| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
UMR INAPG/INRA Environment and Arable Crops, B.P. 01, 78850 Thiverval-Grignon, France
* Corresponding author (coquet{at}grignon.inra.fr)
Received 5 March 2002.
ABSTRACT
Knowledge of the sorption properties of the vadose zone is essential to provide sound risk assessments of groundwater contamination by pesticides. This study explored the sorption properties of the vadose zone of a 187-ha agricultural catchment located 120 km northeast of Paris, France. Eocene geologic materials composing the vadose zone include limestones, marlstones, clays, and sands. Distribution coefficients (Kd) of 14C-labeled atrazine (6-chloro-N2-ethyl-N4-isopropyl-1,3,5-triazine-2,4-diamine), isoproturon [3-(4-isopropylphenyl)-1,1-dimethylurea], and metamitron [4-amino-4,5-dihydro-3-methyl-6-phenyl-1,2,4-triazin-5-one] were measured in batch on 30 vadose zone samples, and their sorption isotherms measured on six representative samples. Kd values for each pesticide were highly variable and ranged from 0.10 to 2.99 L kg-1 for atrazine, 0.06 to 1.27 L kg-1 for isoproturon, and 0.06 to 2.08 L kg-1 for metamitron. Two vadose zone materials, a lignitic clay and a clay with low organic C content, were found to have Kd values higher than those of the topsoil. Most of the sorption isotherms in the vadose zone were found to be linear. Kds could be well predicted by the clay content (particle-size fraction < 2 µm after organic matter and carbonates removal) of the vadose zone. This result suggests that simple but efficient pedotransfer functions may be derived to predict the sorption properties of the vadose zone at the catchment scale.
Abbreviations: AFNOR, Association Française de Normalisation • dpm, disintegrations per minute • PCA, principal components analysis
LIMITED KNOWLEDGE is available about the key physicochemical processes driving the fate and transport of pesticides in the vadose zone. In studies of the fate of pesticides in the environment, the primary focus has been on soil (Bailey and White, 1970; Goring and Hamaker, 1972; Sawhney and Brown, 1989; Cheng, 1990; Linn et al., 1993), which is the main receptacle for pesticide residues and a place where most pesticide degradation processes occur. Interest in aquifers is more recent and has increased with the availability of national and regional inventories that revealed the common occurrence of pesticide residues in groundwater (Garner et al., 1986; Hallberg, 1989; Leistra and Boesten, 1989; Barbash and Resek, 1996; Kolpin et al., 2000). Most experimental studies about pesticides in aquifers have dealt with transformation and degradation processes (Kraft and Helmke, 1991; Agertved et al., 1992; Rodriguez and Harkin, 1997; Johnson et al., 2000; Larsen et al., 2000; Albrechtsen et al., 2001; Kristensen et al., 2001; Papiernik, 2001). Such studies are essential to address the question whether or not detected contaminants are likely to persist in the vadose zone. Retention processes deserve attention as well, since they generally play a major role in the fate and transport of pesticides in the subsurface. Retention in the vadose zone delays the appearance of pollutants in groundwater, and increases their migration time, thereby allowing more time for the chemical to undergo degradation in the vadose zone (Bayless, 2001; Goody et al., 2001). Degradation rates in subsoils and in the vadose zone are generally low (Johnson and Fuhrmann, 1993; Kördel et al., 1995; Accinelli et al., 2001; Mills et al., 2001; Vinther et al., 2001; Wood et al., 2001), whereas a long residence time may result in more significant attenuation. Information about the retention processes in the vadose zone is essential to predict this attenuation.
Most of the data on sorption properties of the vadose zone concern subsoils (Sonon and Schwab, 1995; Felding, 1997; Walker et al., 1999) or sandy and/or Quaternary aquifer materials (Wehtje et al., 1983; Nordmeyer et al., 1992; Roy and Krapac, 1994; Moreau and Mouvet, 1997; Moorman et al., 2001). Sorption data on other geological rocks are rare. These include data on the European Senonian chalk (Johnson et al., 1998; Besien et al., 2000) or other rocks of more local distribution (Palmer et al., 1992; Bersillon et al., 1994). Sorption data are generally measured in batch and expressed by means of the distribution coefficient Kd between the solid and the liquid phases at a single concentration:
 | [1] |
The study presented here provides more information about pesticide sorption in the vadose zone, with specific focus on sorption variability with depth, isotherm linearity, and relationships between sorption and the nature of vadose zone materials. Tertiary geologic materials considered in this study include limestones, marlstones, sands, and clays.
