Tuebingen Groundwater Research Institute (TGF), Tuebingen, Wrangelstr. 49, D-10997 Berlin, Germany; +49 (0) 30 3036-4864; email@example.com
A two-dimensional multicomponent reactive transport modeling approach was used to simulate contaminant transport and the evolution of redox processes at a large-scale kerosene-contaminated site near Berlin, Germany. In contrast to previous site-scale modeling studies that focused either on one or two contaminants or on steady-state redox conditions, multiple contaminants and electron acceptors, including mineral phase Iron (III), were considered with an evolving redox zonation. Inhibition terms were used to switch between the different electron acceptor processes in the reaction scheme. The transient evolution of redox zones and contaminant plumes was simulated for two separate transects of the site, which have different geology and ground water recharge distributions and where quite different downstream contaminant and terminal electron–accepting process (TEAP) distributions are observed. The same reaction system, calibrated to measured concentrations along one of the transects, was used in both cases, achieving a reasonable match with observed concentrations. The differences between the two transects could thus to some extent be attributed to the different hydrological and hydrogeological conditions, in particular ground water recharge distributions. Long-term simulations showed that the distribution of TEAPs evolves as Fe(III) becomes depleted, with conditions becoming increasingly methanogenic, leading to changes in contaminant plume lengths. The models were applied to assess the potential effects of planned changes in land use at the site that may affect the ground water recharge distribution. The simulated redox zonation responded strongly to changes in recharge, which in turn led to changes in the contaminant plume lengths.
Monitored natural attenuation (MNA) is increasingly applied as a remediation strategy at contaminated sites where ground water quality is adversely affected by a contaminant source. Particularly common in this respect are sites contaminated by fuel hydrocarbons, whereby the main compounds of concern are typically monoaromatic hydrocarbons (MAH), which include benzene, toluene, ethylbenzene, and xylene (BTEX) compounds as well as other alkylbenzene isomers such as trimethylbenzenes (TMB). Central issues for the implementation of MNA at a site are the questions of contaminant plume length and spatial and temporal stationarity. Key in this respect are the contaminant degradation processes, which for the implementation of MNA need to be understood and quantified as far as possible.
The oxidation of organic contaminants in ground water leads to changes in redox conditions downstream from the source zone as electron acceptors in the aqueous or solid phase are consumed in the parts of the aquifer affected by the contaminant plume. The development of redox zones in response to the transformation or degradation of organic contaminants has been recognized for at least 30 years and has been described for a large number of field sites (e.g., Chapelle et al. 2002; Christensen et al. 2000; Cozzarelli et al. 2001).
Where the availability of organic substrate and nutrients is not a limiting factor for microbial growth, the utilization of electron acceptors appears to occur sequentially, with the thermodynamically most favorable electron acceptors becoming depleted before the less favorable; dissolved oxygen is consumed first, after which nitrate, Fe(III) mineral phases, and sulfate are consumed prior to the onset of methanogenesis. In a contaminated aquifer, this translates to a distinct spatial zonation of the redox processes downstream from the contaminant source, with oxygen and nitrate reduction as the dominant terminal electron–accepting processes (TEAPs) at the fringes of the contaminant plume, giving way to a zone where Fe(III) reduction dominates followed by sulfate reduction and finally methanogenesis occurring closest to the source zone (e.g., Lovley et al. 1994). While this typical sequential picture of electron acceptor utilization implies the development of clearly defined redox zones, overlapping of anaerobic redox zones has been reported at a number of sites, with simultaneous consumption of sulfate and Fe(III) (Postma and Jakobsen 1996; Vencelides et al. 2007) and Fe(III) consumption and methanogenesis (Chapelle et al. 2002; Cozzarelli et al. 2001).
Solid-phase electron acceptors, in particular Fe(III) mineral phases, play an important role both in the degradation of contaminants and in the temporal and spatial evolution of redox zonation. It has been shown that as Fe(III) minerals become depleted in the aquifer sediments, redox conditions in the plume change, with Fe(III) reduction giving way to methanogenesis as the dominant TEAP (Chapelle et al. 2002; Cozzarelli et al. 2001). Meanwhile, reoxidation of reduced iron species may significantly influence the spatial availability of electron acceptors for contaminant degradation (Vencelides et al. 2007).
The supply of electron acceptors to the contaminant plume, and hence the redox zonation, is also influenced by local hydrogeological and hydrological conditions; Bjerg et al. (1995) reported differences between the observed redox zonation downstream from a landfill for two parallel transects separated by 30 m, attributed in part to an uneven leaching of electron acceptors from the landfill, while Scholl et al. (2006) showed that the TEAP distribution in a contaminant plume can be influenced by the supply of electron acceptors from vertically infiltrating ground water recharge.
As organic hydrocarbons degrade at different rates depending on the dominant TEAP (Aronson and Howard 1997), the spatial and temporal evolution of the redox zonation has significant implications for contaminant transport and should be considered when implementing a long-term remediation strategy such as MNA. For example, increases in plume length have been reported for benzene and ethylbenzene, which generally degrade more rapidly under Fe reducing than methanogenic conditions, at sites where local depletion of Fe(III) minerals is driving a change in the redox conditions from Fe(III) reducing to methanogenic (Chapelle et al. 2002; Cozzarelli et al. 2001). In the development of MNA concepts, numerical models are frequently used to understand and quantify contaminant emission and degradation processes and to predict the likely future behavior of contaminant plumes. In this respect, it is clearly desirable that the factors discussed previously are included in such models. Various aspects of electron acceptor availability, redox zonation, and their influence on biodegradation have been considered in numerical modeling studies, with a broad range of approaches applied to various types of contaminated sites reported in the literature.
