Sorption and Degradation of the Herbicide 2-Methyl-4,6-dinitrophenol

dimensional monitoring network installed in the aquifer downgradient of the injection. The sorption and degradation of DNOC were evaluated based on mo...
0 downloads 3 Views 305KB Size
Download PDF
Environ. Sci. Technol. 2001, 35, 4789-4797

Sorption and Degradation of the Herbicide 2-Methyl-4,6-dinitrophenol under Aerobic Conditions in a Sandy Aquifer in Vejen, Denmark

TABLE 1. Molecular Structure and Relevant Physical Chemical Parameters for DNOC

METTE M. BROHOLM,* NINA TUXEN, KIRSTEN RU ¨ GGE,† AND POUL L. BJERG Environment & Resources DTU, Groundwater Research Centre, Technical University of Denmark, DK-2800, Lyngby, Denmark

A pulse (7 days) and a continuous (216 days), natural gradient field injection experiment with herbicides, including 2-methyl-4,6-dinitrophenol (4,6-dinitro-o-cresol, abbreviated DNOC), and a bromide tracer were conducted in a shallow, aerobic aquifer near Vejen, Denmark. The pulse and continuous plume were monitored in a dense, threedimensional monitoring network installed in the aquifer downgradient of the injection. The sorption and degradation of DNOC were evaluated based on moment analysis of breakthrough curves, cross sections, and snapshots of the DNOC plume and supported by results from laboratory experiments. Significant and spatially variable sorption of DNOC (Kd range, 0.10-0.98 L/kg) was observed due to a specific binding of DNOC to clay minerals. The spatial variation was mainly a result of variation in pH, with stronger sorption at lower pH, whereas other factors such as cation composition on the solid matrix appeared to be negligible. Significant degradation of DNOC in the aquifer was revealed by moment analysis of data from the continuous field injection experiment. Degradation of DNOC in the field was slow and/or subject to long lag phases, and the data suggested spatially varying degradation potentials. This was supported by the laboratory experiments. The potential for natural attenuation of DNOC in aerobic aquifers appears promising.

Introduction 2-Methyl-4,6-dinitrophenol (4,6-dinitro-o-cresol, abbreviated DNOC) has been one of the most commonly used nitroaromatic herbicides, the so-called yellow herbicides. Although the hormone active herbicides (phenoxy acids) have been the dominant herbicides in Denmark, the nitroaromatic herbicides have been used extensively as less selective herbicides. Nitroaromatic herbicides are among the most toxic herbicides that have been commonly applied in Denmark in the past (1). DNOC has been encountered in groundwater from shallow aquifers in Denmark (2). DNOC is a weak acid and present primarily in its anionic form at pH > pKa (4.3). The molecular structure and selected compound characteristics of relevance are given in Table 1. The anionic form is not expected to sorb due to its high * Corresponding author phone: +45 4525 1475; fax: +45 4593 2850; e-mail: [email protected]. † Present address: NIRAS Consulting Engineers and Planners A/S, DK-3450 Allerød, Denmark. 10.1021/es010096c CCC: $20.00 Published on Web 11/16/2001

