Environ. Sci. Technol. 2005, 39, 8453-8459
Mass-Transfer Limitations for Nitrate Removal in a Uranium-Contaminated Aquifer J I A N L U O , * ,† O L A F A . C I R P K A , ‡ WEIMIN WU,† MICHAEL N. FIENEN,† PHILIP M. JARDINE,§ TONIA L. MEHLHORN,§ DAVID B. WATSON,§ CRAIG S. CRIDDLE,† AND PETER K. KITANIDIS† Department of Civil & Environmental Engineering, Stanford University, Stanford, California 94305-4020, Swiss Federal Institute of Aquatic Science and Technology (EAWAG), U ¨ berlandstr. 133, 8600 Du ¨ bendorf, Switzerland, and Oak Ridge National Laboratory, Environmental Science Division, Oak Ridge, Tennessee 37831-6038
A field test on in situ subsurface bioremediation of uranium(VI) is underway at the Y-12 National Security Complex in the Oak Ridge Reservation, Oak Ridge, TN. Nitrate has a high concentration at the site, which prevents U(VI) reduction, and thus must be removed. An acidicflush strategy for nitrate removal was proposed to create a treatment zone with low levels of accessible nitrate. The subsurface at the site contains highly interconnected fractures surrounded by matrix blocks of low permeability and high porosity and is therefore subject to preferential flow and matrix diffusion. To identify the heterogeneous mass transfer properties, we performed a novel forced-gradient tracer test, which involved the addition of bromide, the displacement of nitrate, and the rebound of nitrate after completion of pumping. The simplest conceptualization consistent with the data is that the pore-space consists of a single mobile domain, as well as a fast and a slowly reacting immobile domain. The slowly reacting immobile domain (shale matrix) constitutes over 80% of the pore volume and acts as a long-term reservoir of nitrate. According to simulations, the nitrate stored in the slowly interacting immobile domain in the fast flow layer, at depths of about 12.2-13.7 m, will be reduced by an order of magnitude over a period of about a year. By contrast, the mobile domain rapidly responds to flushing, and a low average nitrate concentration can be maintained if the nitrate is removed as soon as it enters the mobile domain. A fieldscale experiment in which the aquifer was flushed with acidic solution confirmed our understanding of the system. For the ongoing experiments on microbial U(VI) reduction, nitrate concentrations must be low in the mobile domain to ensure U(VI) reducing conditions. We therefore conclude that the nitrate leaching out of the immobile pore space
* Corresponding author phone: (650) 723-8321; fax: (650) 7259720; e-mail:
[email protected]. † Stanford University. ‡ EAWAG. § Oak Ridge National Laboratory. 10.1021/es050195g CCC: $30.25 Published on Web 09/27/2005
2005 American Chemical Society
must continuously be removed by in situ denitrification to maintain favorable conditions.
Introduction Microbial reduction of U(VI) to U(IV) is a promising remediation strategy for uranium-contaminated groundwater. The solubility and mobility of uranium depend on its valence state. Under oxidizing conditions in acidic environments, U(VI) exists predominantly as the uranyl cation UO22- or uranyl-carbonate complexes, which are relatively mobile. Microbial reduction of U(VI) under anaerobic conditions produces U(IV), which rapidly precipitates as highly insoluble uraninite, UO2. Laboratory experiments have shown that this strategy is promising (e.g., 1-4), but its efficiency under complex field conditions remains to be confirmed. Over the past 4 years, we have characterized a site in Area 3 of the Field Research Center (FRC) of the Natural and Accelerated Bioremediation (NABIR) program of the U.S. Department of Energy (DOE) at the Y-12 National Security Complex at Oak Ridge, TN. A biostimulation experiment is ongoing at this site to evaluate the potential for microbial reduction of U(VI) in the contaminated aquifer. This site is adjacent to the former S-3 disposal ponds, which were used for disposal of highly acidic (pH ≈1) U-NO3- waste solutions between 1951 and 1983. It is believed that the ponds received 3.2 × 105 m3 of waste during this period. Nitric acid concentrations were in the range of tens of grams per liter, and uranium concentrations were in the range of hundreds of milligrams per liter. In the early 1980s, the ponds were partially neutralized, partially denitrified, capped, and paved