In a relatively straightforward approach, which included a finite mineral-phase electron acceptor but did not otherwise consider redox zonation, Essaid et al. (2003) simulated BTEX degradation with temporally and spatially constant aerobic and anaerobic first-order degradation rates for the oil spill site where Cozzarelli et al. (2001) reported a growth in the benzene and ethylbenzene plumes as a result of Fe(III) depletion. When they included Fe(III) as an electron acceptor in the simulations, they observed an increase in the benzene plume length as the Fe(III) became depleted.
Studies to consider the effects of redox zonation on plume development have included both steady-state and transient approaches. In a steady-state approach using a predefined redox zonation, Lønborg et al. (2006) simulated the degradation of six MAH downstream from a landfill site. They determined first-order degradation rates for the different compounds in the different zones; however, the approach relied on the assumption of steady-state conditions with only limited overlap of the TEAPs. In a transient approach, Brun et al. (2002) simulated the development of redox zones downstream from a landfill site for a single contaminant, dissolved organic carbon (DOC), with multiple TEAPs using Monod kinetics with inhibition constants controlling the switching and overlap between the different TEAPs. They were able to show that Fe(III) reduction played a major role in the attenuation of DOC at the site and found that the inhibition constants were very important for matching the observed redox zonation, in particular for the SO42− reduction system.
As an alternative to using inhibition constants to control switching between TEAPs, Schreiber et al. (2004) used a sequential electron acceptor model with first-order kinetics to simulate degradation of BTEX compounds under transient redox conditions. They observed an increase in methanogenesis as Fe(III) was depleted, but the sequential model was unable to reproduce overlapping TEAPs observed in the field.
Other studies meanwhile have focused more specifically on the role of mineral-phase electron acceptors in the TEAP system. Prommer et al. (1999) determined the relative importance of iron and sulfate reduction at a BTEX-contaminated site by simulating the degradation of toluene in an approach considering multiple electron acceptors and including secondary reactions of reduced iron species. The effect of iron cycling reactions on electron acceptor availability was studied by Vencelides et al. (2007) who simulated the temporal evolution of redox zonation at a BTEX-contaminated site and showed that secondary reactions of reduced iron species can be a significant sink for electron acceptors.
In contrast to previous site-scale modeling studies that focused either on one or two contaminants or on steady-state redox conditions, in this work, a full set of relevant contaminants are considered with multiple electron acceptors and an evolving redox zonation. The reactive transport simulations for the evaluation of natural attenuation processes at a large-scale kerosene-contaminated site near Berlin, Germany consider Monod and first-order kinetic degradation of the contaminants with the electron acceptors and secondary reactions of reduced iron species. Using the approach first put forward by Widdowson et al. (1988), inhibition terms are used to control the transitions between the different TEAPs. The evolution of redox zones and contaminant plumes is simulated for two separate transects of the site, which have different geology and ground water recharge distributions and where quite different downstream contaminant and TEAP distributions are observed. The kinetic parameters for the reaction system are calibrated to measured concentrations along one transect. The calibrated parameters are then used in simulations for both transects. The contributions of the different TEAPs to the overall degradation, the evolution of the electron acceptor utilization, and the development of the contaminant plumes under the differing hydrological and hydrogeological conditions in the transects can then be compared. A prognosis is made for the future development of the contaminant plumes, whereby the potential effects of planned developments and changes in land use at the site, which will affect ground water recharge distributions, are assessed.
Flughafen Brand is a disused military airfield located approximately 60 km south of Berlin, Germany. The airfield was in operation from the late 1950s until the early 1990s, during which time massive contamination of the subsurface beneath the fuel transfer and storage facilities occurred (Figure 1). The tankfarm, where fuel was stored in aboveground tanks on an unsealed surface, is located in the southeast portion of the site. Immediately to the north of the tankfarm is the railhead used for fuel deliveries, which crosses the site from west to east. In the center of the site is a large area of open ground covered in parts with construction waste, and to the northwest, an unsealed landfill containing construction waste and general detritus. The remainder of the site is covered by mixed forest interspersed with derelict infrastructure. An area of approximately 110 ha is contaminated with kerosene jet fuel, present as mobile and immobile nonaqueous phase in the capillary fringe and unsaturated zone (Figure 1). The total volume of kerosene in the subsurface, estimated from the properties of the aquifer material and apparent kerosene phase thicknesses measured in monitoring wells, is on the order of 500 m3. The site is relatively well characterized, with a total of 39 conventionally bored ground water monitoring wells and fifty 1-inch direct-push multilevel monitoring wells installed since 2004. In a direct-push survey of the site, laser-induced fluorescence (LIF) soundings (U.S. EPA 1997) were carried out at 10- to 20-m intervals to accurately delineate the downstream extent of the kerosene phase. Data from these and other direct-push soundings were used to construct a detailed geological model for the site (Miles et al. 2007).
The site is located at an elevation of 55 to 65 m above sea level in a topographically flat region of Pleistocene glacial sediments. The geology at the site consists of quaternary sediments up to 100 m thick with principal units of fine, medium, and coarse sands and discontinuous layering of silt and clay. The stratigraphy is complex, forming a hydraulically connected system of aquifers over the depth of the sediments. The unsaturated zone at the site is typically about 10 m thick. In the southern part of the site, a silt unit in the unsaturated zone forms the aquitard of a perched aquifer, which in part overlies the kerosene-contaminated zone (Figure 1). This basin-like structure dipping from east to west has a significant influence on the distribution of ground water recharge and the local flow situation at the site (Miles et al. 2007). Ground water recharge to the main aquifer is intercepted over a large area by the perched aquifer, in which the flow direction is east-west. A portion of the intercepted recharge overflows to the main aquifer in a relatively small area in the center of the site, contributing to the divergent character of the south-north flow in the main aquifer, which can be seen in the hydraulic head contours in Figure 1.
