Modeling the Kinetics of the Competitive ... - ACS Publications

Feb 13, 2004 - OLE K. BORGGAARD, AND. PETER SESTOFT. The Royal Veterinary and Agricultural University,. Thorvaldsensvej 40, 1871 Frederiksberg C, ...
0 downloads 0 Views 72KB Size
Environ. Sci. Technol. 2004, 38, 1718-1722

Modeling the Kinetics of the Competitive Adsorption and Desorption of Glyphosate and Phosphate on Goethite and Gibbsite and in Soils ANNE LOUISE GIMSING,* OLE K. BORGGAARD, AND PETER SESTOFT The Royal Veterinary and Agricultural University, Thorvaldsensvej 40, 1871 Frederiksberg C, Denmark

The herbicide glyphosate and inorganic phosphate compete for adsorption sites in soil and on oxides. This competition may have consequences for the transport of both compounds in soil and hence for the contamination of groundwater. We present and evaluate six simple, kinetic models that only take time and concentrations into account. Three of the models were found suitable to describe the competition in soil. These three models all assumed both competitive and additive adsorption, but with different equations used to describe the adsorption. For the oxides, three additional models assuming only competitive adsorption were also found suitable. This is in accordance with the observation that the adsorption in soil is both competitive and additive, whereas the adsorption on oxides is competitive. All models can be incorporated in transport models such as the convection-dispersion equation.

Introduction The herbicide glyphosate (N-phosphonomethylglycine) was first introduced in 1974. Since then it has become very popular, and today it is the worlds leading herbicide. Global sales of glyphosate-based herbicides exceed those of the next six leading herbicides combined (1). It is adsorbed in soil by ligand exchange through the phosphonic acid moiety in a way similar to the adsorption of phosphate (2). The apparent similarity in adsorption mechanisms means that glyphosate and phosphate are probably adsorbed to the same sites in soil and that they compete for adsorption sites (3-6). The competition for adsorption sites may imply that phosphate influences the mobility of glyphosate and hence its transport to drains and to groundwater and surface waters. The influence of phosphate on glyphosate adsorption is of growing importance because of increasing use of glyphosate in the production of both conventional and gene-modified crops (7, 8). Additionally, many Danish and other European soils have been fertilized with phosphate for many years in excess of what is absorbed and removed by the crops (9). Therefore, phosphate has accumulated in the soils, leading to a reduction in the capacity to adsorb phosphate and perhaps also glyphosate. * Corresponding author telephone: +45 3528 2828; fax: +45 3528 23 98; e-mail: [email protected]. 1718

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 6, 2004

In previous studies, we have investigated the adsorption of glyphosate and phosphate to pure oxides, clay silicates, and contrasting Danish surface soils (3-5). These studies showed that pure iron and aluminum oxides adsorb much larger amounts of both glyphosate and phosphate than do pure clay silicates and whole soils and that there is a strong competition between glyphosate and phosphate for adsorption sites on the oxides and a weaker competition on the clay silicates and in the soils. Interactions, such as competition, between compounds in soil are often ignored when modeling the transport in soil, even though such interactions may have serious consequences for transport and leaching and hence for the contamination of groundwater (10, 11). Here we present simple models to describe the kinetics of the competitive adsorption and desorption of glyphosate and phosphate in soil. Such models may be used in transport modeling. As a comparison to the models in soil, models of the competition on the two oxides goethite and gibbsite are also presented.