MATERIALS AND METHODS
Vadose Zone Description and Sampling
Vadose zone samples were taken from three sites (Fig. 1) in the Bruyères-et-Montbérault catchment, located in the north of France 120 km northeast of Paris. This 187-ha catchment is formed by Eocene sedimentary beds of the Parisian Basin (Blondeau et al., 1980; Cavelier et al., 1980). These Eocene beds include, from bottom to top (Fig. 1): Cuisian sands (50-m thickness at the center of the catchment) and Laon clay (2 m) dated from the Late Ypresian, Lutetian coarse limestone (25 m) and heterogeneous marls and gravels (17 m), and Auversian discontinuous Saint-Gobain clay (<1 m) supporting the Beauchamp sands (10 m). Two successive loess deposits covered the outcrops of these materials during the Quaternary Period and were largely eroded and decarbonated except in a few places in the catchment where their thickness still reaches 7 m. The main aquifer of the catchment is the Lutetian limestone aquifer. Its aquitard is formed by the Laon clay layer. About 20 springs seeping from the plateau border constitute the outlet of the aquifer. The piezometric height near the center of the catchment varies between 10 and 15 m, so the vadose zone occupies the major part (
40 m thick) of the catchment. Three sampling sites (Fig. 1) were selected to obtain the main vadose zone materials while limiting the number of drillings. Samples were taken with a rotative dual-shell stainless-steel corer for poorly consolidated materials and with a helicoidal hollow-stem auger for hard rocks (limestone). The sampling progression was limited to segments of about 50 cm, with the sampling equipment cleaned between each segment. From the 20 m of drillings collected, we selected 30 samples that represented most of the facies variations encountered in each bed of the catchment. Samples were sealed in plastic bags or tin cans and stored at 4°C before analysis.
|
|
Distribution coefficients (Kd) of atrazine, isoproturon, and metamitron on each of the 30 vadose zone samples were measured in batch using 14C-labeled molecules. Duplicate 5-g samples were added to 10 mL of 14C pesticide solution in CaCl2 10-2 M in glass centrifuge tubes with Teflon caps. Adsorption of any of the three molecules to the glass tubes and/or the caps was found negligible. Initial pesticide solutions were prepared by mixing 10 to 50 µL of 14C methanolic solution to 500 mL of unmarked aqueous solution, so that any cosolvency effects (such as illustrated by Bouchard, 2002) remained negligible. Initial concentrations were 3.59 mg L-1 for atrazine, 3.81 mg L-1 for isoproturon, and 5.22 mg L-1 for metamitron, with activities of 4760, 4720, and 4760 disintegrations per minute (dpm) mL-1 respectively. Samples were left 24 h in an end-over-end agitator at 22 ± 1°C. Any transformation of the pesticides during that time was considered to be negligible (Coquet, 2003). Samples were subsequently centrifuged for 15 min at 2000 g. An aliquot of 0.5 mL of the supernatant was mixed with 4 mL of scintillating liquid (Ultima Gold XR Packard; PerkinElmer Life Sciences, Boston, MA) and counted 10 min in a Betamatic V Kontron scintillation counter (Kontron Instruments, Montigny-le-Bretonneux, France). The quantity of sorbed solute was next calculated from the difference in counts between the initial pesticide solution concentration and the solution concentration after equilibration with the soil sample (Organization for Economic Cooperation and Development, 2001). Decreases in activities ranged from 240 to 2810 dpm mL-1 for atrazine, 165 to 1830 dpm mL-1 for isoproturon, and 145 to 2480 dpm mL-1 for metamitron.