The ground water contamination at the site consists mainly of MAH, that is, benzene, ethylbenzene, xylenes, ethyltoluenes, and trimethyl- and propylbenzenes; significant concentrations of toluene are not found at the site. Two contour lines for the total MAH concentration are included in Figure 1. Concentrations above 1 mg/L are observed consistently beneath the source zone. Downstream, the concentrations fall to below 0.1 mg/L within a few meters in the western part of the site, whereas in the east, the MAH plume extends much farther, with measured concentrations remaining above 0.1 mg/L for about 100 m downstream from the source zone.
An overview of the distributions of electron acceptors and reduced redox species is given by the interpolated concentrations shown in Figure 2. Because nitrate is not present at significant concentrations anywhere at the site, only the four most important species, oxygen, Fe(II), sulfate, and methane, are included. Anoxic conditions with dissolved oxygen concentrations below 1 mg/L exist beneath the source zone and extend downstream for 200 to 300 m, beyond which the concentrations rise to 3 to 4 mg/L but remain below the background (upstream from the source zone) concentrations of 5 to 6 mg/L. Oxygen is also found in the center of the site, where there is communication between the perched and the main aquifers. Fe(II) is found at high concentrations (greater than 30 mg/L) beneath the entire source zone, with concentrations falling to less than 10 mg/L within 100 to 200 m downstream, compared to concentrations below 0.2 mg/L upstream from the source zone. The Fe(II) concentrations downstream from the source zone follow a similar pattern to the MAH concentrations, extending farther in the east of the site than in the west. In the center of the site where conditions are slightly oxic, Fe(II) is not detected beyond the source area. The Fe(II) distributions indicate that Fe(III) reduction is occurring as a significant biodegradation process in the source zone and contaminant plume, with remineralization or ion exchange processes limiting the transport of Fe(II) away from the areas where it is being produced. Sulfate is significantly depleted in the eastern part of the site, with concentrations below 10 mg/L compared to the background concentrations of 60 to 70 mg/L found outside the contaminated area. To the west, the concentrations are also depleted, though to a lesser extent. In the center of the site, similarly to oxygen, an area of higher concentration is seen. For methane, the highest concentrations (18 mg/L) occur beneath the source zone in the southeast of the site, with elevated concentrations extending downstream, whereas to the west, much lower methane concentrations, mostly below 1 mg/L, are observed. Upstream from the source zone, the measured methane concentrations of 20 to 30 μg/L are 2 and 3 orders of magnitude lower than the peak concentrations.
The observed concentrations of the anaerobic redox species indicate overlapping of anaerobic redox processes at the site. Elevated Fe(II) concentrations are found together with both low sulfate concentrations and high methane concentrations, which may indicate that Fe(III) reduction is occurring concomitantly with both sulfate reduction and methanogenesis. Sulfate reduction and methanogenesis appear to be more separated, with the highest methane concentrations found only where sulfate concentrations are very low. Such overlapping of TEAPs is frequently observed at contaminated sites and may be caused by a number of processes, both abiotic and biotic. The abiotic processes include vertical mixing during sampling, advective transport or adsorption of reaction by-products, and microscale heterogeneity of mineral electron acceptor species (Schreiber et al. 2004), while the biotic might include the simultaneous utilization of electron acceptors, which has been suggested on the basis of observations in field studies of both contaminated and pristine aquifers (Jakobsen et al. 1998; Jakobsen and Postma 1999).
Overall, in terms of contaminant plume lengths and distributions of redox species, the site can be broadly divided into two different areas. The eastern part of the site, where the source zone is completely covered by the perched aquifer, that is, with reduced ground water recharge, is characterized by high methane and low sulfate concentrations, with elevated Fe(II) concentrations extending downstream from the source zone and long MAH plumes. In the western part of the site, where the downstream area of the source zone is not covered by the perched aquifer, sulfate concentrations are higher, methane concentrations are lower, elevated Fe(II) concentrations do not extend as far downstream, and MAH plumes are short.
On the basis of the observed differences in the TEAP distributions and MAH plume lengths, two-dimensional flow and reactive transport models were created for two transects, one in the east and the other in the west of the site (Figures 1 and 2). The transects are oriented approximately parallel to the ground water flow direction. Despite the large number of monitoring wells potentially available, the choice for the transects, in particular the west transect, was limited by the appearance of kerosene phase in some wells and the influence of minor, secondary contaminant sources in others, rendering them unsuitable for inclusion in the model. As a result, the transects are not oriented exactly in the assumed ground water flow direction for their entire length. Because simulating a three-dimensional flow and transport system as two-dimensional profiles introduces uncertainties by the implicit simplification of the flow field, the orientation of the transects may be an added source of uncertainty in the simulations. Both transects begin within the source zone and extend downstream to encompass the contaminant plumes. Although initial simulations were carried out with model domains extending farther upstream and encompassing the entire length of the source zone, in order to reduce computational costs for the calibration process, which focused on the downstream area, the domains were truncated and the results of the initial simulations were used to define the upstream boundary conditions. The multiphase, multicomponent reactive transport code MIN3P (Mayer et al. 2002) was used for the simulations. The following section briefly summarizes relevant theoretical aspects of the code, and the development of the models is described thereafter.