to cut off infiltration. The acidic groundwater (pH ≈3.4) near the source area contains high levels of nitrate (≈10 g/L), and soluble uranium is present at highly toxic levels of 20-50 mg/L. The remediation experiment aims to establish favorable field conditions for the growth of indigenous bacteria and stimulate their activity by delivering appropriate substrate and nutrients to foster U immobilization as uraninite. We have installed a recirculation scheme at the site, consisting of two injection wells and two extraction wells, creating a nested flow cell, which functions as an in situ reactor for the bioremediation experiment (5). A three-phase remediation strategy has been designed for this demonstration (6). It includes first removing nitrate prior to stimulation of U(VI) reduction and then adjusting the pH to levels favorable for activity of U(VI)-reducing bacteria, i.e., to about neutral values, and finally adding electron donor to the in situ reactor to foster reduction and immobilization of U(VI). Nitrate inhibits microbial reduction of U(VI) because dissimilatory metal reducing microorganisms preferentially reduce nitrate before U(VI) and because nitrate and its partial reduction products such as nitrite can reoxidize U(IV) (710). Accessible nitrate must therefore be removed prior to U reduction, and concentrations must be maintained at very low levels during bioremediation. Under current conditions, stimulation of in situ microbial denitrification of the treatment zone is inadvisable, because the extremely high nitrate concentration would result in accumulation of gaseous N2 and excessive biomass in the subsurface. Pump-and-treat is also not effective in creating a treatment zone with acceptably low levels of nitrate concentrations because of mass transfer limitations and proximity to the contaminant source (S-3 ponds) that can continuously release highly contaminated groundwater. Instead, we chose to flush the site with nitrateVOL. 39, NO. 21, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
8453
free water prior to biostimulation (6). This strategy would significantly reduce the average nitrate concentration in the accessible region, i.e., the mobile domain, although it would have little effect on the bulk of nitrate in the immobile domains. Because of the high degree of chemical heterogeneity on this small scale, nitrate would be depleted in some areas and thus not inhibit the U(VI) reduction. Understanding the mechanisms controlling nitrate transport and release from matrix storage is essential in designing such a strategy for a successful U(VI) bioreduction experiment in U(VI) and nitrate cocontaminated aquifers. At our site, the mechanisms may be complex because the subsurface solid matrix, weathered saprolite, is highly heterogeneous and exhibits multiporosity behavior. Most flow occurs in a zone of interconnected fractures on the millimeter scale, referred to as a fast flow layer, which contributes only marginally to the overall void-space, and storage of groundwater and its dissolved constituents is predominantly in stagnant micropores (11). The overall mass flux may be low because of low overall fracture porosity and slow mass transfer from the matrix pores to the fractures (12). In highly heterogeneous aquifers, standard short-duration tracer tests are insufficient to characterize mass transfer because these tests are insensitive to long-term diffusion processes dominating the exchange within slow immobile domains (13). In this study, we conducted a novel forced-gradient tracer test with the well system operating under conditions similar to those of the subsequent U(VI) bioremediation experiments. The tracer test involved the addition of bromide, the displacement of nitrate, and the rebound of nitrate after completion of pumping. To model the observed behavior in the fast flow layer, which is also the bioactive zones, we extended the two-region (mobile-immobile) model by dividing the immobile domain into two subdomains: one with fast exchange to the mobile domain and the other with slow exchange. The tracer test and fitted model allow us to characterize the heterogeneous mass transfer properties in the treatment zone, to understand the limiting processes involved in nitrate removal, and to address the issue of how long nitrate inhibition on U(VI) bioreduction may persist.
