Bekæmpelsesmiddelforskning fra Miljøstyrelsen, 107 – Kolloid-faciliteret transport af glyphosat og pendimethalin – Kvantificering og modellering

Kolloid-faciliteret transport af glyphosat og pendimethalin

Summary

Aim and conclusions

The project (Quantification of Colloid-facilitated Transport of Glyphosate and Pendimethalin) was carried out as a laboratory based follow up of a field based monitoring and modelling study (Facilitated Transport of Pesticides, Holm et al., 2003). They concluded that a study of the basic phenomena behind transport of sorbing pesticides and the related process kinetics was necessary under controlled conditions in the laboratory.

The aim of this project was to investigate colloid generation and to quantify the importance of colloid facilitated macropore transport of glyphosate and pendimethalin in the unsaturated zone in structured agricultural soil. The knowledge gained was used for development of a model component that should be used for simulation of colloid facilitated pesticide transport in the upper part of the root zone on usaturated agricultural soils with macropores (> approx. 1 mm). Furthermore, the colloid composition has been investigated, because it is important for the ability of the colloids to sorb pesticide.

Two additional goals have been to investigate the importance of soil structure, exemplified through different types of soil tillage, on mobilisation and transport of colloids in the unsaturated zone in the presence of macropores, and to try to disclose in which parts of the soil column, the pesticide carrying colloids are generated.

On the basis of the project results in can be concluded that colloid transport can be very important for both glyphosate and pendimethalin. For glyphosate, 63-74% of the total leached mass was bound to colloids when the spraying took place on ploughed soil. For pendimethalin 75-80% of the total leached mass was transported bound to colloids, but doubt remainswith respect to which fractions the pendimethalin is bound to.

The most important effect of soil structure was on the generation of colloids. Considerably more colloids were generated on the ploughed columns than on the minimally treated. Observable splash erosion was found on the ploughed columns only. The total pesticide leaching could not be shown to be different for the ploughed and minimally treated soil. For glyphosate, however, the amount of colloid-bound glyphosate made up only 11-22% on the minimally treated fields. This effect cannot be attributed to the soil structure alone. It was probably also of importance that dissolved glyphosate was washed off the plant material on the surface and that it, due to sorption kinetics and the transport velocity, did not reach equilibrium with the surrounding soil and colloids.

None of the generally accepted leaching models and procedures can describe the observed colloid facilitated transport of glyphosate and pendimethalin as well as of the dissolved glyphosate.

On the basis of colour trials, turbidity measurements and pesticide content of the leached colloids it is concluded that the pesticide carrying colloids primarily are generated in the upper centimetres of the soil. Further down in the soil columns blue coloured areas were found only along active macropores. Colloids with pesticide may move down through the profile stepwise, but due to the flow conditions in the columns during the experiment (negative suction in the soil matrix under the A-horizon) it will be difficult to re-mobilise colloids that has been sucked into the matrix.

The modelling showed that the improvements in process descriptions that were carried out in the project led to good simulation of flow conditions and bromide transport. When the flow and solute exchange parameters were determined, it was possible to simulate the two very different columns with the same soil treatment with a limited variation of the colloid parameters. For columns with different soil treatment, the parameters were considerably more different. This shows that much of the difference between two similarly treated columns was dictated by the flow pattern.

The observed patterns of colloids and pesticide transport could be generated by the model with different combinations of parameters for colloid generation and kintics. It is therefore not clear exactly how the processes complement each other. It was expected that there could be time differences in when colloid generated by the different processes would occur in the effluent, but while some of the results from the columns indicated this, this pattern was not found in the modelling work.

The investigation

The leaching of two pesticides through macropores in the unsaturated zone was investigated and the dependency of the leaching on colloid-facilitated transport was evaluated. The selected pesticides were glyphosate and pendimetalin, which are both strongly sorbing, but chemically very different. Each pesticide was added to 4 columns sampled from the soil surface to 50 cm’s depth. Of these, 2 columns were sampled from a ploughed field and 2 from a minimally tilled field. The columns were structurally very different due to the different treatement, but are not expected to be representative of all aspects of the different types of soil tillage. For each column the flow pattern was first investigated using bromide transport. In these initial experiments, the relationship between turbidity and colloid content was also determined. Radioactively tagged pesticide mixed with formulated pesticide was added to selected columns and 3 irrigation events were performed. The effluent was investigated for pesticide content, pH, conductivity and turbidity, and filtered through a 0.02 mm-filter, after which also the filtrate was analysed for tagged pesticide. A few effluent samples were analysed for total organic carbon (TOC) and dissolved organic carbon (DOC). From the ploughed columns, splash eroded material from the surface of the column splashed onto a collar fixed on the column. The splashed material was collected. With the chosen methods, colloids in this study are greater than 0.02 mm and smaller than 30-50 mm.

The amount of dispersible material was determined for the soil and for the splash-eroded material with the aim of evaluating the potential amount of colloids. Furthermore, K_d-values were determined for the sorption of the two pesticides to the soil used in the experiment and to straw. Chemical analyses were made on selected stock solutions, effluents and filtrate from the dispersion experiments in order to test the agreement between these and concentrations calculated based on measurement of radioactivity. Furthermore, the colloids in the effluents were studied and sorbing constituents characterised through infrared spectroscopy (IR), carbon-13-NMR spectroscopy, transmission electron microscopy combined with element analysis, and X-ray diffraction.

