Environ. Sci. Technol. 2008, 42, 2793–2798
Organic Matter and Modeling Redox Reactions during River Bank Filtration in an Alluvial Aquifer of the Lot River, France MONIKA A. M. KEDZIOREK,* STEPHANE GEOFFRIAU, AND ALAIN C. M. BOURG Environmental Hydro Geochemistry, Department of Earth Sciences, University of Pau, BP 1155, 64013 Pau Cedex, France
Received September 25, 2007. Revised manuscript received December 23, 2007. Accepted January 24, 2008.
A 3 year study of the infiltration of Lot River water into a well field located in an adjacent gravel and clay alluvial aquifer was conducted to assess the importance of organic matter regarding the redox processes which influence groundwater quality in a positive (denitrification) or negative (Mn dissolution) manner. Chloride was used to quantify the mixing of river water with groundwater. According to modeling with PHREEQC, the biodegradation of the infiltrated dissolved organic carbon (DOCi) is not sufficient to explain the observed consequences of the redox reactions (dissolved O2 depletion, denitrification, Mn dissolution). Another electron donor source must therefore be involved: we propose solid organic carbon (SOC) as a likely candidate, if made available for degradation by prior hydrolysis. Its contribution to redox reactions could be significant (30-80% of the total organic carbon consumed by redox reactions during river bank filtration). We show here also that even though the first few meters of infiltration are highly reactive, significant redox reactions can take place further in the aquifer, possibly affecting groundwater quality away from the river bank.
Introduction In alluvial settings, groundwater and river water, when they are hydraulically connected, act as a single resource (1). While under normal conditions the river can be a gaining stream, flow can be reversed during the operation of a well field, leading to infiltration of river water into the aquifer. Although this induced recharge (2) can enable the abstraction of large volumes of water near high-demand areas (3), it also can result in the infiltration of undesirable substances such as organic matter, suspended solids, microorganisms, and inorganic and organic pollutants. Several natural processes (e.g., biological activity, particle filtration, and redox reactions), grouped under the generic term of river bank filtration (RBF), are able to improve the quality of the infiltrating water by removing many of these undesirable compounds. Among the potential beneficial effects of RBF are the degradation of dissolved organic carbon (DOCi) and the lowering of nitrate concentrations by denitrification and by mixing (groundwater with river water) (e.g., see refs 4–6). Because RBF is very attractive (e.g., see ref 7), it is widely used for producing * Corresponding author phone: 33 559 40 74 16; fax: 33 559 40 74 15; e-mail:
[email protected]. 10.1021/es702411t CCC: $40.75
Published on Web 03/15/2008
2008 American Chemical Society
drinking water in many countries. At the end of the 20th century, drinking water produced by RBF was about 16% in Germany (8), 50% in the Slovak Republic, 45% in Hungary, 75% in Berlin (7), and 50% in France (3). Redox reactions observed during RBF are definitively the result of microbiological activity (e.g., see refs 8–10). They produce the energy needed for metabolic processes and cellular construction of bacteria. Redox reactions involve an organic carbon compound (acting as an electron donor) and an electron acceptor (eqs 1-5, where CH2O stands for the organic matter). The electron acceptors available in alluvial aquifers are (by decreasing standard redox potential) dissolved O2, NO3, manganese oxides, iron oxyhydroxides, and sulfate (e.g., see refs 11–13). During RBF the infiltrating river water, which usually contains dissolved O2, involves aerobic conditions in the first few centimeters to tenths of meters of the river bank. Further along the flow path, anaerobic conditions may appear when this oxygen is totally consumed (e.g., see refs 14 and 15). In most cases, dissolved O2 and NO3 are the electron acceptors used by bacteria. If there is an abundant source of organic carbon and insufficient oxygen and nitrate, other electron acceptors are involved, leading to water quality problems such as manganese and/or iron dissolution (e.g., see refs (4) and 14–16). Such undesired effects can be encountered under three circumstances: rivers carrying a high organic carbon load, rivers whose course is dammed (which increases the deposition of organic-matterrich sediments and decreases the dissolved oxygen load) and well fields located in a hydrogeological context with an impermeable or semipermeable soil cover (14, 15, 17). respiration CH2O + O2 f CO2 + H2O
(1)
denitrification 5CH2O + 4NO3- + 4H+ f 5CO2 + 2N2 + 7H2O (2) Mn reduction CH2O + 2MnO2(s) + 4H+ f 2Mn2+ + 3H2O + CO2 (3) Fe reduction CH2O + 8H+ + 4Fe(OH)3(s) f 4Fe2+ + 11H2O + CO2 (4) sulfate reduction 2CH2O + SO42- + H+ f HS- + 2H2O + 2CO2 (5) It is now recognized that organic matter is the controlling factor for redox reactions and their impact on groundwater quality (e.g., see refs 12 and 14). Infiltrating DOCi does not seem sufficient to explain these effects (3, 6). At the Aubergenville well field, 40 km downstream from Paris near Mantes in the Seine River alluvial plain (France), Doussan et al. (3), observing a marked decrease in nitrate concentration between the river and the wells, hypothesized that this decrease occurs in the river sediments. In a study of the Torgau aquifer associated with the Elbe River (Germany), Grischek et al. (6) identified denitrification in infiltrating river water using δ15N analysis, and proposed, based on a mass balance calculation using the Redfield equation (18), that the oxidizable organic C required for this denitrification comes from both the infiltrating river water and solid matter in the river bed sediments and aquifer material. In this paper, we revisit the Capdenac-Gare well field (south western France) utilizing 3 years of data and provide a quantitative explanation by modeling the consumption and production of redox reactive elements and compounds using PHREEQC (19). VOL. 42, NO. 8, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2793