Methods and Materials Tracer Test. To create a treatment zone with favorable conditions for U(VI) bioreduction, we installed a recirculation system (Figure 1) consisting of two injection wells and two extraction wells, creating an inner flow cell nested within an outer flow cell (5, 6). In the bioremediation operations, chemical amendments, such as electron donors or pH modifiers, are added to the nested inner cell in injection well FW104, while the outer cell is flushed with a nitrate- and uranium-free solution at a pH that favors retention of sorbed U(VI). This setup prevents loss of U(VI) from the system and creates favorable conditions for U(VI) reduction inside the inner cell. The pumping wells were aligned approximately along the strike, and the multilevel sampling wells (MLS) were aligned along the dip within the pumping well field. More detailed geologic characterization of the aquifer has been presented by Solomon et al. (11) and Fienen et al. (14). Water was injected at FW104 and FW024 and extracted at FW026 and FW103 with the pumping rates as labeled in Figure 1. Figure 1 also shows a conceptual diagram of the flow field boundaries, based on a homogeneous isotropic model. Though heterogeneity and anisotropy impact the geometry of the flow field, some qualitative conclusions may be drawn from this diagram. The domain is divided into four flow zones: zone I is the nested inner cell connecting the inner two pumping wells and functioning as an in situ reactor for the subsequent bioremediation experiment, zone II is a transition zone connecting the inner and outer well pairs, 8454
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 21, 2005
FIGURE 1. Plan view of the well system for the bioremediation experiment and the tracer test. FW024 and FW103 are referred as the outer wells and FW026 and FW104 as the inner wells. Down arrows indicate injection, and up arrows indicate extraction. FW100, FW101, and FW102 are MLS wells. Solid lines are separation streamlines based on a homogeneous and isotropic model. The separations streamlines divide the domain into different flow zones. zone III is an outer cell connecting the outer wells, and zone IV connects the outer wells with the exterior regional flow. Throughout the test, acidified tap water was injected to well FW024. FW104 was injected with the acidified water before and after the tracer test. The acidified water was acidified with HCl to pH 3.6 and contained CaCl2 (5mM) to maintain the ion strength to prevent decomposition of the solid phases. The injection of acidified water was to maintain the same pH as the groundwater to prevent the aquifer from being clogged by in situ precipitation of aluminum hydroxide. During the tracer test, injection water at FW104 was switched from the CaCl2 water to acidified MgBr2 water (10mM Br-, pH 3.6), which has the same ion strength as 5 mM of CaCl2. The test included four phases. Phase 1 (hour 0 to 18:08): an acidified solution with 5 mM CaCl2 was injected into both injection wells while pumping at both extraction wells. Phase 2 (hour 18:08 to 33:52): an acidified solution with 5 mM MgBr2 was injected as tracer into FW104 and an acidified solution with 5 mM CaCl2 was injected into FW024, while pumping at both extraction wells. Phase 3 (hour 33:52 to 96): the addition of bromide tracer was stopped, and tracer-free acidified solution with 5 mM CaCl2 was injected into both injection wells while pumping at both extraction wells. Phase 4 (after hour 96): both the injection and extraction wells were stopped, so that the ambient flow regime was reestablished. Samples were periodically collected at all sampling points to monitor changes in chemical composition as a function of time and space. This phase lasted for 10 months. Modeling. In classical two-domain models, a porous medium is conceptualized as consisting of two overlapping continuous media: a mobile domain, in which advectivedispersive transport occurs, and an immobile one. Mass transfer between the two domains is kinetically controlled and, commonly, a first-order approximation is used to model diffusive mass transfer. Two-domain models have been used primarily to represent systems composed of dead-end pores, porous particles, aggregates, fractures, or macropores (1517). The classical two-domain model, which assumes perfect mixing within the immobile domain, did not fit the data
FIGURE 2. Hydraulic conductivity and contaminant concentration profiles over depth at different wells: (a) hydraulic conductivity, (b) aqueous U(VI) concentration in the mobile domain, (c) total U concentration in sediment, and (d) nitrate concentration. adequately. In reality, the immobile domain is not well mixed so that a distribution of exchange rates rather than a single value may be more appropriate (18, 19). Haggerty and Gorelick (13) presented a multirate model to describe mass transfer in very heterogeneous aquifers. The development and application of the multirate model can also be found elsewhere (e.g., 20, 21). Here, we found that a simpler model fit our data well: We modified the classical two-domain model by dividing the immobile domain into two subdomains in series: the first with fast exchange to the mobile domain and the other interacting slowly with the first. By conceptualizing transport to occur in noninteracting stream tubes (22), the multiple-region transport model for a single stream tube may be formulated as
∂cm ∂cm ∂2cm λm ) -v + D 2 + (cim1 - cm) ∂t ∂x θm ∂x λim ∂cim1 λm (c - cim1) + (c - cim1) ) ∂t θim1 m θim1 im2
cim1,Br(x,0) cim2,Br(x,0) cm,Br(x,0) ) 0, ) 0, )0 c0,Br c0,Br c0,Br cm,NO3- (x,0) c0,NO3-
) 1,
cim1,NO3- (x,0) c0,NO3-
) 1,
cim2,NO3- (x,0) c0,NO3-
(2)
)1 (3)
in which c0,Br is the injected bromide concentration, and c0,NO3is the initial nitrate concentration at the site. We solve eq 1 numerically using finite differences for spatial discretization (upwind differentiation for advection and central differentiation for dispersion) and a fourth-order Runge-Kutta scheme for temporal integration (23).