 2001 American Chemical Society

aqueous solubility and the common negative charge of clay minerals. In low pH groundwater aquifers, part of the DNOC is present as the neutral (undissociated) form. Due to the low hydrophobicity (expressed by the log Kow value) of DNOC, hydrophobic sorption of DNOC is expected to be very limited in low organic carbon (foc) aquifers. However, the neutral forms of a large number of nitroaromatic compounds, including DNOC, can be sorbed significantly in low foc aquifer materials due to a specific binding mechanism with clay minerals that involves an electron donor-acceptor (EDA) complex (3, 4). Experiments with manipulated conditions showed that this EDA complex formation depends strongly on the structure of the compound, the clay minerals, the type of exchangeable cations, and the pH for ionizable nitroaromatic compounds such as DNOC (3, 5, 6). However, which of the factors that control the sorption under natural conditions is currently unknown. Sheremata et al. (7) observed significant sorption of the nitroaromatic compound 2,4,6-trinitrotoluene to the clay mineral illite (pH 6) and negligible sorption to Borden aquifer sand (low foc, negligible clay content, pH 8). Tuxen et al. (8) observed significant sorption of DNOC to Vejen aquifer sand (low foc, low clay content, and inlet pH 6.4) in a laboratory column study. Little is known about aerobic degradation of DNOC under aquifer conditions. Most of the literature regarding degradation of DNOC is related to topsoils (e.g., refs 1 and 9-11), where bacteria capable of metabolizing DNOC have been isolated. In these experiments, nitrite was identified as a metabolite (9). Other probable aerobic degradation products are the corresponding cresols such as 2-methyl-4-nitrophenol, 2-methyl-6-nitrophenol, and 2-methylphenol (12); however, degradation pathways are not well-documented. The only previous studies of DNOC degradation with aerobic groundwater aquifer materials we are aware of are the laboratory column study by Tuxen et al. (8) and the microcosm study by Ru ¨ gge et al. (13). These experiments were carried out with aquifer material and groundwater from the Vejen site, Denmark, in conjunction with the current study. Degradation of DNOC was observed in the columns but only in nutrient-amended microcosms. Two natural gradient field injection experiments, one pulse and one continuous, were conducted in the shallow aerobic groundwater aquifer at Vejen, Denmark. The continuous injection was conducted in the same field plot as the pulse injection about 6 months later. Supplementary laboratory batch experiments were performed with groundwater and aquifer material from the site to aid in the understanding of the mechanisms involved and the interpretation of the field data. The aim was to provide field data concerning migration and degradation of DNOC under aerobic condiVOL. 35, NO. 24, 2001 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

4789

FIGURE 1. Plan view of injection site (Vejen, Denmark) with illustration of sampling points and bromide plume at day 175. tions. DNOC was injected together with the herbicides MCPP (Mecoprop, (()-2-(4-chloro-2-methylphenoxy)propanoic acid), dichlorprop ((()-2-(2,4-dichlorophenoxy)propanoic acid), bentazone (3-(1-methylethyl)-1H-2,1,3-benzothiadiazin-4(3H)-one-2,2-dioxide), and isoproturon (N,N-dimethylN′-[4-(1-methylethyl)phenyl]urea) and BAM (2,6-dichlorbenzamide), a dichlobenil (2,6-dichlorobenzonitrile) metabolite. The herbicides were injected at low concentration levels, and their migration was monitored in three dimensions for ∼9 months. This paper describes the migration, sorption, and degradation of DNOC observed in the field injection experiments and the accompanying laboratory experiments. The experimental setup; the tracer movement; and the migration, sorption, and degradation of MCPP, dichlorprop, bentazone, isoproturon, and BAM was described in a previous paper (14).

Experimental Section Field Site. The experimental site is located in the western part of Denmark near the town of Vejen (Figure 1). The herbicides were injected into a shallow, unconfined aerobic aquifer. The aquifer is underlain by clay about 10 m below ground surface (bgs). The upper aquifer is a glaciofluvial sand and gravel aquifer (Weicselian, Quaternary period). The groundwater table was located about 4.5 m bgs, ∼36 m above sea level (asl), with seasonal fluctuations of about 1 m. The pesticides were injected over a 1-m interval between 33.5 and 34.5 m asl. The dominant flow direction was from north to south. The geology, hydrogeology, and groundwater chemistry of the site were described in more detail in Broholm et al. (14). A field injection experiment to study multicomponent cation-exchange processes had previously been conducted at a site adjacent to this experiment (15, 16). The cationexchange capacity was determined to be 1.0 mequiv/100 g based on 35 aquifer sediment samples. Pedersen et al. (17) determined the clay content of selected samples to be in the order of 0.19-1.7%. Field Injection Experiments. The feed solutions for the field experiments consisted of aqueous solutions of the herbicides DNOC, MCPP, dichlorprop, bentazone, and isoproturon; the dichlobenil metabolite BAM; and the nonreactive tracer bromide (as LiBr). Feed solution was injected continuously for 7 days in the pulse injection experiment and 216 days in the continuous injection experiment, which was injected in the same field plot about 6 moths 4790