originates from both the river and the aquifer—the proportion of each depending on the pumping rate, the distance between the river and the well, permeability, porosity, hydraulic heads, etc. Chloride, a mobile and conservative element (e.g., see ref 4), was used as a natural tracer to determine the extent of mixing, thus enabling us to calculate the fraction of Lot River water pumped in each bore hole. This was possible at Capdenac-Gare because the dissolved chloride contents in the Lot River and in the alluvial aquifer were both different and quite stable during the sampling campaigns. Chloride concentrations in well FE9 and in the Lot River were taken as representative of groundwater (Caquifer) and river water (CLot), respectively. Chloride concentrations in all of the bore holes results from the mixing of river water and groundwater, and the mixing ratio x is calculated using eq 6. The water in all bore holes except one piezometer (Pz2) comes mainly from the river (Table SI2 of the Supporting Information and Figure 1). Cwell ) xCLot + (1 - x)Caquifer
FIGURE 1. Capdenac-Gare well field (Aveyron, France) in 1992, from Bourg and Bertin (21) Reprinted with permission from ref 21. Copyright 1994 American Chemistry Society. White circle ) well; black circle ) piezometer; dotted lines ) piezometric levels in November 1992.
FIGURE 2. Observed DOC in the Lot River and its alluvial aquifer between 1992 and 1994 (month 1 ) February 1992). (There are less data for Pz1 because it was drilled only in July 1992.)
Site and Methods Field Site. The Capdenac-Gare well field (Figure 1), located in the Lot River alluvial plain (Aveyron, Southwestern France), has been studied and described previously (4, 20–22). The alluvions consist of gravel with clay lenses overlying marly limestone and dolomite from the Lias overlying an impermeable Lower Hettangian formation (23). The saturated and unsaturated zones are about 6 and 4 m thick, respectively. Quasi-monthly sampling (22 campaigns) was carried out between February 1992 and December 1994 in the Lot River and in all wells and piezometers in the well field (i.e., 11 sampling points; Figure 1). They included on-site measurement of pH, Eh, alkalinity, conductivity, temperature, and dissolved O2, and laboratory analyses to determine dissolved (filterable through 0.45 µm) Cl, Mn, Fe, NO3, DOC (dissolved organic carbon), and SO4, following proper conservation measures (4, 20). Mixing of River Water and Groundwater. The water pumped in a well field located in an alluvial aquifer generally 2794
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 42, NO. 8, 2008
(6)
Average groundwater residence times (6.4 ( 1.0 days in Pz1, 12.2 ( 2.4 days in FE3, and 16.5 ( 0.8 days in FE5), and velocities were obtained using radon-222 as a tracer of the infiltration of river water into the aquifer (22). Determining Concentrations of Elements Involved in Redox Reactions. The concentration of a dissolved redox reactive element or compound observed in a given bore hole is the result of both the mixing of river and aquifer water and the effect of redox reactions. Assuming a chemical steady state (i.e., invariant chemistry both at the time of sampling and over the period of time needed for the river water to reach the wells (4)), the theoretical concentration, the concentration that should be observed in each bore hole if water mixing alone between river water and groundwater (i.e., no reaction) had occurred, was calculated using the MIXING function of the geochemical code PHREEQC (19). This assumption is acceptable because even though the water chemistry of the Lot River is not stable, variations from one month to the next rarely exceed 25% (with the exception of DOC in July 1992, January 1993, and February 1993 (Figure 2 and Bourg et al. (24)) and transit from the river to the furthest point mentioned here (FE5) takes about 2 weeks. Theoretical concentrations of dissolved O2, NO3, SO4, DOC, and dissolved Mn were determined for each well and for each sampling campaign. The concentration of each redox reactive element or compound that had reacted along the flow path was calculated by subtracting those observed from theoretical concentrations. A positive value indicates that the element or compound is consumed (by biological activity, precipitation, or adsorption), whereas a negative value indicates that the element is produced (by biological activity, dissolution, or desorption). Since the oxidation of organic carbon is responsible for redox reactions, the quantity of electron acceptors consumed (determined by PHREEQC-in batch mode) is expressed in equivalent biodegraded carbon. Neither significant iron nor significant sulfate reduction was observed in the well field; therefore, only eqs 1-3 were used for the calculations.