The same procedure was used for the sorption isotherms, but with five different initial concentrations ranging from 1.10 to 6.87 mg L-1 for atrazine, 1.13 to 22.7 mg L-1 for isoproturon, and 1.16 to 23.2 mg L-1 for metamitron, and three repetitions for each initial solution concentration. These initial concentrations represent one-one hundredth to one-third of the maximum pesticide concentration in soil after application assuming standard agricultural doses (1, 1.8, and 2.8 kg ha-1 for atrazine, isoproturon, and metamitron, respectively) that reach the soil surface, and a depth of mixing in the soil of 1 cm. These pesticides may be expected to reach groundwater at concentrations much smaller than the values above. For instance, Kolpin et al. (2000) reported a maximum concentration of atrazine of 4.2 µg L-1 in U.S. groundwaters. Concentrations of atrazine >2 µg L-1 were recorded only six times in more than 2300 French groundwater samples (Institut Française de l'Environnement, 2001). However, accurate sorption measurements at low pesticide concentrations are difficult to achieve (e.g., Boesten, 1990; Bersillon et al., 1994; Moreau and Mouvet, 1997). A trade-off had to be found between the accuracy of the sorption measurements and their relevance to vadose zone transport processes. This last point will be addressed below in the discussion of the results. Initial activities of the solutions ranged from 1550 to 1980 dpm mL-1 for atrazine, 4520 to 5460 dpm mL-1 for isoproturon, and 2190 to 2340 dpm mL-1 for metamitron. Decreases in activities ranged from 35 to 510 dpm mL-1 for atrazine, 100 to 850 dpm mL-1 for isoproturon, and 180 to 980 dpm mL-1 for metamitron. Sorption isotherms were fitted to the Freundlich equation in its logarithmic form:
 | [2] |
Accuracy of the Sorption Measurements
Accuracies of the sorption data were calculated as indicated by Boesten (1990), except for errors in the initial and final activities of the solution (
a and e in Boesten, 1990), which were evaluated in a more rigorous way. Recognizing the Poisson distributed nature of radioactive emission, relative accuracy of the solution activity is given by:
 | [3] |
The relative accuracies of the Kd values were between 2 and 18% for atrazine with an average of 8%, between 3 and 38% for isoproturon with an average of 9%, and between 3 and 52% for metamitron with an average of 14%. Absolute accuracies of the Kds were between 0.016 (Sample A13) and 0.061 (Sample B8) for atrazine with an average of 0.021 L kg-1, between 0.019 (Sample A6) and 0.036 (Sample B8) for isoproturon with an average of 0.021 L kg-1, and between 0.016 (Sample B11) and 0.051 (Sample B8) for metamitron with an average of 0.020 L kg-1. These accuracies were consistent with the repeatability of the measurements. Standard deviations that may be computed from the Kd duplicates were on average 0.020 L kg-1 for atrazine, 0.010 for isoproturon, and 0.028 for metamitron, which is in reasonably good agreement with the accuracy calculations given the limited number of replicates.
Relative accuracies of the sorbed concentrations S in the isotherm experiments were between 4 and 76% for atrazine with an average of 12%, between 5 and 46% for isoproturon with an average of 14%, and between 2 and 15% for metamitron with an average of 7%. Absolute accuracies of the sorbed concentrations ranged from 0.03 to 0.18 mg kg-1 from the lowest to the highest concentrations for atrazine, from 0.02 to 0.42 mg kg-1 for isoproturon, and from 0.03 to 0.55 mg kg-1 for metamitron. Standard deviations of S computed from the triplicate values were always smaller than these absolute accuracies.
Sample Selection for the Isotherm Measurements
Not all of the 30 vadose zone samples initially collected were characterized for their sorption isotherms. It was thought that the 30 samples should not be all different, particularly those taken from the same geological layer, but with slightly different facies. Therefore, the sorption isotherms may be measured on a limited number of representative samples and still give a fairly good picture of the variation of the isotherm parameters in the vadose zone of the entire catchment. To do this, a principal components analysis (PCA) of all Kd values and other physicochemical data was first performed to investigate similarities among the samples. Principal components analysis is a standard mathematical tool for analysis of multivariate data, especially large tables of variables and observations (see e.g., Blackith and Reyment, 1971; Mardia et al., 1979; Insightful Corporation, 2001, chap. 20). Second, a clustering method (Ward, 1963) was used to group the samples according to their similarity. In each group, the sample nearest to the barycentre of the group was then selected for isotherm characterization.