Results and Discussion (1)
∂cim2 λim (c - cim2) ) ∂t θim2 im1 in which c is the concentration, t is time, v is the mean pore water velocity, D is the hydrodynamic dispersion coefficient, θ is the effective porosity, and x is the spatial coordinate in the direction of flow. The subscripts m, im1, and im2 on c and θ represent the mobile, first immobile (fast exchange), and second immobile (slow exchange) domains, respectively. λm is the first-order mass-transfer constant between the mobile and first immobile domains, and λimis the mass transfer constant between the first and second immobile domains. There is no direct exchange between the mobile domain and the second immobile domain. The initial concentrations for bromide and nitrate transport are assumed zero and unity, respectively.
Initial Conditions. The initial concentration distributions over the aquifer depth were measured in the wells, and values of hydraulic conductivity were obtained by conducting flow meter tests (14). A narrow fracture zone with high hydraulic conductivity was found at about 12 m depth (Figure 2a). The distributions of U(VI) in the aqueous and solid phase correlate with the hydraulic conductivity profile, indicating that uranium migrates predominantly along the strike within this narrow region of preferential flow (Figure 2b,c). Unlike the uranium distribution, nitrate is found in highest concentrations of up to 27 g/L at deeper depth, likely due to long-term penetration into the matrix and the higher density of the nitrate brine (Figure 2d). Bromide Tracer Test. Figure 3 shows the breakthrough curves of bromide at different depths of the MLS wells. The responses at depths 12.2 and 13.7 m are more pronounced than those at other depths. This is consistent with the vertical profile of hydraulic conductivity, shown in Figure 2a, where hydraulic conductivity data below 13 m are unavailable. It confirms that the fast flow layer is at these depths. Furthermore, less bromide was observed at FW100 than at FW101 VOL. 39, NO. 21, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
8455
FIGURE 3. Bromide breakthrough curves at the MLS wells and the extraction wells. c0 is the injected bromide concentration.
FIGURE 4. Measured bromide and nitrate data and simulations. Circles are measurements, and lines are fitted results. and FW102. The profiles at MLS depths 12.2 and 13.7 m of FW101 and FW102 can be simulated through a simple model that envisions one-dimensional flow along single streamlines from the injection well FW104 to each MLS well, neglecting transverse dispersion. However, long tails can be found at the two extraction wells, FW026 and FW103, because these wells capture a set of streamlines reflecting a larger distribution of arrival times. The bromide mass captured at FW026 and FW103 was 55% and 39% of the total injected bromide at FW104, respectively, i.e., only 6% of the injected bromide was not recovered during the 96 h of operation. Nitrate. Responses of nitrate concentrations in wells were also consistent with the hydraulic conductivity distribution. Faster displacement of nitrate was observed in the fast flow layer. Figure 4 shows that the nitrate concentration at depths 12.2 and 13.7 m of FW101 and FW102 dropped to 1% of the initial concentrations, indicating that the fast flow layer responded to flushing quickly and the nested inner cell was not penetrated by exterior regional flow. After the pumps were stopped, nitrate concentrations recovered to almost 60% within 2 months and 80% within 10 months. The rebound was mainly caused by the intrusion of indigenous groundwater and local mass transfer from the immobile domains to the mobile domain. Commonly applied tracer tests usually are of short duration and do not measure the rebound. To 8456
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 21, 2005
TABLE 1. Fitted Parameters and Standard Deviations of Estimation parameters
FW101 (12.2 m depth)
FW102 (13.7 m depth)
λm/θm (10-5/s) λm/θim1 (10-5/s) λim/θim1 (10-7/s) λim/θim2 (10-7/s) v (10-4 m/s) D (10-5 m2/s)
7.54((1.28) 6.53((1.44) 6.06((0.44) 0.96((0.11) 3.02((0.18) 2.725((0.47)
12.38((2.70) 17.20((4.99) 15.38((1.64) 1.48((0.18) 12.35((0.62) 8.43((3.39)
obtain reliable estimates of bulk mass transfer rates, such tests need very precise measurements at late times (13). The rebound data measured in our tracer test include valuable information about late-time behavior and thus allow us to identify and calculate mass-transfer properties. Mass-Transfer Limitation. The model parameters listed in Table 1 are estimated jointly from bromide and nitrate data through nonlinear least-squares fitting applying GaussNewton optimization. The numbers in parentheses are the standard deviations of estimation for the fitted parameters. Due to heterogeneity, we do not expect a single set of parameters to fit all breakthrough curves. The mean pore water velocity obtained at FW102 is much larger than the
TABLE 2. Time Scales for Mass Transfer between the Mobile and Immobile Domains domain considered
interacting with
half-time, t1/2
mobile first immobile first immobile second immobile
first immobile mobile second immobile first immobile
2.5 h 3h 13.3 d 83.5 d