Material and Methods Topsoil samples from five Danish soils were used in the experiments. The soils are representative of the soils found in Denmark (5, 12) and exemplify five soil orders including Alfisols, Inceptisols, Mollisols, Spodosols, and Ultisols that cover large areas epecially in temperate and boreal regions (13). Table 1 shows some basic characteristics of the soils. Besides from the soils the iron oxide goethite and the aluminum oxide gibbsite were used. Both of the oxides were synthetic, and the surface areas were 40 m2 g-1 for goethite and 45 m2 g-1 for gibbsite. A batch technique was used to study the adsorption of glyphosate and phosphate. 800 mg of goethite or gibbsite (giving 2 g of oxide L-1) or 40 g of soil (giving 100 g of soil L-1) was transferred to a glass bottle, and 381.5 mL of 0.1 M KCl was added as background electrolyte and 16 mL of 2.5% NaN3 (sodium azide) solution was added to prevent microbial growth. For one of the soils (Avedøre), an additional experiment was made with 50 g of soil L-1. Sample pH for the oxides was adjusted to 7.0 by addition of KOH or HCl. At time zero, 2.5 mL of 80 mM glyphosate containing 14Clabeled glyphosate or 80 mM phosphate solution was added to the mineral or soil suspensions, which were kept under constant magnetic stirring. The glyphosate or phosphate concentration in the reaction bottle was 0.5 mM. To follow the reaction between the minerals or soil and phosphate or glyphosate, 5.00 or 4.00 mL suspension aliquots were taken from the reaction bottles and filtered through a 0.45-µm filter, and the filtrate was used for phosphate or glyphosate determination. After 5 days for the oxides and 7 days for the soils, 80 mM glyphosate was added to the reaction bottles with phosphate, and 80 mM phosphate was added to the reaction bottles with glyphosate. The added volumes brought the concentration in the bottles to 0.5 mM. Again, samples were taken during the following 5 or 7 days. The glyphosate concentration in the clear filtrates was measured by liquid scintillation counting, and the phosphate concentration was measured by the molybdenum blue/stannous chloride method. The adsorption experiments were done in triplicate. All analyses were carried out with pro analyse or similar purity chemicals, and the water was triple de-ionized. More details about the experiments can be found in refs 4 and 5. 10.1021/es030572u CCC: $27.50

 2004 American Chemical Society Published on Web 02/13/2004

TABLE 1. Basic Characteristics of the Five Soils Useda

soil

classification, USDA

Avedøre Fladerne Foulum Jyndevad

Aquic Hapludalf Typic Fragiorthod Oxyaquic Hapludult Humic Psammentic Dystrudept Typic Argiudoll

Tåstrup

pH (CaCl2)

C (%)

NaHCO3 extractable P (mg kg-1)

total P (mg kg-1)

Feox (%)

Alox (%)

FeDCB (%)

AlDCB (%)

CEC8 (cmol + kg-1)

clay (%)

silt (%)

4.6 5.7 5.7 5.9

3.1 6.0 1.2 1.6

6.6 36.3 24.9 28.1

491 595 916 444

0.35 0.36 0.23 0.16

0.12 0.16 0.16 0.08

0.66 0.53 0.34 0.31

0.15 0.19 0.14 0.10

23.0 25.1 14.4 13.3

15 4 10 3

23 5 8 1

6.0

1.3

15.6

620

0.27

0.08

0.57

0.06

22.0

14

11

a

Feox and Alox are oxalate extractable iron and aluminum. FeDCB and AlDCB are dithionite-citrate-bicarbonate extractable iron and aluminum. CEC8 is the cation exchange capacity determined at pH 8.

The models are all simple models taking into account concentrations and time only. All models are given by differential equations with several parameters. The parameters in the models are estimated by a Java program that fits the models to a given set of data. For a given parameter estimate, the program solves the differential equations numerically and computes the differences between predicted and observed data. The best parameter estimate is that which minimizes the sum of squares of these differences, so the parameter estimation problem is a nonlinear least squares minimization problem. The differential equations are solved numerically by the routines Odeint and Stiff from Numerical Recipes in C (14), which were translated into Java by us. Minimization of the sum of squares of m nonlinear functions of n variables was performed using the Lmdif routine from a public domain version of MINPACK, originally developed by Garbow, Hillstrom, and More at Argonne National Laboratories. Lmdif was translated from Fortran to C by Steve Moshier and from C to Java by us (15). The Java program is included as Supporting Information.

FIGURE 1. Competitive adsorption and desorption of glyphosate and phosphate in the Foulum soil. Panel A represent scenario 1 where phosphate has been applied to the soil, there is equilibrium between adsorbed phosphate and phosphate in solution, and then glyphosate is applied at time 0. Panel B is scenario 2 where glyphosate has been applied to the soil, equilibrium between adsorbed glyphosate and glyphosate in solution has been established, and then phosphate is applied at time 0.