RESULTS AND DISCUSSION
Kd Variations
The Kd values of atrazine, isoproturon, and metamitron measured on the 30 vadose zone samples are listed in Table 2 and reported in Fig. 2, together with the basic stratigraphy of the three sampling profiles. Kd values for the topsoil (0–18 cm) at each site were added for comparison (Table 3). Two vadose zone materials were found to have Kd values higher than that of the topsoil. The 3.6- to 3.8-m-depth sample in Profile B, noted B8 in Fig. 2b, was identified as a lignitic clay, and had an atrazine Kd 3.6 times that of the topsoil. In Profile C, Sample C5 (depth 2.65–2.85 m) corresponding to a clay bed had a metamitron Kd 1.5 times that of the topsoil. A trend of decreasing Kds from the soil surface to deeper depths as observed by various authors (Felding, 1997; Jenks et al., 1998; Jacques et al., 1999) was not uniformly present along the three vadose zone profiles shown in Fig. 2, although it was valid for the top 1.5 m. Kd variations vs. depth in the vadose zone were mostly determined by the type of geological material present.
|
|
|
|
|
The relative sorption capacities of the vadose zone materials for the three herbicides were different from those of the topsoil (Fig. 2). Whereas atrazine tended to be less sorbed than isoproturon and metamitron in the topsoils, it was predominantly sorbed by all vadose zone materials of Profiles A (except outlier A5) and B (except near-surface sample B1). In Profile C, isoproturon has a Kd well below those of atrazine and metamitron. It is worth noting that Madsen et al. (2000) observed the same pattern for a similar type of carbonate-free sandy vadose zone samples from Denmark.
Similar orders of magnitude were found for atrazine Kd or Kf values by Moorman et al. (2001) for alluvium, loess, and oxidized till subsurface materials; by Roy and Krapac (1994) for Quaternay sands and tills; and by Moreau and Mouvet (1997) and Jenks et al. (1998) for subsoils. Agertved et al. (1992) and Widmer and Spalding (1995) measured atrazine retardation in sandy aquifers that corresponded to Kds between 0.03 and 0.05 L kg-1. Such low Kds are probably the result of the low organic C content (
0.2 g kg-1) and low clay content (1 g kg-1) of the aquifers. The isoproturon Kds measured by Johnson et al. (1998) in the Cretaceous chalk of England between the 2.1- and 9.9-m depths ranged between 0.02 and 0.46 L kg-1. Madsen et al. (2000) found isoproturon Kds between 0.07 and 0.85 L kg-1 for Quaternary aquifer sands and the Cretaceous chalk of Denmark. Metamitron Kds of the same materials ranged between <0.1 and 0.97 L kg-1, plus two unexplained outliers at 1.1 and 3.1 L kg-1.
Principal Components Analysis
Principal components analysis is a simple and efficient way to investigate the relationships between variables (i.e., correlations), between observations (i.e., similarities), or between observations and variables (i.e., how similarities among observations may be explained by the variables). Principal components analysis is especially useful when analyzing a large number of observations described by a large number of variables, such as in our case: 30 samples described by 11 variables (Table 2). As stated by Blackith and Reyment (1971)(Chapter 12), PCA is essentially the statistical application of eigenvectors decomposition and stems from the fact that it is generally easier to investigate data by considering a small number of well-chosen linear combinations of the original variables, rather than all of the variables separately (which may be too cumbersome when the number of variables is large). Through the principal components' decomposition, it is possible to explore much of the variability of the data by looking only at a few graphs using the first principal components (the first eigenvectors) as their axes in place of the original variables. In our case, the first principal component (noted PRIN1, Fig. 3) accounted for 46% of the total variance, the second principal component (PRIN2) 27%, and the third (PRIN3) 15%. In all, the first three principal components were enough to account for 88% of the original variability corresponding to all 11 variables. The discriminatory power of PCA is illustrated in Fig. 3, which shows the coordinates of the 30 vadose zone samples in the first two new axes (PRIN1, PRIN2). These first two principal components include 73% of the original variability of the data, thus giving a better representation of the original variability than any one graph that could be made by selecting two of the original variables. In Fig. 3, one can easily identify similarities among the samples that tend to group together, whereas for instance sample B8 (the lignitic clay) clearly appears specific.