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 35, NO. 24, 2001

later. The feed solutions were injected through six injection wells screened over 1 m (33.5-34.5 m asl) and covering a 1.5 m wide section of the aquifer perpendicular to the overall flow direction. The preparation of the feed solutions, well installation, and injection system was described in detail in Broholm et al. (14). In the pulse injection experiment, a 20% increase in DNOC concentration in the injection solution was observed over the 7-day injection period, which was explained by incomplete dissolution of DNOC prior to initiation of the injection. The concentrations at the injection wells were not measured. The DNOC:bromide ratio in the injection solution was 1:2000 (approximately 50 µg/L DNOC and 100 mg/L bromide at the injection wells). In the continuous injection experiment, the injection solution was replaced with fresh solution every 2 weeks. Analysis of the injection solution at the end as well as beginning of each 2-week period, with average end/start concentrations >98%, did not indicate any DNOC degradation in the solution. The resulting maximum concentrations in the aquifer at the injection wells were about 40 µg/L DNOC and 40 mg/L bromide. The cross-sectional mass of DNOC for the monitoring points on the injection wells remained fairly constant over time (data not shown), suggesting that no significant degradation of DNOC occurred in or at the injection wells. Monitoring and Data Interpretation for the Field Experiments. The pulse was monitored in selected points of a three-dimensional network of 36 multilevel samplers (MLSs) with nine sampling points each for 78 days. The monitoring network was expanded prior to and during the continuous injection to provide complete capture of the plume. The plume developed by the continuous injection was monitored in three dimensions in 96 MLSs of 6-9 points each (illustrated in Figure 1) for 230 days; selected individual points were monitored for a longer period. The transects of monitoring points downstream of the injection wells perpendicular to the overall direction of groundwater flow are referred to as fences. The MLSs monitored in the pulse injection experiment were those with width coordinates -1, -0.5, 0, 0.5, and 1 m in all six fences and width coordinates -1.5 and 1.5 m in the 15-, 20-, and 25-m fences (see Figure 1). For the pulse injection experiment, retardation factors for DNOC were determined by half mass breakthrough determined by moment analysis (breakthrough curve integration) for the four monitored points in the 1-m fence, which