value at FW101, indicating a preferential flow path between FW104 and FW102. The mass transfer coefficients related to the same aqueous region for wells FW101 and FW102 are generally on the same order of magnitude. For the mobile domain, λm/θm and λm/θim1 are on the order of 10-5/s ≈ 1/d, and for the immobile domains, the mass transfer coefficients λim/θim1 and λim/θim2 are 2 orders of magnitude smaller. Figure 4 shows that the calibrated model fits both the nitrate and bromide data very well. Note that we have fit the classical two-domain model to the data, but this model could not capture both the flushing period and the rebound after completion of active pumping. In the following analysis, we use the estimated parameters and measured data at FW101 as a representative set to analyze mass-transfer limitations. The estimated parameters indicate that the mobile and first immobile porosities are similar, whereas the second immobile porosity is about six times larger than the mobile porosity: θm/θim1 ) 0.9, θim2/θim1 ) 6.3. Using a realistic value of the total porosity, 0.4 (24), we arrive at a small value for the mobile porosity, 0.05, which is consistent with the result found by Solomon et al. (11). Our conceptual understanding of the system is that about 80% of the nitrate is in regions that are practically not accessible over the time scale of the flushing experiment. The flushing removed the nitrate in the mobile and first immobile domains but left the nitrate mass in the slowly reacting immobile domain essentially untouched. After flushing was stopped, the remaining mass had 10 months to diffuse from the second immobile domain, through the first immobile domain, and into the accessible mobile domain, eventually leading to a concentration rebound within the mobile aqueous region to about 80% of its preflushing value. At high flow rates, such as during active bioremediation, the slowly reacting immobile domain contributes very little to the solute concentrations in the mobile water. However, it acts as a long-term source for those compounds stored within it, and reducing the flow rates should lead to a rebound in their concentrations, exactly as observed after the flushing of the tracer test was terminated. Since the pore volume (and nitrate mass) is much greater in the second immobile domain than the other domains, the relative rate of diffusive mass transfer between the domains appears to be faster in the mobile domain than in the second immobile domain. To convey this result, we evaluate the half-times (t1/2) (Table 2) of mass transfer for all domains:
t1/2 ) ln(2)
θ λ
(4)
in which the porosity θ is specific to each domain and λ is specific to each mass-transfer process. Obviously, the limiting mass-transfer process is between the two immobile domains. A concentration change in the second immobile domain leads to a 50% response in the mobile domain within about 2 weeks, whereas a concentration change in the mobile domain leads to a 50% response in the second immobile domain within almost 3 months. This indicates that solute mass stored within the slow immobile domain was essentially unaffected by the tracer test, which lasted only 4 days. To significantly reduce nitrate
FIGURE 5. Nitrate concentration and removal ratio in different domains in the fast flow layer at FW101 in an imaginary prolonged flushing experiment. mass requires flushing the aquifer over many months. The latter is illustrated by a numerical simulation of a prolonged flushing shown in Figure 5. The concentration of the mobile region decreases quickly, followed by the first immobile region within a reasonable time frame. Nitrate within the second immobile region decreases slowly to a low level. The lower part of Figure 5 shows the removal ratio R for the area between FW104 and FW101, defined by
Ri(t) )
mi(t0) - mi(t)
(5)
mi(t0)
in which the index i refers to a domain (mobile, first immobile or second immobile), mi(t) is the nitrate mass in the domain at time t, and mi(t0) is the nitrate mass in the domain at the beginning of the tracer test. While the mass in the mobile aqueous domain appears to be removed almost completely once the concentrations in the first immobile domain have dropped significantly, the mass in the second immobile domain remains at high values over a prolonged period of time. This mass is transferred via the first immobile domain to the mobile domain and flushed out rapidly. To claim successful removal of nitrate at the site, the total removal ratio, weighted by the volume fractions of the various domains, must approach unity. The latter is controlled by the mass stored within the slowly reacting immobile domain. It would take about a year to significantly reduce the total nitrate mass in the fast flow layer. Figure 5 shows that the nitrate concentrations in the three domains are quite stable over a long period of time (302000 h). Within this period, using an initial concentration of 10 g/L (Figure 2), the concentrations in the mobile and first and second immobile regions approximated by the model are about 20, 100, and 10 000 mg/L, respectively. This pseudosteady-state can be explained by the concentration change rate due to mass transfer.
