulos, P.G., Suidan, M.T., 1997. In: Alleman, B.C., Leeson, A. (Eds.), In Situ and On-Site Bioremediation, vol. 1. Battelle Press, Columbus, pp. 353 – 359. Shah, J.K., Sayles, G.D., Suidan, M.T., Mihopoulos, P.G., Kaskassian, S.R., 2001. Anaerobic bioventing of unsaturated zone contaminated with DDT and DNT. Water Sci. Technol. 43 (2), 35 – 42. Sleep, B., 1998. Modeling transient organic vapor transport in porous media with the dusty gas model. Adv. Water Resour. 22 (3), 247 – 256. Van Eyk, J., 1997. Petroleum Bioventing. Balkema, Netherlands. Zyvoloski, G.A., Robinson, B.A., Dash, Z.V., Trease, L.L., 1997. User’s manual for the FEHM application—a finite element mass and heat transfer code. Los Alamos National Laboratory report LA-13306-M.