|
|
From the seven clusters, seven representative samples were selected by taking from each cluster that sample (when the cluster included more than one) located closest to the cluster barycentre in 11-dimensional data space. The seven selected samples are italicized in Table 2. The Kd values, the clay content, and the organic C content of the B8 sample were highly divergent from those of the other samples, which explains the remote position of this sample in Fig. 3. Sample C5 had a high clay content but a low organic C content. Sample A1 corresponded to the decarbonated loam layer at the upper part of Profile A (Fig. 2). Sample A10 corresponded to the limestone. The marlstones and the top of Profile B were represented by Sample B2, and the sandier and less organic layers of Profile B by Sample B6. The materials derived from the Beauchamp sand were represented by Sample C4.
Isotherm Descriptions
Unfortunately, the quantity of material recovered during the drilling was not sufficient to characterize the sorption isotherms of Sample B8. The sorption isotherms of atrazine, isoproturon, and metamitron on the six other vadose zone samples are presented in Fig. 5. Fitted linear isotherms are also shown. The Freundlich model was fitted to the isotherms, and its parameters are listed in Table 4.
|
|
level, except for the atrazine isotherm of Sample B2 ( = 0.027) and for the metamitron isotherm of Sample B6 ( = 0.048). Hence, most of the vadose zone sorption isotherms of Fig. 5 could be fitted to a linear isotherm. These values are also presented in Table 4, together with their 95% confidence intervals.
Some discrepancies between the Kd values derived from the isotherms and those measured initially can be seen by comparing Tables 2 and 4. Student tests were used to check whether or not these discrepancies were significant by comparing the normalized residues of the initial Kd duplicates to the residual variance of the linear isotherms. Excluding Samples B2 for atrazine and B6 for metamitron, whose isotherms were not linear, the initial Kd values were found to be consistent with those derived from the isotherms in 70% of the cases (23/32), considering a 0.05 probability. In the case of atrazine, both duplicate initial Kd values of Samples A1 and B6 and one of the duplicates of Sample B4 overestimated the isotherm-derived Kd. Only one of the duplicate initial Kd values of Sample A1 overestimated the isotherm Kd of isoproturon. Both duplicate points of Sample A10 and one of Sample C5 significantly underestimated the metamitron isotherm Kd value. The consistency check increased to 87% when considering a 0.01 probability for the Student tests, with only four initial Kd measurements different from the isotherm Kd out of 32. This consistency might be expected to be even higher if one accounted for the uncertainties on the measured values (see Materials and Methods). The differences observed for the atrazine Kd of Sample B6 and the metamitron Kd of Sample A10 at the 0.01 probability level may also be partly the result of heterogeneities of the samples, despite the standard mixing–splitting preparation of the samples. However, these discrepancies do not substantially modify the general trends observed within each profile (Fig. 2).
Most of the sorption isotherms in the topsoil of the same catchment were found to be nonlinear, with Nf < 1 (Coquet, 2003). Such nonlinearity in the topsoil may be due to the presence of highly aromatic organic matter in which specific sorption occurs at internal sites (Pignatello, 1991). The linearity of vadose zone sorption isotherms could be caused by the absence of highly condensed microcrystalline organic matter fractions (Weber et al., 1992; McGinley et al., 1993; Kleineidam et al., 1999; Karapanagioti and Sabatini, 2000; Karapanagioti et al., 2000) such as those created through diagenesis in shales (Young and Weber, 1995; Johnson et al., 2001). Herwig et al. (2001) found Nf values close to unity for atrazine isotherms on clay minerals. However, Moreau-Kervévan and Mouvet (1998) and Clausen et al. (2001) found an atrazine Nf value for kaolinite slightly less than 1, while Laird et al. (1992) suggested that Nf generally tended to be <1 for smectites. On the same kaolinite, Clausen et al. (2001) found an isoproturon Nf of 1. These same authors also could not detect any adsorption of atrazine or isoproturon on calcite and quartz (except at pH 2.4). Our results (Table 4) suggest that clay minerals, especially if the dominating type is kaolinite, may dominate the sorption processes in the vadose zone of the Bruyères-et-Montbérault catchment.