fell within the plume. The short pulse did not allow for evaluation of DNOC degradation. For the continuous injection experiment, complete breakthrough curve data for DNOC were analyzed for six specific monitoring points (three in each of the 1- and 5-m fences). These provided additional information on the retardation of DNOC relative to bromide. Integration of the breakthrough curves provided information on the degradation of DNOC. Cross sections (vertical, perpendicular to flow direction) were analyzed for two dates (days 175 and 230) for the 1- and 5-m fences, seven dates (days 56, 92, 121, 175, 216, 230, and 345) for the 10-m fence, three dates (days 175, 230, and 297) for the 15-m fence, and two dates (days 175 and 230) for the 20and 25-m fences. Integration of the time-series of cross sections provided a mass versus time curve less sensitive to spatial variation in DNOC concentration levels than curves for specific point concentration versus time and thus better for evaluation of the degradation. Spatial moment analysis was conducted for the two dates, where all sampled fences had been analyzed for DNOC. The moment analysis provided information on the total mass of the herbicide within the monitoring network, the center of mass, and the spreading of DNOC in three dimensions. The integration of cross sections and moment analysis of snapshots was described in Broholm et al. (14). Laboratory Degradation Experiments. DNOC degradation was investigated in laboratory batch experiments for 10 locations in the aquifer. Sediment and groundwater were collected from six locations within the former location of the bromide and herbicide plume at the site of the field injection (A1-C2) and four locations outside the former plume location (NX1-4), see Figure 1. Aquifer material as well as groundwater were collected 40 days after the termination of the continuous injection. At this time, DNOC was still present at the upper of the monitored points in the 1-m fence and at all monitored points in the 5-m fence. The aquifer material for the batch degradation experiments was collected in undisturbed cores with a stainless steel piston sampler (Ø: 5.6 cm) modified after Starr and Ingleton (18). The cores were stored at 4 °C until subsampled (less than 6 days). To obtain uncontaminated material, 2 cm of each end of the core was cut off, and the outer 0.5 cm was pared off using a paring device modified after Wilson et al. (19). All aquifer materials were collected between 1 and 4 m below the groundwater table. A sample of 200 g of wet aquifer material and 300 mL of groundwater from each location were incubated in infusion bottles (500 mL) and spiked with a pesticide solution including DNOC (25 µg/L of each pesticide) similar to the pesticide solution used in the injection experiment. The headspace consisted of air, resulting in an aqueous oxygen concentration of 1011 mg of O2/L. Controls were amended with 100 mg/L HgCl2 to inhibit microbial activity. All bottles were incubated in the dark at 10 °C. Water samples were collected at regular time intervals. The bottles were shaken for 2 min and then allowed to settle for 6 h before they were sampled. Water samples were collected with sterile syringes and needles. Sterile, filtered atmospheric air was added as compensation. Laboratory Sorption Experiments. Aquifer materials for sorption experiments were obtained by hand-augering to the desired depth and then collecting the material with a bailer. The effect of pH on sorption was investigated on sediment and groundwater collected approximately 20 m east of the injection wells (Figure 1, F10). The experiment was conducted by equilibrating suspensions (8 g of sediment and 15 mL of groundwater) to a given pH (for 5 days), whereafter they were spiked with DNOC to a concentration of 100 µg/L. The pH values ranged from 2.5 to 7.0 and were adjusted by addition of 0.25 M HCl or 0.25 M NaOH (original pH ≈5).

The vertical variation in sorption was investigated on aquifer material collected as two profiles (between levels ∼32 and 35 m asl) from the adjacent field (Figure 1, F9 and F12). The collection of the aquifer materials was described in Pedersen et al. (17). The experiments were performed by equilibrating 20 g of sediment and 19 mL of groundwater in Pyrex vials, and then each vial was spiked with DNOC to an initial concentration of 25 µg/L. For the pH experiment as well as the vertical variation experiment, the suspensions were equilibrated for 96 h in an end-over-end tumbler in the dark at 10 °C. The amount of sorbed DNOC was calculated from the difference between the initial and the equilibrium aqueous concentrations. Additional Kd values were calculated from the data from the laboratory degradation study (see above) for the batches in which no degradation of DNOC was observed. All Kd values reported from the laboratory experiments in this paper are for 96 h. Chemical Analysis. DNOC analyses were performed on filtered (45 µm, Sartorius, Minisart SRP 15, PTFE) samples by high-pressure liquid chromatography (HPLC) on a Hewlett-Packard series 1100 HPLC system equipped with a Phenomenex Luna 5 µm C18(2) column (250 × 4.60 mm) at a wavelength of 370 nm. A mixture of two eluents was used: eluent I consisted of 4 g of tetrabutylammonium hydrogen sulfate and 6.3 g of dipotassium hydrogen phosphate trihydrate in 1 L of Milli-Q water, and eluent II consisted of 450 mL of acetonitrile, 450 mL of methanol, and 100 mL of Milli-Q water. An eluent mixture of 47% I and 53% II was used for analysis of samples from the field injection experiment, and an eluent mixture of 40% I and 60% II was used for samples from the laboratory experiments. Bromide was analyzed on a Dionex ion chromatograph DX120. An Ion Pac 144 mm (10-32) column (P/N 46124) was used in combination with an anion suppresser (ASR II 4 mm, self-regenerating) with 3.5 mM Na2CO2/1 mM NaHCO3 buffer solution as eluent at a flow velocity of about 1.45 mL/min. The analysis also provided data for nitrate, nitrite, sulfate, and chloride. Oxygen and pH values were monitored in the field by electrodes. The fraction of organic carbon on the sediment was determined according to Heron et al. (20). DNOC for the experiments was purchased from Riedel-deHa¨en, D-30926 Seeize, Germany, and had a chemical purity g99%. Specific clay minerals were prepared on glass slides and determined using X-ray diffraction on a Philips PW1050/25 diffractometer with monochromator, Cu radiation. The analysis involved air-drying, saturation with ethylene glycol, and heating the samples to 350 °C and subsequently to 550 °C.