We tested 12 models for their ability to model the changes in concentration over time following application of either glyphosate or phosphate as described for scenarios 1 and 2 above. Only the most promising of the models tested are presented here. The first model (model 1) for the competition in soil is based on the assumption that three reactions are taking place. The first is a competitive reaction where the adsorbed species is desorbed by the added species, the second reaction is an adsorption of phosphate, and the third is an adsorption of glyphosate. Below the reactions are shown for scenario 2: k11

Modeling We wish to model two scenarios: In scenario 1, phosphate has been applied to the soil (or oxide), there is equilibrium between adsorbed phosphate and phosphate in solution, and then glyphosate is applied at time 0. Figure 1A shows the solution concentrations for one of the soils corresponding to this scenario. The model aims at predicting what happens to the phosphate and glyphosate solution concentrations after the application of glyphosate. This scenario simulates what happens when glyphosate-based herbicides are used on soils that have been heavily fertilized with phosphate. In scenario 2, glyphosate has been applied to the soil (or oxide), equilibrium between adsorbed glyphosate and glyphosate in solution has been established, and then phosphate is applied at time 0. Figure 1B shows the solution concentrations for one of the soils corresponding to this scenario. Again the model aims at predicting changes in solution concentrations after the application. This second scenario simulates what happens when a field is first sprayed with glyphosate and then fertilized with phosphate afterward.

Sg + Cp {\ } Sp + Cg 1 k2

k31

Cp {\ } Sp 1 k4

k51

Cg {\ } Sg 1 k6

where Sg is adsorbed glyphosate; Cp is phosphate in solution; Sp is adsorbed phosphate; Cg is glyphosate in solution; and k11, k21, k31, k41, k51, and k61 are rate constants. The model is given by

dCp ) -k11SgCp + k21SpCg - k31Cp + k41Sp dt dCg ) k11SgCp - k21SpCg - k51Cg + k61Sg dt

(model 1)

The second model (model 2) is based on the same assumptions as model 1, but the adsorption of phosphate VOL. 38, NO. 6, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

1719

and glyphosate is considered as being irreversible: k12

} S p + Cg Sg + Cp {\ 2

The fifth model tested for adsorption on the oxides (model 5) simply assumes that the only reaction taking place is a competition:

k2

k15

Sg + Cp {\ } Sp + Cg 5

k32

k2

Cp 98 Sp

The model is in this case given by

k42

Cg 98 Sg The adsorption of phosphate and glyphosate is modeled by an irreversible second-order model and is given by:

dCp ) - k12SgCp + k22SpCg - k32Cp(Q - Sp) dt dCg ) k12SgCp - k22SpCg - k42Cg(Q - Sg) dt

(model 2)

where k12, k22, k32, and k42 are rate constants according to the reactions and Q is the adsorption maximum. The third model (model 3) assumes that adsorption of the species applied is independent of the other species already present. The model also assumes that a small amount of the species already present will be desorbed by the added species and that this desorption will depend only on the concentration of the added species and not on how much is present of the adsorbed species. The desorbed species will then be adsorbed again. The reactions for the third model, scenario 2, are as follows: k13

Sg + Cp 98 Sp + Cg

k3

k43

Cg {\ } Sg 3 k5

The model is given by

dCp ) k33Sp - k23Cp dt (model 3)

where k13, k23, k33, k43, and k53 are rate constants according to the reaction scheme. As a comparison, we have also modeled the competitive adsorption on the two oxides gibbsite and goethite. For goethite and gibbsite, we tested models 1-3 and three other models because the mechanisms seem to be different on the oxides as compared to the soils (4, 5). The fourth model tested for the oxides (model 4) was adopted from ref 16, and it assumes that one of the species, in this case phosphate, is dominating in the competitive adsorption (has higher affinity for the adsorbent). The model is given by

dCp ) - k14Cp(Q - Sp) + k24Sp dt dCg ) - k34Cg(Q - Sp) + k44Sg dt

(model 4)

where k14, k24, k34, and k44 are rate constants and Q is the total adsorption maximum. 1720

9

dCg ) k15SgCp - k25SpCg dt

(model 5)

where k15 and k25 are the rate constants according to the reaction scheme. The last model (model 6) is much like model 4 but implies that the two species are more equal in the competition (16):

dCp ) - k16Cp(Q - Sp - Sg) + k26Sp dt dCg ) - k36Cg(Q - Sp - Sg) + k46Sg dt

(model 6)

When using the Java program to estimate the parameters in the models, the initial solution concentration of the species applied was also estimated by the program, whereas the initial concentration of the species already present and the amount adsorbed of this species were determined from measurements of the concentrations (4, 5). The initial adsorbed amount of the added species was set to zero.