dcim1 λim λm (c - cim1) + (c - cim1) ) dt θim1 m θim1 im2 ≈ 10-5/s × (-102 mg/L) + 10-7/s × 104 mg/L ≈ 0 (6) Thus, the first immobile domain acts as a transition region, where the replenishment from the second immobile domain is roughly equal to the release to the mobile domain. Stimulation of denitrification in the mobile domain would only slightly speed up nitrate removal in the second immobile domain because of mass-transfer limitation. Acidic Flushing. Acidic flushing of nitrate was started on August 22, 2003, by injecting a 2 mM KCl solution (pH 3.8) VOL. 39, NO. 21, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
8457
FIGURE 6. Nitrate concentrations at the depth 12.2m of FW101 and FW102 during acidic flushing. at a rate of 0.084 m3/h at FW024 and extracting water with 0.042 m3/h at FW103. Figure 6 shows the nitrate concentrations at the depth of 12.2 m of FW101 and FW102. Within 1 week, nitrate concentrations measured in the wells were reduced to 1% of the initial values. At the end of a month, the concentrations decreased to values in the range of tens of milligrams/liter. At that time, recirculation was initiated, where the extracted groundwater was reinjected rather than 2 mM KCl. Within 4 days (see arrows in Figure 6), the nitrate concentrations jumped to over 1000 mg/L. As anticipated from the analysis, flushing for only 1 month is not sufficient to significantly reduce the nitrate in all the aquifer domains in the fast flow layer. By recirculating the extracted water, the nitrate transferred from the immobile domains into the mobile domain accumulated and thus approached the concentration in the immobile domain. Therefore, flushing with nitrate-free water was reestablished after 4 days of recirculation, and the nitrate concentrations decreased again (Figure 6). Impacts on Bioremediation. For the subsequent U(VI) bioreduction experiment, recirculation of the water extracted at FW026 would bring into the treatment zone water from different flow zones, including slow ones where significant removal of nitrate may take a long time (Figure 3). Thus, short periods (up to about 1 year) are unlikely to be sufficient for the removal of the nitrate in the immobile domains within the treatment zone. The crucial conclusion is that any remediation scheme at the site must be designed considering that the stimulation of U(VI) reduction must be implemented in the presence of a continuous source of nitrate. Since the mobile domain responds rapidly to flushing, low nitrate concentration may be established therein by maintaining a well-designed and reasonably vigorous recirculation scheme while the nitrate is removed as it is transferred to the mobile domain. Let us explore the consequences. Consider that the nitrate flux between the first immobile domain and the mobile domain is
JA ) R(cim1 - cm)
(7)
in which JA is defined as the flux of nitrate in units of mass per volume per time and R ) λm/θm and is the transfer coefficient. A nitrate-dependent degradation rate may be approximated using conventional saturation kinetics:
JB )
qmax,NO3Xcm Ks,NO3 + cm
(8)
in which qmax,NO3 is the maximum specific rate of substrate utilization, Ks,NO3 is the half-rate concentration, and X is the biomass concentration. An intermediate value, 0.2 mg/L 8458
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 21, 2005
FIGURE 7. Contour lines of flux versus nitrate concentration in the mobile domain. Solid lines indicate flux due to mass transfer (JA), and the dashed lines indicate flux due to biodegradation (JB). NO3--N is adopted for Ks,NO3 (25, 26). Both the mass transfer rate JA of nitrate and the degradation rate JB are functions of the mobile concentration. These rates are plotted in Figure 7. Consider first the contour lines of mass transfer (solid lines). The x-intercept defines the state where the nitrate concentration in the mobile domain equals the immobile domain concentration and the mass-transfer flux is zero. Assuming that we start from this point, the degradation flux (dashed lines) tends to decrease the concentration in the mobile domain (see arrows in Figure 7). The JA increases and the JB decreases until the two fluxes become equal, JA ) JB, a condition we refer to as “steady state”. For every combination of qmax,NO3X and cim1 a different steady-state concentration cm in the mobile domain will be approached:
cm )
(
qmax,NO3X 1 cim1 - Ks,NO3 + 2 R qmax,NO3X - Ks,NO3 cim1 R
x(
)
2
)
+ 4cim1Ks,NO3 (9)
Another salient point is that, if a defined nitrate concentration cm must be maintained within the mobile domain, the required concentration of biomass is
X)
R(cim1 - cm)(Ks,NO3 + cm) qmax,NO3cm
(10)
This equation can be used to estimate the critical level of biomass concentration required in the mobile domain. Our previous analysis has indicated that pseudo steadystate conditions in the mobile and first immobile domain are reached over long time periods when accounting for advective-dispersive transport and kinetic mass transfer, leading to concentrations of 20 and 100 mg/L of nitrate in the mobile and first immobile domain, respectively (Figure 5). Thus, we may take 100 mg/L as a representative value for cim1. Assuming a typical value for the maximum specific reaction rate qmax,NO3 of 12 mg OD/mgVss-d (27), Figure 8 shows the required biomass concentration X in equilibrium with the nitrate concentration cm in the mobile region at different cim1. With an increase of cm, the required biomass concentration decreases. For cim1 ) 100 mg/L, biomass concentrations of 100 and 200 mg/L can decrease the nitrate concentration to
FIGURE 8. Required biomass concentration X for the nitrate concentration in the mobile region cm at different cim1. be less than 1 and 0.1 mg/L, respectively. In addition, the required X decreases with the decrease of cim1. Adding a sufficient amount of electron donor to the recirculated water may achieve sufficient biomass to prevent inhibition of U(VI) reduction. If such a biomass concentration cannot be established, the onset of U(VI) reduction will be delayed until the nitrate reservoir within the slowly reacting immobile domain is exhausted.
Acknowledgments This work was funded by the United States Department of Energy (DOE) Natural and Accelerated Bioremediation Research (NABIR) Biological and Environmental Research (BER), through grant number DE-F603-00ER63046. The authors appreciate the help of Paul Bayer, the NABIR program manager, and thank four anonymous reviewers for their constructive comments on the manuscript.
Literature Cited (1) Lovley, D. R.; Phillips, E. J. P. Microbial reduction of uranium. Nature 1991, 350, 413-316. (2) Lovley, D. R.; Phillips, E. J. P. Reduction of uranium by desulfovibrio desulfuricans. Appl. Environ. Microbiol. 1992, 58, 850-856. (3) Lovley, D. R.; Phillips, E. J. P. Bioremediation of uranium contamination with enzymatic uranium reduction. Environ. Sci. Technol. 1992, 26, 2228-2234. (4) Gorby, Y. A.; Lovley, D. R. Enzymatic uranium precipitation. Environ. Sci. Technol. 1992, 26, 205-207. (5) Luo, J.; Wu, W.-M.; Fienen, M. N.; Jardine, P. M.; Mehlhorn, T. L.; Watson, D. B.; Cirpka, O. A.; Criddle, C. S.; Kitanidis, P. K. A nested-cell approach for in situ remediation. Ground Water (in press). (6) Wu, W.-M.; Carley, J.; Fienen, M.; Mehlhorn, T.; Lowe, K.; Nyman, J.; Luo, J.; Gentile, M. E.; Rajan, R.; Wagner, D.; Hickey, R. F.; Gu, B.; Watson, D.; Cirpka, O. A.; Kitanidis, P. K.; Jardine, P. M.; Criddle, C. S. Field-scale bioremediation of uranium in a highly contaminated aquifer I: Conditioning of a treatment zone. Environ. Sci. Technol. Submitted. (7) Finneran, K. T.; Housewright, M. E.; Lovley, D. R. Multiple influences of nitrate on uranium solubility during bioremediation of uranium-contaminated subsurface sediments. Environ. Microbiol. 2002, 4, 510-516.