Results and Discussion The bromide plume followed a relatively complex path through the network downgradient of the injection wells in the continuous injection experiment (Figure 1). The plume movement was controlled by spatially varied hydraulic conductivities of the sand deposit and influenced by asynchronous seasonal variation in the elevation of the groundwater table (14). Slower migration of bromide was observed in the uppermost points of the sampling network in the 1and 5-m fences, caused by lower hydraulic conductivity of the deposit in this section (14, 21). The lower hydraulic conductivity was likely related to higher clay content of the aquifer material. The average pore flow velocities at the site were 0.1-0.5 m/day, and the majority of the bromide plume had left the outermost (25-m) fence approximately 100 days after termination of the injection. Aquifer Geochemistry. The aquifer is aerobic (oxygen 2-10 mg/L) in the part of the aquifer affected by the plume injected. As expected, oxygen concentrations were not affected significantly by the degradation of herbicides at low VOL. 35, NO. 24, 2001 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

4791

FIGURE 2. Vertical profiles of pH distribution within the bromide plume at the 1- and 5-m fences prior to the initiation and on day 120 of the continuous injection experiment. initial concentrations in the continuous injection experiment. An acidic front was observed in the upper part of the monitoring network with a sharp increase in pH (4.5-6.0) with depth (Figure 2). The injection of Li+ in connection with the tracer bromide caused a slight increase in the aqueous concentration of the major cations except Na+ due to cation-exchange reactions. The approximate concentrations were (before/after injection): K+ (3-4/7-10 mg/L), Ca2+ (25-30/35-45 mg/L), Mg2+ (3-3.5/4-5 mg/L), and Na+ (12-14 mg/L). Other groundwater quality parameters include NVOC (1-4 mg/L), alkalinity (0-2.5 mequiv/L), chloride (1232 mg/L), sulfate (5-9 mg of S/L), and nitrate (2.5-20 mg of N/L). The content of organic carbon (foc) of the aquifer sediment samples for the laboratory degradation experiments were in the range of 0.007-0.015%. This is in accordance with data from Tuxen et al. (8) reporting a foc of 0.02% at an adjacent part of the aquifer. The clay minerals smectite, illite, and kaolinite were identified in four selected sediment samples from the field site. Sorption of DNOC in Field Experiments. Significant and spatially variable retardation of DNOC in the aquifer was observed in both the pulse and the continuous injection experiments, as illustrated in Figures 3 and 4. The Kd values were calculated using a linear sorption isotherm, as linearity has previously been observed for the low concentration range (3). Retardation factors and corresponding Kd values based on breakthrough curves for specific points for the pulse and