Results and Discussion

k23

Cp {\ } Sp 3

dCg ) k13Cp + k53Sg - k43Cg dt

dCp ) - k15SgCp + k25SpCg dt

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 6, 2004

Table 2 shows the sum of squares (SS) for the different models tested on the soils, and Table 3 shows the SS for the models tested on the oxides. All parameters estimated can be found in the Supporting Information. The SS is the sum of the square of the deviations between the model and the data, and the lower the SS the better the model will fit the data. Generally, there are no great differences in the SS, which indicates that all models are suited to describe the competitive adsorption and desorption of glyphosate and phosphate. It should also be noted that the estimate of the initial concentration of the added species generally assumes reasonable values. For all soils, model 1 has the lowest SS for both scenarios 1 and 2. Figure 2 shows model 1 with data for the Foulum soil for scenario 2. The assumption behind this model was that, beside from competition between glyphosate and phosphate, there is also independent adsorption of the two species taking place. The model is in good agreement with what can be observed about the adsorption and desorption of glyphosate and phosphate: The added species is able to desorb the other species, but adsorption of the added species takes place in higher amounts than what can simply be explained by competition, and the species that was desorbed is re-adsorbed (5). For the Jyndevad and Avedøre soils, model 2 has a low SS for scenario 1 (model and data are shown in Figure 3). The assumption behind model 2 is the same as for model 1, but the independent adsorption was modeled by an irreversible second-order model. For scenario 2, model 3 had the lowest SS for the Foulum soil (Figure 4 shows this model and data). This model is based on the same observations as for model 1, but it is assumed that the species adsorbed has no influence on the adsorption of the species applied and that a small amount of the species already present will be desorbed by the added species and then will be readsorbed. All three models presented here are good at

TABLE 2. Sum of Squares for the Three Models Tested on the Five Soilsa model 1

model 2

model 3

soil

scenario 1

scenario 2

scenario 1

scenario 2

scenario 1

scenario 2

Avedøre Fladerne Foulum Jyndevad Tåstrup

0.015 0.0031 0.0041 0.0028 0.0043

0.0091 0.0084 0.0048 0.0084 0.014

0.018 0.0089 0.0058 0.0028 0.0051

0.012 0.0091 0.0047 0.0085 0.015

0.034 0.0038 0.0046 0.0028 0.0070

0.011 0.012 0.0046 0.0085 0.015

a Scenario 1 is a scenario where phosphate was applied first and glyphosate second. Scenario 2 is where glyphosate is applied first and phosphate second.

TABLE 3. Sum of Squares for the Six Models Tested on Goethite and Gibbsitea model 1

goethite gibbsite a

model 2

model 3

model 4

model 5

model 6

scenario 1

scenario 2

scenario 1

scenario 2

scenario 1

scenario 2

scenario 1

scenario 2

scenario 1

scenario 2

scenario 1

scenario 2

0.0025 0.0033

0.00063 0.0039

0.0026 0.0034

0.0013 0.0052

0.0026 0.0035

0.00092 0.0051

0.0026 0.0038

0.00085 0.0050

0.0026 0.0063

0.0034 0.0050

0.0026 0.0043

0.00090 0.0050

Scenario 1 is where phosphate was applied first and glyphosate second. Scenario 2 is where glyphosate is applied first and phosphate second.

FIGURE 2. Model 1 for scenario 2 shown with data from the Foulum soil.

FIGURE 4. Model 3 for scenario 2 with data from the Foulum soil.

FIGURE 5. Model 1 for scenario 2 with data for goethite. FIGURE 3. Model 2 for scenario 1 with data from the Jyndevad soil. describing the adsorption/desorption of glyphosate and phosphate, and the mechanisms behind the models agree with what can be observed about the adsorption/desorption of glyphosate and phosphate for both scenarios 1 and 2. Which model should be chosen will depend on the soil type and the mechanisms assumed to take place. Also, drawing data and the model may assist in choosing since the models may look quit different, as can be seen by comparing Figure 2 and Figure 4, which both show the data for the Foulum soil, scenario 2, but with two different models. It should be noted that model 1 has one parameter more than the other two models; therefore, it may be easier for the model to fit the data.