To test the efficiency of the clustering method, we compared pair by pair the sorption isotherms of all selected samples for each chemical (Table 4). Since two of the isotherms were nonlinear, the comparison was done on the Freundlich isotherms in their logarithmic form by comparing their slopes (Nf) and intercepts (Kf). All selected samples had different sorption isotherms for at least one herbicide (Table 4). Atrazine Kf values were all different, except for Samples B2 and C4, and all atrazine Nf values were not significantly different ( > 0.15). Isoproturon Kf values were all different at the 0.05 -level, except Sample B6 when compared with Samples B2 or A1. Isoproturon Nf were not significantly different at the 0.05 -level, except Samples B2 and A10. All metamitron Kf were different from each other at the 0.02 -level. Metamitron Nf were not significantly different at the 0.05 -level, except Sample B6 when compared with all other samples but Sample B2. This confirms that all the samples selected by the clustering process had different sorption properties in regard to the three herbicides considered in this study.
Implications For Vadose Zone Transport
In our sorption experiments, pesticide concentrations at equilibrium ranged from 0.69 to 23 mg L-1. Such concentrations cover the upper part of the range of concentrations that can be expected in the subsoil or in the upper part of the vadose zone (e.g., the first 2 m) after standard pesticide applications. Goody et al. (2001) found that isoproturon concentrations in a chalk aquifer did not exceed 0.5 µg L-1, while also measuring a maximum subsoil solution concentration of 0.5 mg L-1 287 d after the application. Bayless (2001) similarly found atrazine concentrations below 0.001 µg L-1 in a 4.5- to 5.8-m-deep till plain aquifer but measured a maximum concentration of 4 mg L-1 at the 1.5-m depth within the same period.
The relevance of the pesticide concentrations used in this study could be evaluated by applying attenuation factor (AF) calculations (Rao et al., 1985) based on the properties of the three vadose zone profiles (Fig. 2). Most of the parameters needed for the AF calculations of the three herbicides at 1 m depth were measured locally: Kds (Table 2), average field capacities and bulk densities of the loess-derived (Profiles A and B: 0.32 cm-3 cm-3 and 1.57 kg dm-3) and sand-derived (Profile C: 0.22 cm-3 cm-3 and 1.48 kg dm-3) materials, and average (1991–1997) annual recharge rate (December–April: 275 mm). The field half-lives were estimated from the French Ministry of Agriculture database (INRA, 2001): 77 d for atrazine, 33 d for isoproturon, and 38 d for metamitron. Degradation was assumed to occur down to 1 m depth, whereas half-lives were considered to be inversely proportional to the amount of organic matter in each horizon. Attenuation factors were found to be 0.067 for atrazine, 0.013 for isoproturon, and 0.018 for metamitron in Profile A. Considering standard application doses (see Materials and Methods), the concentrations at 1 m depth were 2.2, 1.2, and 1.1 mg L-1 for atrazine, isoproturon, and metamitron. In the same manner, AFs at 1 m depth in Profile B were 0.053, 0.0013, and 0.0023, and the corresponding concentrations were 1.7, 0.11, and 0.14 mg L-1 for atrazine, isoproturon, and metamitron, respectively. At 1 m in Profile C, the AFs were 0.29, 0.086, and 0.070, and the expected concentrations were 9.7, 7.7, and 4.2 mg L-1. Such concentrations are consistent with observations by Bayless (2001) and Goody et al. (2001) and belong to the lower part of the sorption isotherms presented in Fig. 5. These concentrations may be expected at 1 m depth in the vadose zone profiles of the Bruyères-et-Montbérault catchment and to some extent (depending on the degradation and dispersion properties of the vadose zone) below that depth. However, the sorption data presented here may be less relevant for the lower concentrations (<100 µg L-1) that can be expected deeper in the vadose zone.