the continuous injection experiments are given in Table 2. The data showed an order of magnitude variation in Kd, ranging from 0.10 to 0.98 L/kg. The Kd values determined in the laboratory experiments for samples from within and immediately outside the plume ranged from 0.16 to 0.42 L/kg. The observed Kd values for DNOC are orders of magnitude higher than expected based on the hydrophobicity (Kow) of DNOC and the foc of the aquifer material. Hence, hydrophobic partitioning of DNOC to organic matter appears to be negligible relative to other sorption mechanisms. Investigations by Haderlein and Schwarzenbach (5) and Weissmahr et al. (6) have demonstrated that nitroaromatic compounds can be bound by a EDA complex formation between siloxane oxygen’s of the clays and the nitroaromatic substituents. This sorption mechanism is of particular importance for DNOC as it contains two nitro groups, which enhance the electronwithdrawing and electron-delocalizing properties and thereby enhance sorption (3). The extent of sorption by this mechanism also depends on the type of clay minerals present in the sediment. Clay minerals identified in the Vejen aquifer (smectite, illite, and kaolinite) have been shown to be strong sorbents by Weissmahr et al. (4). Sorption and pH. Generally, the retardation of DNOC appeared to decrease with depth in the aquifer (compare breakthrough curves for points at different depths in Figures 3 and 4). The vertical variation of Kd observed in the batch experiments (F9) also showed a clear decrease in Kd with depth (Figure 5). An acidic front with low pH in the upper part and higher pH at greater depth were observed in the aquifer (Figure 2). This implied that pH in the aquifer might be related to DNOC sorption. Haderlein and Schwarzenbach (5) demonstrated a correlation between sorption and pH for dissociable nitroaromatic compounds. They showed that only the neutral (nondissociated) forms of nitroaromatic compounds were subject to sorption. Thus, the sorption of DNOC may be correlated to the neutral fraction of DNOC (R0), which again is directly correlated with pH as given by the following equations (5):

Kd ) R0KdHA

(1)

R0 ) 1/(1+ 10(pH-pKa))

(2)

This model assumes that the Kd value of the nondissociated species (KdHA) is constant over the whole pH range considered.

FIGURE 3. Breakthrough curves for DNOC in four monitoring points in the 1-m fence observed in the pulse experiment. Coordinates x, y, and z (in m, m, and m asl) for the monitoring points are given on each graph. 4792

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 35, NO. 24, 2001

FIGURE 4. Breakthrough curves for DNOC in three monitoring points in each of the 1- and 5-m fences observed in the continuous injection experiment. Coordinates x, y, and z (in m, m, and m asl) for the monitoring points are given on each graph.

TABLE 2. Retardation Factors (R), Sorption Coefficients (Kd), and pH for Selected Monitoring Points in the 1- and 5-m Fences for Pulse and Continuous Injection Experimentsa monitoring point x, y, z coordinates (m, m, m asl) 1, 0.0, 34.25 1, 0.0, 33.75 1, 0.5, 34.25 1, 0.5, 33.75 1, 0.0, 34.75 1, 0.0, 34.25 1, 0.0, 33.75 5, -1.5, 34.75 5, -1.5, 34.25 5, -1.5, 33.75

retardation factor, Rb

Kd c (L/kg)

pH (measured in November 1998)

Pulse Injection Experiment 3.2 0.48 1.5 0.10 2.3 0.29 1.7 0.14

4.61 5.68 4.88 5.72

Continuous Injection Experiment 4.7 0.78 3.2 0.48 1.6 0.12 6.3 0.98 2.8 0.33 2.2 0.22

4.50 4.61 5.68 4.78 4.81 5.06

a

Coordinates for monitoring points refer to Figure 1. b Retardation based on the following: pulse, half mass breakthrough curve; continuous, half concentration increase of breakthrough. c Calculated using a porosity of 0.38 and a bulk density of 1.78 g/cm3.

A simple laboratory batch experiment with pH adjustment was performed to verify that this mechanism was of importance for the Vejen aquifer. The obtained Kd data are plotted against the neutral fraction (R0) of DNOC for the corresponding pH of each batch in Figure 6a. The slope of the linear regression line is then KdHA. Figure 6b illustrates the Kd data as a function of pH. The solid line corresponds to the regression line from Figure 6a. Close correlation of the trend of the data with the regression line confirms the correlation of sorption with the neutral fraction of DNOC (eq 1) and thereby the importance of spatial pH variations for DNOC sorption in the aquifer. As predicted by the model, half of the KdHA value is reached at the pKa value. Similarly, the Kd values observed for all our laboratory measurements and field observations values are plotted versus the neutral