In the soils, glyphosate and phosphate adsorption is similar, with almost equal affinities, and the interactions are both competitive and additive (5). In contrast to this, the adsorption/desorption on the oxides is almost exclusively competitive, with phosphate being the strongest competitor (4). We tried to model adsorption/desorption on the oxides with the same models as for the soils along with three additional models assuming a more strict competition than in models 1-3. The three additional models have also been tested on the soils (data not shown) but did not perform well. In Table 2, the SS for the six models tested on the oxides are presented. From the table, it can be seen that model 1 has the lowest SS for both oxides and both scenarios, indicating that this model is also capable of adequately VOL. 38, NO. 6, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

1721

describing the competition on the oxides. Figure 5 shows this model with data for goethite, scenario 2. As for the soils, the low SS may be due to the number of parameters. Model 5, which assumes only a simple competitive reaction, gives high SS for goethite scenario 2 and gibbsite scenario 1. This indicates that it is not just a simple exchange of glyphosate for phosphate or phosphate for glyphosate that takes place; additional reactions are involved. Model 4, which was proposed by Murali and Aylmore (16), and model 6, which is a modification of model 4, yield relatively low SS values, indicating that these two models are also suited to describe the competition on the oxides. Moreover, the theory behind these two models agree very well with what is known about the competition between glyphosate and phosphate for adsorption sites (4, 5) because model 4 assumes that phosphate is the dominating species, which agrees with the adsorption on goethite, and model 6 assumes that both adsorbing species have comparable affinities, which is the case for gibbsite. Altogether model 1 gives the best description (lowest SS) of adsorption and desorption on both oxides and in soil for both scenarios 1 and 2, and the assumptions behind the model closely simulate observed characteristics of the adsorption (4, 5). The main drawback of this model is the rather high number of adjustable parameters. Model 3 is also capable of describing adsorption/desorption for both scenarios for oxides as well as soils, and this model is also based on assumptions agreeing with observations. For the oxides, models 4 and 6 proposed by Murali and Aylmore (16) also fit the data well in accordance with the observed strong competition between glyphosate and phosphate for adsorption sites, in particular on goethite (4). All the six models presented and evaluated here may be incorporated into transport models such as the convectiondispersion equation to better simulate the risk of leaching of glyphosate (and phosphate).

(1) Monsanto. Monsanto Company 2002 Annual Report; Available from http://www.monsanto.com (accessed July 2003). (2) Sheals, J.; Sjo¨berg, S.; Persson, P. Environ. Sci. Technol. 2002, 36, 3090-3095. (3) Gimsing, A. L.; Borggaard, O. K. Clays Clay Miner. 2001, 49, 270-275. (4) Gimsing, A. L.; Borggaard, O. K. Clay Miner. 2002, 37, 509515. (5) Gimsing, A. L.; Borggaard, O. K.; Bang, M. Eur. J. Soil Sci. 2004, 54 (1), available online at: http://www.blackwell-synergy.com/ rd.asp?code)EJS&goto)journal. (6) Hance, R. J. Pestic. Sci. 1976, 7, 363-366. (7) Dion, H. M.; Harsh, J. B.; Hill, H. H. J. Radioanal. Nucl. Chem. 2001, 49, 385-390. (8) Franz, J. E.; Mao, M. K.; Sikorski, J. A. Glyphosate, a unique global herbicide; American Chemical Society: Washington, DC, 1997. (9) Del Campillo, M. C.; van der Zee, S. E. A. T. M.; Torrent, J. Eur. J. Soil Sci. 1999, 50, 391-399. (10) Zsolnay, A. Chemosphere 1992, 24, 663-669. (11) Kalatzis, A.; Garcı´a-Delgado, R. A.; Pang, T.-K.; Koussis, A. D.; Bowers, A. R. Water Resour. Res. 1993, 29, 2241-2248. (12) Møberg, J. P.; Petersen, L.; Rasmussen, K. Geoderma 1988, 42, 295-316. (13) Soil Survey Staff. Soil Taxonomy, A Basic System of Soil Classification for Making and Interpreting Soil Surveys; United States Department of Agriculture: Washington, DC, 1999. (14) Press, W. H.; Teukolsky, S. A.; Vetterling, W. T.; Flannery, B. P. Numerical Recipes in CsThe Art of Scientific Computing, 2nd ed.; Cambridge University Press: Cambridge, UK, 1992. (15) More´, J. J.; Garbow, B. S.; Hillstrom, K. E. User Guide for MINPACK-J; Technical Report ANL 80-74; Argonne National Laboratories: 1980; Software from http://www.netlib.org/ minpack/. (16) Murali, V.; Aylmore, L. A. G. Soil Sci. 1983, 135, 143-150.

Supporting Information Available

Received for review July 31, 2003. Revised manuscript received December 15, 2003. Accepted December 26, 2003.

The Java program used for fitting and tables with all the estimated parameters. This material is available free of charge via the Internet at http://pubs.acs.org.

1722

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 6, 2004

Literature Cited

ES030572U