MIN3P considers variably saturated flow according to the Richards’ Equation and uses a generalized formulation for biogeochemical reactions in the aqueous phase. The reaction formulation allows the reaction progress to depend on the total aqueous concentrations and/or activities of any number of dissolved species. Any reaction order can be accommodated with respect to any of the species. Monod and inhibition terms expressed as functions of total aqueous concentrations are also included and can be used to describe microbially mediated reactions or applied to activate or deactivate reactions in response to changing geochemical conditions. The rate expression is given by (Mayer et al. 2002):
The term kkis the rate constant for the kth aqueous-phase kinetic reaction to which any of the other multiplicative terms may be applied in defining a reaction scheme for Nccomponents. Three fractional-order terms are included for the total aqueous component concentration , aqueous species concentration , and aqueous complex concentration for the jth component. The total aqueous component concentration is the sum of the aqueous species and complex concentrations. and are the activity coefficients for the aqueous species and complexes, respectively. , , and define the reaction orders with respect to the total aqueous component, species, and complex concentrations, respectively. In the Monod and inhibition terms, defines the Monod half-saturation constant, while is an inhibition constant. Bacterial growth and die-off are not included in the kinetic formulation. The affinity term, where IAPkis the ion activity product of the reaction and Kkis the equilibrium constant, may be excluded to define an irreversible reaction.
Linear sorption of aqueous-phase components is included in MIN3P as a nonkinetic equilibrium with a distribution coefficient, Kd, for the component mass in the aqueous- and the solid-phase volumes and is implicitly assumed to occur equally in all aquifer materials.
The first step in the modeling approach was the creation of flow models for the two transects. The seasonal fluctuations in ground water levels at the site are minimal (on the order of up to 10 cm), whereas observations for the past decade show longer-term fluctuations within a range of 0.5 m, however, without a change in the ground water flow direction. On the basis of these relatively stable hydraulic conditions, a steady-state flow model was considered to be adequate for the subsequent reactive transport simulations. The reaction system was then developed and calibrated to observed concentrations in the east transect. The calibrated reaction system was subsequently implemented in the west transect without further adjustment of the reaction parameters.
The model domains for the 365 × 11.3 m east transect and the 270 × 11.3 m west transect are depicted in Figure 3. Specific information for the model domains, including the exact dimensions and parameter values, is available in the online version of the article. Each transect includes four monitoring wells positioned either directly in or projected onto the transect (Figure 1). The wells within the source zone are conventional 4-inch monitoring wells, screened over a 2-m interval, while the downstream wells are 1-inch diameter installed by direct push, where multilevel ground water samples were taken using a mobile inflatable packer system with a 90-cm sampling screen. The geological parameter distributions for the transects were generated directly from the existing three-dimensional GIS-based geological model for the site (Miles et al. 2007) using a preprocessing tool developed for this task. In this way, the irregular forms that make up the stratigraphy of the five material types (fine, medium and coarse sand, silt, and clay) in the geological model could be accurately reproduced in the regular structure of the finite-volume grid (Figures 3a and 3b). The finite-volume grids for the domains consisted of 26,000 and 30,000 cells for the east and west transects, respectively. The grid spacing was 0.75 m horizontally, with local refinement in the central part of the west transect, where numerical difficulties were encountered with the flow model due to the geological parameter distribution, whereas vertically a variable spacing was used, with a minimum cell size of 0.1 m in the upper portion of the domains to account for steep geochemical gradients, increasing to 0.5 m at the base of the domain.
The boundary conditions for the model domains are depicted in Figures 3c and 3d, with values for the parameters given in Table 1. For the steady-state flow models, the ground water recharge distributions for the upper specified flux boundaries were taken from the existing three-dimensional site-scale flow model (Miles et al. 2007). The boundary heads and hydraulic conductivities for the five material types were then calibrated to measured hydraulic heads along the transects.
Table 1. Component Concentrations in Ground Water Recharge, Source Zone, and Background (Inflow Boundary Concentrations and Initial Condition for Pristine Aquifer)
Background (Pristine Aquifer)
Distribution Coefficient for Linear Sorption (ML−3/ML−3) Kd (—)
x = 240 to 365 m, mixed ground cover, concentrations assumed and adjusted in calibration. For 0 < x < 240 m over the extent of the source zone and where recharge is low due to overlying aquitard, delivery of electron acceptors via infiltration was assumed to be insignificant.
x = 0 to 125 m, infiltration through landfill and above source zone. Low levels of O2(aq) with elevated SO42(aq) concentrations due to leaching from construction waste were assumed.
x = 125 to 270 m, mixed ground cover, concentrations assumed and adjusted in calibration.
The reactive transport models consider a nondepleting source of organic contaminants at constant concentrations and homogeneously distributed in a source zone in an initially pristine aquifer with background concentrations of electron acceptors.
The four principal MAH contaminants at the site, benzene, ethylbenzene, m/p-xylene, and 124-trimethylbenzene, and a bulk DOC component were included in the simulations with four TEAPs: oxygen, sulfate and solid-phase Fe(III) reduction, and methanogenesis. Nitrate and manganese reduction were not included, because the available concentration data do not indicate that these are significant processes at the site. DOC was included as a bulk component representing soluble hydrocarbon components of the kerosene not explicitly considered in the simulations. The initial and boundary concentrations for all components in the simulations are given in Table 1.