FIGURE 5. Kd and pH as a function of depth for a vertical profile (F9, see Figure 1) approximately 20 m east of the field injection site. fraction of DNOC (R0) in Figure 7. The neutral fraction of DNOC was calculated based on the groundwater pH measured for the specific point prior to the injection experiments. The data tend to group in two: one group with the laboratory data from locations F9 and F12 and one group with the field data and the laboratory data from locations close to or inside the injection site. Although, some scatter is observed, there appears to be a reasonable correlation with the neutral fraction of the DNOC and hence with pH for each group. For the field data, one data point (5 m G2, Table 2) apparently showed significantly stronger sorption than expected. A likely explanation for this observation is that this Kd value was determined from a late breakthrough for which influence of degradation could not be ruled out, therefore making determinations of Kd values uncertain (see later discussion of degradation). The most probable reason for the relative higher Kd values for the batches of locations F9 and F12 as compared to the other data is the difference in experimental setup. Kd values for F9 and F12 were determined in continuously shaken batches where new sorption sites could be made available during the experiVOL. 35, NO. 24, 2001 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

4793

FIGURE 6. Kd for DNOC in aquifer sample suspension with adjusted pH. (a) Kd as a function of the neutral fraction of DNOC (r0). The solid line is a linear regression to the data, where the slope 14.0 L/kg corresponds to KdHA in eq 1. (b) Kd as a function of pH in the aquifer samples. The solid line is modeling of the data using eq 1 with KdHA equal to 14.0 L/kg.

FIGURE 7. Comparison of Kd vs neutral fraction of DNOC (r0) calculated from pH values determined in the field injection experiments and from laboratory batch experiments. mental period and thereby result in an enhanced sorption. This corresponds well with the findings of Haderlein et al. (3), who observed a correlation between the Kd values and the available area of the siloxane surfaces of the clay minerals. Nevertheless, the overall relationship between pH and Kd suggests that pH is a significant factor, which should always be taken into account when sorption of DNOC is evaluateds even in sandy aquifers with low clay content. Other Factors Affecting Sorption. Weissmahr et al. (22) demonstrated that the equivalent fraction of specific cations on the clay mineral surfaces had significant effects on sorption. The coverage of the clay mineral surfaces by H2O in connection with strongly hydrated cations (such as Al3+, Mg2+, Ca2+, Na+, and Li+) prevented the formation of the 4794

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 35, NO. 24, 2001

EDA complexes, whereas minor coverage by H2O in connection with weakly hydrated cations (such as NH4+, K+, Rb+, and Cs+) did not. All major cations were measured in the aqueous samples of the pH laboratory experiment. These aqueous concentrations were converted to the corresponding equivalent fractions on the clay mineral surfaces (Ncation) by use of the Gaines-Thomas convention and selectivity coefficients previously determined for the Vejen aquifer by Bjerg et al. (16). Weissmahr et al. (22) observed a significant increase of the Kd values of nitroaromatic compounds related to DNOC when NK (equivalent fraction of K+ on the clay mineral surfaces) exceeded 40%. In the pH experiments, NK ranged between 3% and 12%, and in previous investigations of the aquifer (16), the average NK value was 6% and NK never exceeded 13%. The slight increase in the aqueous K+ concentration caused by ion exchange with Li+ corresponds to a decrease in NK of only 0.5%. This indicates that the variation in cation composition is unlikely to have had a significant effect on DNOC sorption at this site. Hence, it cannot explain the large variation in Kd values for DNOC. The method used to identify the specific clay minerals in this study was not quantitative; however, the lower hydraulic conductivity and visual inspection of samples suggests a higher clay content in the upper part of the aquifer in the 1- and 5-m fences. Since, sorption of DNOC is expected to be dominated by sorption to clay minerals, we suggest that the significantly stronger sorption is related to a higher content of clay minerals. This is supported by a fair correlation between Kd and the clay content determined on a subset of the F12 samples (data not shown). The foc values of all samples showed a negative correlation with Kd. Weismahr et al. (22) reported the lowering of Kd values for nitroaromatic compounds due to sorbed organic matter on the siloxane plane of the clay. However, the correlation in our case is weak and can also be ascribed to other factors. Thus, the significant (Kd range of 0.05-1.05 L/kg) and variable sorption of DNOC seems primarily to be governed by the pH, but the microscale variability of the mineralogy/geology of the Vejen aquifer plays a role as well. Degradation of DNOC in Continuous Field Injection Experiment. For the continuous field injection experiment, the degradation of DNOC was evaluated based on breakthrough curves for specific sampling points in the 1- and 5-m fences (Figure 4), on plan-view mass distribution in the entire network for two snapshots (Figure 8), and on DNOC distributions and masses in cross sections at various dates (10-m fence data for seven dates are shown in Figures 9 and 10). The breakthrough curves for the mid (1, 0.0, 34.25) and lower part (1, 0.0, 33.75) of the 1-m fence indicated almost full mass recovery, whereas the breakthrough curves for the upper part of the 1-m fence (1, 0.0, 34.75) and the 5-m fence suggested removal of DNOC (Figure 4). In particular, determination of the mass below the upper point 5-m breakthrough curves (1, -1.5, 34.75) supported this with a DNOC mass of about 30% (ratio between bromide and DNOC) of the corresponding bromide mass. If DNOC was affected only by sorption, we would observe a delay in breakthrough, tailing in the breakthrough curves, but full recovery of the DNOC mass. Comparison of masses from breakthrough curves demands that the flow paths for DNOC and bromide are identical. The plan-view mass distribution in the entire network for the two snapshots (Figure 8) reveals that DNOC in the first about 5 m of the network downstream of the injection wells followed near the same flow path as bromide. Thus, the breakthrough curves strongly indicate that degradation took place between the injection and the 5-m fence. Further away from the injection wells, the flow path for DNOC and bromide seems to be fairly similar, but the