The extrapolability of the isotherm linearity found in this study may be questionable at low concentrations. Based on thermodynamic considerations (e.g., Karickhoff, 1984), linear isotherms may be expected for dilute sorbates and homogeneous sorbents, whether or not they involve partitioning or adsorption processes (Mingelgrin and Gerstl, 1983). These findings were initially supported by available experimental data (Chiou et al., 1979; Ball and Roberts, 1991). However, invoking a distributed reactivity concept (Weber et al., 1992; McGinley et al., 1993) to describe sorption by soils and subsoils, and evidence that isotherms may be nonlinear at low concentrations in the case of strong specific sorbate–sorbent interactions (Spurlock and Biggar, 1994) suggests that one should be particularly careful when extrapolating sorption isotherms. For instance, Chiou and Kile (1998) showed that the sorption isotherms of polar solutes, such as the three herbicides considered here, on a "mineral" soil (organic C content of 2.6 g kg-1) were linear down to a relative solute concentration (C/Sw, where Sw is the solubility of the solute in water) of 0.1 to 0.3, and nonlinear below these values. Relative concentrations for the isotherms of Fig. 5 ranged from 0.0003 to 0.35, but no indication of nonlinearity was detected at the lowest concentrations in the case of the linear isotherms. Most of the vadose zone samples of the Bruyères-et-Montbérault catchment had a low organic C content (<5 g kg-1, Table 2), with the clay fraction generally being the dominant sorbent. This could explain that most of the isotherms were found to be linear. However, it remains to be shown experimentally if this linearity can be extrapolated to low herbicide concentrations (<100 µg L-1).
Dependence of Sorption on Vadose Zone Properties
Possible relationships between the sorption parameters of atrazine, isoproturon, and metamitron and physicochemical properties of the vadose zone materials listed in Table 2 were explored using correlation analysis. For all three herbicides, Kd was found to be strongly correlated to clay content. Spearman rank correlation coefficients were 0.87 for atrazine ( < 0.0001), 0.93 for metamitron ( < 0.0001), and 0.73 for isoproturon ( = 0.0001). No significant correlation between the Kds and the organic C content was observed ( > 0.37). This confirms the dominant role of the clay minerals in the sorption of organic pollutants on sediments having low organic C contents, as shown by Karickhoff (1984) and others. In particular, the high Kd values measured for Samples B8 and C5 are easily explained by the high clay content of these samples (Table 2). A negative correlation was found between the Kds and the pH ( .0002) as well as the CaCO3 content ( < 0.04). However, it is likely that these correlations are indirectly due to the strong negative correlation between clay content and pH or CaCO3 content ( < 0.002). This correlation is probably a consequence of the deposition process of the sedimentary materials under the prevailing carbonated environmental conditions of the Lutetian sea, with a balance between clays and carbonates. This balance ranged from clays with low carbonates in the Marl and Gravels beds (e.g., Sample B8) to limestones with low clay content (e.g., limestone of Profile A) (Blondeau et al., 1980). No correlation was found to be significant at the 0.05 level between the Nf parameters (Table 4) and any of the physicochemical properties of the samples.