For the aqueous-phase electron acceptors, the initial concentrations correspond to the average concentrations measured outside the contaminated area, while the concentrations in ground water recharge were calibrated. Because the upstream boundaries are located within the source zone, based on concentration profiles from the initial simulations with model domains extending farther upstream, boundary concentrations of zero were set for the uppermost part.
Gas chromatography/mass spectrometry (GC/MS) analyses of kerosene samples from different parts of the site showed differences in the kerosene composition, whereas the equilibrium concentrations calculated on the basis of the analyses were found to be lower in many cases than the concentrations detected in monitoring wells within the source zone, both of which point to spatial heterogeneity of the kerosene composition. Source zone concentrations for the four contaminants were hence calibrated in the simulations to reproduce measured concentrations in the source zone monitoring wells in the transects. The assumed concentration of DOC in the source zone was based on the sum of mole fraction–weighted solubilities for compounds identified in GC/MS analysis of the kerosene found at the site excluding those explicitly considered in the model. The longitudinal and vertical extents of the source zones in the transects, x = 0 to 235 m and x = 0 to 122 m in the east and west transects, respectively, with a thickness of 0.7 m, were based on data from the LIF survey.
The initial quantity of Fe(III) in the simulations corresponds to the average amorphous Fe(OH)3 content of aquifer sediment samples determined by HCl extraction (Jahn 2005b). Microbial microcosm experiments carried out with aquifer material from the site showed that of the total mineral and amorphous Fe(OH)3 present, the amorphous form was bioavailable and completely used by the Fe(III)-reducing bacteria found in the aquifer, whereas the crystalline forms were barely used (Jahn 2005a). There was a large variability in the amorphous Fe(III) contents determined for the 21 samples of aquifer material analyzed, which did not show any trend that could be correlated to sampling locations (uncontaminated area, plume, or source zone) or sediment type. Thus, in the simulations, an average value was used and a homogeneous distribution in the aquifer assumed. In using the average present-day Fe(III) content as an initial condition for the pristine aquifer, it is assumed that, although the average Fe(III) content of the sediments has undoubtedly decreased due to iron reduction to date, this change is small compared to the variance for the samples used to calculate the average. In fact, the maximum depletion of Fe(III) in the model domains after 40 years was found to be approximately 30% of the standard deviation for the iron contents of the analyzed samples. Values of the distribution coefficient Kd used to represent linear sorption in the model were calculated for each component from the organic carbon-water partition coefficient Koc and the average organic carbon content for nine samples of aquifer material.
The reactions considered in the simulations are listed in Table 2. Reactions are included for each of the five organic species for each of the TEAPs, with a mixture of Monod and first-order kinetics. Monod kinetics, which are a more accurate representation of biodegradation reactions than first-order kinetics where a wide range of substrate concentrations is to be considered, were applied for reactions with oxygen and sulfate. Methanogenesis, for which it was not necessary to consider the availability of electron acceptors, was simulated using first-order kinetics approximated by setting an extremely high value for the half-saturation concentration in the Monod expression. Degassing of methane as well as secondary reactions that it may undergo, such as reoxidation by ferric oxides or hydroxides (Baedecker et al. 1993), were not considered. On the basis of the laboratory microcosm experiments (Jahn 2005b), where it was found that the degradation of contaminants with Fe(III) reduction could be adequately described by a simple first-order kinetic expression, the reaction of aqueous-phase substrates with Fe(III) was represented in the model using first-order kinetics. Aqueous phase Fe(II) ions produced by the reduction of Fe(III) minerals are known to take part in numerous secondary reactions such as precipitation as Fe(II) sulfides or reoxidation to Fe(III) (Tuccillo et al. 1999; Vencelides et al. 2007) and in ion exchange processes (Appelo et al. 1999; Freedman et al. 1994) that limit their transport away from areas where Fe(III) reduction is active. In the simulations, the complex geochemistry of Fe(II) was simplified to two processes: reoxidation of Fe(II) to Fe(III) in the presence of oxygen and a first-order decay term applied to Fe(II) to represent the various other sink reactions. Ion exchange processes were considered by including linear sorption of Fe(II) with an assumed distribution coefficient (Table 1). The reaction of Fe(II) with oxygen was considered a separate process, because it was found that the resulting depletion of oxygen affected the transport of the organic contaminants in the simulations.
Table 2. Reactions Included in the Simulations
Note: The kinetic parameters correspond to those in the generalized formulation of the reaction system (Equation 1). kkis the calibrated reaction rate (mol/L H2O/s) or (1/s), the half-saturation constant for substrate and in brackets for the electron acceptor (mol/L). is the aqueous or mineral inhibition constant, (mol/L) for aqueous or (—) for mineral phase. (aq) denotes a dissolved aqueous species, and (m) a mineral.
Inhibition constant for inhibition by O2(aq).
Inhibition constant for inhibition by Fe(OH)3(m), units (—).
Inhibition constant for inhibition by SO42−(aq).
First-order decay rate applied to Fe2+(aq), units (1/s).
The inhibition terms in the generalized reaction scheme (Equation 1) were used to control switching between the different TEAPs by inhibiting the reactions to various degrees in the presence of other electron acceptors. Fe(III) reduction was considered to be inhibited by the presence of oxygen, sulfate reduction by oxygen and the mineral phase Fe(III), and methanogenesis by oxygen, Fe(III), and sulfate.