FIGURE 8. Plan view of mass distribution (depth integrated) for bromide and DNOC within the monitoring network on days 175 and 230. concentrations of DNOC relative to bromide were significantly lower (Figure 8). The cross-sectional distribution of DNOC in the 10-m fence over time (Figure 9) revealed relatively early arrival of DNOC at this distance but with a continuously smaller affected area and lower overall concentrations of DNOC than bromide. Late arrival and low concentrations were especially observed in the left-hand side of the cross section. The significant and spatially variable sorption of DNOC did not facilitate discrimination between sorption and degradation. For cross-sectional mass breakthrough, sorption may cause delayed breakthrough. However, if no degradation was taking place the total injected mass should be preserved and pass each of the fences. Simulations with a three-dimensional reactive solute transport model reported in Højberg (21) showed that peak cross-sectional mass (mg/m) in the 10-m cross section (using Kd values determined from the 1- and 5-m fence breakthrough curves) was approximately 75% (ratio between DNOC and bromide) of the corresponding bromide peak cross-sectional mass (g/ m). Determination of cross-sectional mass of DNOC for various dates revealed continuously low DNOC relative to bromide mass (∼20%) (Figure 10). In addition, DNOC was not detected on day 345 in the limited section sampled in the 10-m fence (129 days after the termination of the injection). If the entire mass of DNOC to pass the 10-m fence was to reach the mass injected (or corresponding mass of bromide to pass the fence), high concentrations of DNOC should still have been present at this time. The presence of DNOC further out in the aquifer was evaluated from the plan-view distribution (Figure 8) and by sampling of a complete cross section (data not shown) at the 15-m fence at day 297. The cross-sectional mass for these three dates are given in Table 3. The mass of DNOC in the 15-25-m fences increased between day 175 and day 230, but no DNOC was found in any sampling points in the 15-m fence at day 297 showing an efficient removal of the remaining DNOC at this date. In conclusion, the detailed monitoring of the DNOC plume has shown that degradation occurred throughout the aerobic aquifer and that beyond 10 m downgradient of the injection DNOC had been completely degraded at day 297.

Laboratory degradation experiments revealed fast degradation of DNOC in experiments with aquifer material from two sampling locations between the injection wells and the 1-m fence (Figure 11). Within 15 days, more than 90% of the initial DNOC had disappeared, and after 56 days, concentrations were below detection (