Regression analysis showed that the clay content explained 91.1, 83.1, and 97.8% of the Kd variations for atrazine, isoproturon, and metamitron, respectively. Removing the peculiar limestone sample, A5, improved the correlation with clay content, especially for isoproturon Kd. Without the A5 outlier, the clay content explained 91.7, 90.7, and 98.0% of the Kd variations for atrazine, isoproturon, and metamitron, respectively. The polarity of the three herbicides may relate to the quality of the relationship of Kd with clay content. Metamitron, whose Kd variability was explained in large part by the clay content, also has the largest solubility in water (Table 1). Stepwise forward regression did not reveal any other explanatory variables for atrazine (5% threshold level for the Fisher test). The coarse silt content could be considered as a secondary explanatory variable for metamitron Kds, but only explained <0.5% of the variability. On the other hand, the organic C content could explain a further 2.1% of the isoproturon Kd. Satisfactory regressions (R2 > 90%) could be obtained for topsoil Kds in the same catchment but with the organic C content as the main explanatory variable, instead of clay content (Coquet and Barriuso, 2002). These results suggest that efficient and simple pedotransfer functions may be developed for the prediction of sorption coefficients of the vadose zone at the catchment scale, but that they may require different basic information (e.g., clay content) than those usually required for soils (organic C content).
The regressions between the Kd values and vadose zone characteristics obtained here may be improved by including other types of variables, such as clay mineralogy (Fruhstorfer et al., 1993; Cox et al., 1996; Herwig et al., 2001) and specific surface area (Madsen et al., 2000). The cost of measuring these characteristics is high, which may make their direct implementation in pedotransfer functions a counterproductive strategy. For instance, the regression between clay content and metamitron Kd had an R2 of 98% in our study, which is larger than that obtained by Madsen et al. (2000) for the same pesticide (96%) considering total Fe content as a principal explanatory variable, and several other secondary variables. The fact that these authors did not find any good relationship to the clay content may be explained by the low clay content range of their samples (<5–90 g kg-1) compared with the samples used here (14–782 g kg-1, Table 2).
The relationships found for the atrazine, isoproturon, and metamitron Kds are probably specific to the Bruyères-et-Montbérault catchment, and hence may be a good tool for sorption and transport studies within that catchment, but uncertain for extrapolation outside the catchment, especially when other types of geologic materials are present. Developing pedotransfer functions for vadose zone sorption properties will require databases with more general information, as has been done for the establishment of KOCs (Kds normalized by the organic C fraction of the soil) for pesticide sorption in topsoils (Hamaker and Thompson, 1972; Calvet et al., 1980; Hornsby et al., 1996). While important, such an approach is likely to be subject to the same type of drawbacks as those for the KOCs (Weber et al., 2000; Wauchope et al., 2002), such as differences in the measurement techniques, biases in the geographical distribution of the samples, incompleteness, and large uncertainties when used for predictions at local or regional scales. However, developing such databases is a necessary first step for objective risk assessments of groundwater contamination by pesticides. More experimental data, such as those presented here, are needed for this purpose.
CONCLUSIONS
Vadose zone sorption properties were explored at the scale of the 187-ha Bruyères-et-Montbérault catchment in France. The adsorption coefficients of three herbicides, atrazine, isoproturon, and metamitron, were found to be highly variable, with CVs >100%. These variations were driven by variability in the geological materials encountered in the catchment. The sorption isotherms of the three herbicides were found to be linear, so that the sorption properties could be characterized effectively using only the distribution coefficients Kds. Having linear sorption isotherms will greatly simplify the modeling of pesticide transport through the vadose zone to groundwater. The Kds could be predicted fairly accurately using the clay content (particle-size fraction < 2 µm, after organic matter and carbonate removal) of the materials. This result suggests that simple but efficient pedotransfer functions may be derived to predict vadose zone sorption properties.
ACKNOWLEDGMENTS
The author wishes to thank C. Labat, P. Vachier, and the INRA Agronomy Unit of Laon for their help during sampling of the vadose zone; J. Maucorps for his insights into the local geology of the catchment; and V. Etiévant for the pesticide adsorption measurements. Financial support was provided by ADEME, the French Agency for Environment and Energy Management, under contract no. 97.01.002.
REFERENCES
-alumina. J. Environ. Qual. 30:846–857.[ISI][Medline]This article has been cited by other articles:
|
|
 |
|
  A. Mermoud, J. M. F. Martins, D. Zhang, and A. C. Favre Small-Scale Spatial Variability of Atrazine and Dinoseb Adsorption Parameters in an Alluvial Soil J. Environ. Qual., August 8, 2008; 37(5): 1929 - 1936. [Abstract] [Full Text] [PDF]
|
|