Results and discussion
The reactive transport models for the two transects share a single reaction scheme, developed for the east transect. The model was manually calibrated to observed concentrations along the east transect by adjusting reaction rates, inhibition constants, and the concentrations of electron acceptors in ground water recharge. The reaction scheme was then implemented in the model for the west transect, which was calibrated to observed concentrations by adjusting only the concentrations of electron acceptors in ground water recharge. Source zone concentrations for the contaminants were calibrated to measured concentrations as described in the previous section. The parameters for the calibrated reaction scheme are given in Table 2. Half-saturation constants for the Monod-controlled reactions were assigned assumed values of 10% of the maximum or background concentration for the substrates and electron acceptors, respectively, with the aim of ensuring that the reaction kinetics would not tend too strongly to first or zero order over the concentration ranges used. For aerobic degradation, which for MAH is rapid compared to anaerobic processes, a maximum utilization rate was assigned such that the maximum reaction rates in the simulations were significantly higher than those for the anaerobic processes. For Fe(III) reduction, laboratory-determined first-order degradation rates for the organic components were found to be too high when applied in the simulations, and so values for the rates were determined in the calibration, with an average rate used for DOC. The relative magnitudes of the rates, which were similar for all the compounds except ethylbenzene, which was found to be significantly less degradable, were not changed in the calibration. The absolute values for the calibrated rates were two orders of magnitude lower than those determined in the laboratory but fall within the ranges of first-order rates reported in field studies for benzene, ethylbenzene, and xylene (Aronson and Howard 1997). The reason for the large difference is not clear. However, the rates determined in the laboratory apply to a microcosm where a small quantity of aquifer material is mixed completely with water containing dissolved-phase contaminants. In the model, these rates are then applied to a bulk volume in the model cells, implicitly assuming a similar complete mixing and Fe(III) reduction for the entire volume of the cell. In reality, it is likely that the Fe(III) reduction is taking place only in limited areas (microsites) within such a volume. For sulfate reduction, maximum utilization rates were determined by calibration, resulting in similar rates for benzene, m/p-xylene, and 124-TMB and a somewhat higher rate for ethylbenzene. First-order rates for methanogenesis were also determined by calibration; however, as it is widely reported, benzene is poorly degraded in comparison to other monoaromatic compounds under methanogenic conditions (Aronson and Howard 1997), subject to the constraint that the degradation rate for benzene was lower than those for the other compounds. The calibrated rates for the remaining compounds were somewhat higher than the rates for Fe(III) degradation, for which a similar trend is seen in the data reported by Aronson and Howard (1997). For the secondary reactions of Fe(II), it was assumed that in comparison to the biodegradation reactions, oxidation by dissolved oxygen was effectively an instantaneous process. The rate for the first-order decay of Fe(II) was determined by calibration and was found to be important for reproducing the observed concentrations downstream from the source zone, indicating that secondary reactions such as sulfide formation are a significant sink for Fe(II) downstream from the source zone.
Similar to Brun et al. (2002), the inhibition constants, which control the degree of separation or overlap of the TEAPs, were found to be critical for reproducing the measured concentrations and the different TEAP distributions in the two transects. For inhibition of the anaerobic processes by oxygen, a very low concentration was used for the inhibition constant, ensuring that the reactions were suppressed until dissolved oxygen concentrations approached zero. Using inhibition by Fe(III) and sulfate, it was possible to reproduce the observed situation where high concentrations of methane or sulfate are found together with high Fe(II) concentrations, but high concentrations of methane and sulfate are not found together. The calibrated inhibition constants for inhibition of sulfate reduction and methanogenesis allow both processes to occur whereas mineral-phase Fe(III) is present, with sulfate reduction inhibited significantly more than methanogenesis, while methanogenesis is additionally strongly inhibited by sulfate. Used in this way, the inhibition constants act as a lumped parameter combining the biotic and abiotic factors that may be responsible for the apparent overlap of the redox processes.
The electron acceptor concentrations applied to ground water recharge in the simulations are given in Table 1. Where the main aquifer was overlain by the perched aquifer and recharge to the main aquifer is minimal, it was assumed that the supply of electron acceptors was insignificant. Over the extent of the source zone, it was assumed that recharge reaching the main aquifer would be anoxic due to biodegradation in the unsaturated zone, which has been shown to be the case for sites contaminated with light nonaqueous phase liquids (LNAPL) (Chaplin et al. 2002; Essaid et al. 2003). Elsewhere, the oxygen concentration was assumed to be the same as the background concentration in the aquifer. The concentrations of sulfate at the site were assumed to be variable depending on the ground cover, with high concentrations due to leaching from construction waste in the landfill area of the west transect and lower concentrations in wooded areas (Figure 1).
The simulated and measured concentrations of the organic components and anaerobic redox species at monitoring wells in the east and west transects are shown in Figures 4 and 5, respectively. The simulated concentrations shown are vertical profiles after 40 years’ simulation time, which is considered to be a reasonable current age for the contamination based on the history of the site, with average concentrations shown for the screened sampling intervals. The locations of the monitoring wells in the transects can be seen in Figures 1 and 3. Looking first at the measured contaminant concentrations, it can be seen that in the east transect, benzene forms the longest plume, being the only one of the four contaminants detected at GWM5, the farthest downstream of the four monitoring wells. The remaining contaminants all have similar plume lengths, being present at significant concentrations close to the source zone at D9 and at low concentrations 50 m further downstream at E13. In the west transect, a quite different situation is seen, with shorter plumes for all of the contaminants. Generally, the contaminant concentrations fall very rapidly downstream from the source zone; ethylbenzene is not detected in any of the downstream monitoring wells, whereas at A20, only benzene and 124-TMB are present at low concentrations (less than 10 μg/L). None of the contaminants are detected at GWM1, about 100 m downstream from the source zone, with the exception of 124-TMB, where the concentrations of 1 to 2 μg/L are close to the detection limit.