Results and Discussion Biogeochemical Reactions Observed in the Capdenac-Gare Well Field. DOC in the Lot River and Its Alluvial Aquifer. Since the drainage basin of the Lot River is little anthropized, in agreement with the Redfield equation (expressed under its simplified form by CH2O + O2 S CO2 + H2O (18)), DOC in the river is higher during the summer due to excess photosynthetic activity compared to respiration (24) (Figure 2). Except for one event in February 1993 where DOC was higher in Pz1 (3.7 m from the river bank) than in the river,
FIGURE 3. Observed and theoretical concentrations of dissolved oxygen for boreholes Pz1, FE3, and FE5 between 1992 and 1994 (month 1 ) February 1992). Black squares and lines ) theoretical concentrations; white triangles ) observed concentrations. the DOC concentration generally decreases rapidly as distance from the river bank increases (Pz1, FE3, FE5). High reactivity (i.e., degradation) of DOCi is observed in the first meters of the infiltration flow path. In Pz1, DOC appears to follow the seasonal trends of DOC in the river (i.e., higher summer values). Redox Reactions Involving O2. When comparing theoretical and observed concentrations, dissolved O2 evolves with time in two ways (Figure 3). In bore holes near the river (e.g., Pz1 and FE3), there is an apparent seasonal difference between observed and theoretical concentrations. Observed O2 concentrations are always lower than expected, which indicates that redox reactions have occurred, but there is less difference during the winter months when water is colder (7–9 °C) than during the summer when water temperature ranges from 11 to 19 °C. This is in agreement with the DOC data in Figure 2. This temperature dependence suggests the involvement of biological activity. In bore holes farther from the river (e.g., FE5), observed O2 concentrations are lower than theoretical concentrations even during the winter. The quite constant year-round water temperature in these wells of 9–12 °C (21) probably favors a constant biological activity and dissolved O2 consumption, even in winter. Redox Reactions Involving Nitrate (Figure 4). Nitrate appears to behave like dissolved oxygen, but a spatial difference is observed. In the first meters of the infiltration path, where conditions are oxidizing (dissolved oxygen concentrations between 0.1 and 0.3 mM), observed and theoretical nitrate concentrations are similar, indicating that nitrate reduction does not take place (e.g., Pz1 and FE3). Because redox reactions are more energetic with dissolved oxygen than with nitrate, dissolved oxygen is consumed first by facultative aerobic bacteria as long as O2 concentrations remain high, and consequently denitrification is low. A difference between observed and theoretical nitrate concentrations is visible only at a distance of about 50 m from the riverbank (FE5) where less than 0.1 mM dissolved oxygen remains in the water samples. Reactions Involving Manganese Oxides. Little dissolved manganese is observed in most of the bore holes, showing that there is little or no reductive dissolution of this element (maximum value, 1.4 µM in FE3). In FE2, located 40 m from
the river, observed O2and nitrate concentrations are much lower than in FE5 (54 m from the river), indicating that redox reactions associated with bacterial activity are more prevalent near FE2 than near FE5. The high dissolved manganese concentrations observed in FE2 (13-25 µM), together with the low dissolved oxygen (median value, 0.01 mM) and nitrate (median value, 0.01 mM) concentrations, indicate that manganese oxide is used as an electron acceptor for bacterial activity. Two circumstances might explain the difference observed between FE2 and FE5: (a) the occurrence of an organic-rich zone near FE2, which would lead to greater redox reactivity (Figure 1) (when the bore holes were drilled, gray clay was observed at depths of 5.7 and 7.2 m between the Lot River and FE2 (23); (b) a lower water velocity between the river and FE2 than between the river and FE5. FE2 is located at about the same piezometric level as FE3, and the water velocity around these two wells, therefore, could be similar (ca. 0.8 m day-1) and lower than at FE5 (3.3 day-1) (22). Since contact time can influence reductive dissolution, a lower water velocity will favor manganese dissolution (e.g., see ref 4). In conclusion, as river water and its DOCi infiltrate into the aquifer, microbial activity results in a spatial distribution of groundwater chemistry in accordance with the decreasing standard redox potentials of the various electron acceptors available (O2, NO3, and Mn oxides). As expected, nitrate is consumed only when the dissolved oxygen concentration is low, and Mn oxide reductive dissolution takes place only when both dissolved oxygen and nitrate concentrations are low. Electron acceptors with high redox potentials therefore play an inhibiting role in redox reactions involving electron acceptors with lower redox potentials. Role of SOC on Redox Reactions. As indicated above, DOC infiltrating from the river induces redox reactions. As usual in infiltration systems, the Lot River is richer in DOC (0.1–0.37 mM) than the aquifer (