(8) Abdelouas, A.; Lu, Y.; Lutze, W.; Nuttall, H. E. Reduction of U(VI) to U(IV) by indigenous bacteria in contaminated groundwater. J. Contam. Hydrol. 1998, 35, 217-233. (9) Senko, J. M.; Istok, J. D.; Suflita, J. M.; Krumholz, L. R. In-situ evidence for uranium immobilization and remobilization. Environ. Sci. Technol. 2002, 36, 1491-1496. (10) Istok, J. D.; Senko, J. M.; Krumholz, L. R.; Watson, D.; Bogle, M. A.; Peacock, A.; Chang, Y. J.; White, D. C. In situ bioreduction of technetium and uranium in a nitrate-contaminated aquifer. Environ. Sci. Technol. 2004, 38, 468-475. (11) Solomon, D. K.; Moore, G. K.; Toran, L. E.; Dreier, R. B. A hydrologic framework for the Oak Ridge Reservation; ORNL/ TM-12026 Environmental Sciences Division, Publication No. 3815; U.S. Department of Energy: Oak Ridge, TN, 1992. (12) Jardine, P. M.; Jacobs, G. K.; Wilson, G. V. Unsaturated transport processes in undisturbed heterogeneous porous media: I. Inorganic contaminants. Soil Sci. Soc. Am. J. 1993, 57, 945-953. (13) Haggerty, R.; Gorelick, S. M. Multiple-rate mass transfer for modeling diffusion and surface reactions in heterogeneous media. Water Resour. Res. 1995, 31, 2383-2400. (14) Fienen, M. N.; Kitanidis, P. K.; Watson, D. B.; Jardine, P. M. An application of Bayesian inverse methods to vertical deconvolution of hydraulic conductivity in a heterogeneous aquifer at Oak Ridge National Laboratory. Math. Geol. 2004, 36, 101-126. (15) Coats, K. H.; Smith, B. D. Dead-end pore volume and dispersion in porous media. Soc. Petrol. Eng. J. 1964, 4, 73-81. (16) van Genuchten, M. T.; Wierenga, P. J. Mass-transfer studies in sorbing porous-media: 1. Analytical solutions. Soil Sci. Soc. Am. J. 1976, 40, 473-480. (17) Gerke, H. H.; van Genuchten, M. T. Evaluation of a 1st-order water transfer term for variably saturated dual-porosity flow models. Water Resour. Res. 1993, 29, 1225-1238. (18) Ball, W. P.; Roberts, P. V. Long-term sorption of halogenated organic-chemicals by aquifer material: 2. Intraparticle diffusion. Environ. Sci. Technol. 1991, 25, 1237-1249. (19) Cunningham, J. A.; Roberts, P. V. Use of temporal moments to investigate the effects of nonuniform grain-size distribution on the transport of sorbing solutes. Water Resour. Res. 1998, 34, 1415-1425. (20) Brusseau, M. L.; Jessup, R. E., Rao, P. S. C. Modeling the transport of solutes influenced by multiprocess nonequilibrium. Water Resour. Res. 1989, 25, 1971-1988. (21) McKenna, S. A.; Meigs, L. C.; Haggerty, R. Tracer tests in a fractured dolomite 3. Double-porosity, multiple-rate mass transfer processes in convergent flow tracer tests. Water Resour. Res. 2001, 37, 1143-1154. (22) Cirpka, O. A.; Kitanidis, P. K. An advective-dispersive stream tube approach for the transfer of conservative-tracer data to reactive transport. Water Resour. Res. 2000, 36, 1209-1220. (23) Lomax, H.; Pulliam, T. H.; Zingg, D. W. Fundamentals of Computational Fluid Dynamics; Springer: New York, 2001. (24) Jardine, P. M.; Wilson, G. V.; Luxmoore, R. J. Modeling the transport of inorganic ions through undisturbed soil columns from two contrasting watersheds. Soil Sci. Soc. Am. J. 1988, 52, 1252-1259. (25) Betlach, M. R.; Tiedje, J. M. Kinetic explanation for accumulation of nitrite, nitric-oxide, and nitrous-oxide during bacterial denitrification. Appl. Environ. Microbiol. 1981, 42, 1074-1084. (26) Almeida, J. S.; Reis, M. A. M.; Carronda, M. J. T. A unifying kinetic model of denitrification. J. Theor. Biol. 1997, 186, 241249. (27) Rittmann, B. E.; McCarty, P. L. Environmental Biotechnology: Principles and Applications; McGraw-Hill: New York, 2001.
Received for review January 28, 2005. Revised manuscript received August 25, 2005. Accepted August 29, 2005. ES050195G
VOL. 39, NO. 21, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
8459