Looking at the simulated concentrations for the east transect, a generally reasonable match is seen between the measured and the simulated concentrations for all four compounds. The model successfully reproduces the general characteristics of the observed situation with the long benzene plume and shorter plumes for the other compounds, but somewhat underestimates the plume length for ethylbenzene, with simulated concentrations at E13 much lower than the measured. The measured concentrations for some of the sampling intervals, however, are higher than in D9, which cannot be accounted for by the model. The concentrations for benzene in the lowest sampling intervals at E13, which lie below a thick silt layer, are also underestimated in the model, indicating that there may be in reality more flow to the deeper parts of the aquifer than is seen in the flow model. For the west transect, the model again reproduces the general characteristics of the observed situation, with no downstream plume for ethylbenzene and short plumes for the other compounds. Concentrations are overestimated close to the source zone for benzene at A19 and for benzene, xylene, and TMB in the lower intervals at A20. The proximity of these wells to the downstream edge of the source zone makes them particularly sensitive to uncertainties in the lateral extent of the source zone or heterogeneities in its composition.
For the redox species, in the east transect, the situation with low sulfate concentrations and high Fe(II) and methane concentrations is reproduced quite successfully, although the simulated concentrations of methane beneath the source zone at HT8/02 are lower than the measured, as are the Fe(II) concentrations adjacent to the source zone at D9. In the west transect, the model reproduces the general features of the observed situation quite well, with higher sulfate and lower methane concentrations than the east transect. The simulated methane concentrations are generally slightly higher than the measured, whereas the simulated sulfate concentration below the source zone at Bdep4/98 is significantly lower than the measured value. For Fe(II), the simulated concentrations adjacent to the source zone at A19 are, similarly to the east transect, lower than the measured. That this is the case in both transects suggests that the first-order decay term applied to Fe(II) to account for the various geochemical sinks may be resulting in too much mass loss at higher concentrations. Using a zero-order term for the decay was also tried, but this was found to be less effective than the first-order term. The model could also not account for the measured Fe(II) concentrations at GWM1, which are equal to or higher than those at A20. It is possible in this case that there is a separate source of Fe(II), possibly related to the landfill or an assumed second source zone in the vicinity of the well southwest of GWM1.
To look at the medium- to long-term evolution of conditions in the two transects, the calibrated models were run for a period of 400 years. Although it is considered likely that for such a time scale, particularly the soluble MAHs will be depleted in the source zone and the contaminant emission will decrease accordingly, a nondepleting source at constant concentration was assumed, being a conservative approach in terms of MNA.
Figure 6 shows the temporal development of plume lengths from the downstream edge of the source zone for the two transects, together with the contributions of the four TEAPs to the total degradation.
The TEAP contributions (Figures 6a and 6b) show that anaerobic processes dominate the degradation in both transects throughout the simulated period, with the contribution from aerobic degradation at the plume fringes less than 10%. Of the anaerobic TEAPs, sulfate reduction accounts for the largest fraction of the total degradation, remaining fairly constant at about 50% of the total in the east transect and 70% in the west transect, which is comparable to a large number of field studies reported by Wiedemeier et al. (1999). In the east transect, this occurs principally at the upstream plume fringe, where there is a supply of sulfate from water entering the model domain across the upstream boundary. In the west transect, the supply of sulfate from ground water recharge accounts for the additional contribution to degradation. This additional supply of sulfate is also responsible for the smaller contribution of methanogenesis to the degradation in the west transect than the east, because it inhibits methanogenesis in the source zone. The contribution of Fe(III) reduction is similar in both transects, though slightly higher in the east due to the larger source zone, and decreases over time as Fe(III) becomes depleted in the source zone. The two-step changes in Fe(III) reduction seen for the east transect (after 60 and 120 years) correspond to the complete depletion of Fe(III) in different regions of the source zone with different material types; because the porosities and hence volumetric Fe(III) contents differ, the complete depletion of Fe(III) occurs at different times for different material types. As the iron becomes depleted, the contribution of methanogenesis increases in the source zone and body of the plume where the supply of sulfate, which would otherwise be the next TEAP in the series, is restricted.
Figure 6c shows that in the east transect, plumes develop rapidly for all four contaminants, reaching maximum lengths of 10 to 50 m within 50 years for ethylbenzene, xylene, and 124-TMB, whereas the benzene plume reaches a length of 120 m. These are comparable with the plume lengths currently measured at the site; for benzene, concentrations of 5 to 10 μg/L at D9 suggest a plume length greater than 95 m for a 5-μg/L threshold, whereas for the other components, the measured concentrations at D9 and E13 indicate plume lengths of between 10 and 60 m for the same threshold concentration. After 50 years, as Fe(III) becomes depleted in the source zone and conditions become more methanogenic, a gradual increase is seen in the plume length for benzene, which degrades less well under methanogenic conditions, while the plumes for xylene and 124-TMB, which in the calibrated reaction scheme degrade more rapidly under methanogenesis, begin to shrink. In the west transect, where conditions are initially less methanogenic and electron acceptors are supplied by ground water recharge, shorter plumes develop initially for benzene, xylene, and 124-TMB, with the same behavior seen as Fe(III) becomes depleted in the source zone, in that the plume lengths increase for benzene and decrease for xylene and 124-TMB. For ethylbenzene, which in the calibrated reaction scheme is more degradable than the other contaminants under sulfate-reducing conditions, a plume develops only after 80 years, once sulfate concentrations have been significantly depleted in the wake of the other plumes. In the east transect, where sulfate is rapidly depleted, the ethylbenzene plume develops from the beginning of the simulation. As with the east transect, the simulated plume lengths are comparable with those inferred from the measured concentrations; for benzene, the observed concentrations of 5 to 10 μg/L at A20 suggest a plume length of 15 to 20 m for a 5 μg/L threshold. For xylene and 124-TMB, the concentrations measured at A19 and A20 indicate current plume lengths of 10 to 20 m, somewhat smaller than the simulated, whereas for ethylbenzene, the measured concentrations at A19 are close to the detection limit, indicating that there is currently no significant plume. Changes in contaminant plume length in response to changing TEAPs as Fe(III) becomes depleted have been reported at field sites and in modeling studies (Chapelle et al. 2002; Cozzarelli et al. 2001; Essaid et al. 2003), with increases in plume lengths for benzene and ethylbenzene. The simulations described in this work similarly show this behavior in the medium to long term under the assumption of steady-state flow conditions with a constant contaminant source. They also show that the converse may also occur in that a contaminant that degrades more rapidly under a subsequent TEAP exhibits plume shrinkage as Fe(III) becomes depleted.
The models for the two transects were then applied to assess the potential impact of possible future developments, which relate to a large neighboring recreational facility. Two possible options are likely in particular to impact the distribution of ground water recharge, the first being replacement of the topsoil and redevelopment of the site as a camping ground and the second being paving the site for use as a parking area. These two scenarios were considered in simulations by changing the rate of ground water recharge after 45 years for all parts of the profiles. For the camping ground scenario, a rate of recharge of 130 mm/yr was assumed, which based on the Bagrov method (Bagrov 1953) used for the hydrogeological site model is consistent with grass ground cover for the site. For the paved parking lot scenario, a minimal ground water recharge of 10 mm/year was assumed. The results are shown in Figure 7. For the camping ground scenario (Figures 7a and 7b), the assumed change in recharge has virtually no effect in the east transect, where the recharge distribution is dominated by the perched aquifer. In the west transect, where the change from mixed ground cover to grassland results in a decrease in recharge of between 30 and 70 mm/yr for most parts of the transect, an increase in the plume length is seen for all four contaminants as the supply of sulfate and oxygen to the plume is reduced. The greatest increase is seen for benzene, with an increase in the plume length of around 20 m.
For the parking lot scenario (Figures 7c and 7d), the assumed change in recharge is much greater than for the camping ground scenario. For the east transect, where the change in recharge applies to only a small part of the transect, the effect is once again minimal with a small increase (less than 5 m) in the benzene plume length. In the west transect, the change in recharge results in a significant reduction in the supply of electron acceptors, in particular sulfate, to the plume. This is reflected in the response of the simulated contaminant plume lengths, with initial increases of between 20 and 40 m for xylene, 124-TMB, and ethylbenzene, whereas for benzene, the plume length increases by more than 80 m, reaching the downstream boundary of the model domain. Following this initial response, in the absence of sulfate, methanogenesis subsequently increases, leading to a slight decrease in the plume lengths for xylene, 124-TMB, and ethylbenzene over the following 40 years.
Summary and conclusions
At a large-scale kerosene-contaminated site, reactive transport models were developed to investigate observed differences in contaminant plume lengths and TEAP distributions in different parts of the site. In contrast to previous site-scale modeling studies that focused either on one or two contaminants or on steady-state redox conditions, in this work, a full set of relevant contaminants were considered with multiple electron acceptors and an evolving redox zonation. Models were developed for two transects of the site with a reaction scheme calibrated to measured concentrations of contaminants and redox species along one of the transects. The observed overlapping of anaerobic TEAPs at the site could be successfully simulated using Monod-type inhibition terms. With the same reaction scheme used in both models, the differences in contaminant and TEAP distributions for the two transects can be attributed to some degree to hydrogeological and hydrological differences. In particular, the supply of electron acceptors in ground water recharge was found to play a critical role in the distribution of TEAPs at the site, with more methanogenesis and less sulfate reduction in the transect where an overlying perched aquifer prevented recharge to the main aquifer. This resulted in longer contaminant plumes, particularly for benzene, which degrades poorly under methanogenic conditions, compared to the second transect where the supply of sulfate in ground water recharge inhibited methanogenesis in the source zone and allowed more sulfate reduction in the plume. The important role of the supply of electron acceptors via ground water recharge was also demonstrated by simulations to assess the potential effects of changes in the recharge distribution that may result from a change in land use at the site. These showed that the simulated contaminant plumes respond rapidly to changes in ground water recharge, with in part a significant increase in plume lengths as the supply of electron acceptors to the plume is reduced. Long-term simulations for a further 350 years for the two transects showed that even under constant boundary conditions, the distribution of TEAPs and the contaminant plume lengths are not steady state, with benzene plume lengths increasing as Fe(III) is depleted in the vicinity of the source zone and conditions evolve from iron reducing to methanogenic. This transient plume behavior in the medium to long term even under constant recharge conditions and the sensitivity of the redox zonation and plume lengths to potential changes in ground water recharge are essential factors to be considered for the implementation of MNA at the site.
This work was funded by the German Ministry of Education and Research (BMBF) under grant 02WN0352 as a part of the KORA priority program, subproject 1.2 and additional funding from the Brandenburgische Boden Gesellschaft für Grundstücksverwaltung und –verwertung mbH (BBG). The authors would like to thank the three anonymous reviewers for their helpful comments.