At the end, the colour Brilliant Blue was added to the columns in order to enhance visualisation of the flow pattern. Samples were taken from each layer of four columns for pesticide determination with the aim of studying the relationship between the transport pathways for pesticides and colour and to establish a mass balance.

It was assumed that glyphosate could be marked with ¹⁴C, while pendimethalin could be marked with ³H. Hence, trials could be carried out on the same columns with both pesticides. However, the ³H-marked pendimethalin turned out to be unstable. Measurement of pendimethalin thus also had to be based on ¹⁴C-marking and, as a consequence, the trials with the two different compounds had to be carried out on different columns. This meant that some columns had to be used, although the flow on these columns had been shown by the bromide trials to be limited. The number of replications became less than originally planned and the variation between columns became greater.

Furthermore, it was expected that the colloid-bound pendimethalin could be collected on the selected filters. This turned out not to be the case, (at least not for the tagged fraction). The same phenomenon was reported by Holm et al.(2003), but it was explained to be caused by desorption during the time elapsed between sampling and analysis. To investigate whether the pendimethalin was present in solution, two extra column experiments were carried out, from which the effluent was filtered and nano-filtered.

Results

- colloid generation

The generation of colloids was larger on the ploughed soil than on the minimally tilled soil. In average over three events 50-97 mg/l leached from the minimally tilled columns and 183-295 mg/l from the ploughed columns in the experiments with pesticides.

For the minimally treated columns where damage to the surface was registered during the sampling of the column, the maximal turbidity was considerably larger than for the undamaged columns (about 20 times larger than for the undamaged columns) during the initial trial. The difference became smaller with time, which is likely to be due to a consolidation of the surface and the fact that moss on the surface grew during the experiment.

Soil moisture content appears to have an effect on colloid generation. The moisture content in the columns was higher at the beginning of event 2 than at the beginning of event 1, and except for one column the colloid generation was larger for event 2 than for event 1. The total amount of colloid bound pesticide was, however, constant or fell with event number.

The content of the colloids leaving the columns was investigated with a high degree of temporal resolution, and it was found that the mineralogical content and content of functional organic groups (carboxylic groups and alkenes) was constant over time. It is therefore probable that the colloid generation took place in identical areas in the column (the same horizon) and was caused by similar mechanisms during the experiment. This is in contrast to earlier performed field trials, where the content of organic matter was largest in the first samples of a drain period.

The mineralogical studies of the colloids aimed at identifying specific reactive structures and surfaces of minerals that can function as sorption sites for the pesticides. The study of natural, highly polymeric carbon compounds showed a completely new type of sorption site in Danish soils: Carbon having a graphite-like structure and with a significant oxygen content. These compounds may be very important for the sorption of organic compounds with aromatic structures. They were found in the soil in all horizons, and probably in a wide size range.

Mineralogical investigations of whole colloid samples showed the presence of microcrystalline goethite in physical close contact with organic carbon compounds. Goethite is the sorption site in the soil with the highest expected affinity for glyphosate, but it is unknown what the close association means for the sorption properties. Such physical associations may explain that the properties are not proportional to the content of specific minerals. Aluminium silicates were the most common type of mineral identified.

After the leaching trials the very highest concentrations of both glyphosate and pendimethalin was found in the upper 0.5 cm of the soil column, 11 and 8 mg/kg for glyphosate and 14 and 25 mg/kg for pendimethalin, for the ploughed and minimally tilled columns, respectively. The concentration in the surface layer was slightly higher in the minimally tilled field, which agrees with the findings of the colour trial, where the thickness of the upper, fully coloured layer was thinnest in the minimally tilled soils, but this relation was not found for glyphosate.

No tagged pesticide was found in randomly sampled soil samples outside the coloured areas. Only in one sample from 25-30 cm’s depth from the minimally tilled soil was glyphosate identified above the detection limit. Tagged pesticide found under the upper soil layer was found in blue-coloured areas around flow-active biopores and cracks. The pesticide concentration in these blue-coloured areas decreased with depth. The same trend was observed for the average concentration in randomly sampled soil.

The generation of splash-eroded material was significantly different between the ploughed and minimally treated columns. This could indicate that a larger part of the colloids on the minimally treated columns is generated in the soil. However, the pesticide concentration of colloids generated on the minimally treated columns was larger than the pesticide concentration of colloids generated on ploughed columns, except for column 2 (minimally treated), which had a very low leaching. For glyphosate the concentration was, as mentioned approximately 9 and 11 mg/kg soil in the upper 0.5 cm of column 6 (minimally treated) and column 4 (ploughed) and the concentration of the leached colloids in 3^rd event was 10 and 14 mg/kg, respectively. Below 0.5 cm’s depth, the concentration in scrapings from coloured macropores was always below 1 mg/kg and decreased strongly with depth. For pendimethalin the concentration on leached colloids was lower (approximately 4 and 2 mg/kg for column 4, minimally treated) and column 2 (ploughed), but the largest concentration measured in coloured macropores was 0.4 mg/kg, again decreasing strongly with depth. The results thus indicate that the colloids in both cases primarily were generated in the top layer.

The fact that the electrical conductivity in these trials corresponded to the conductivity in the soil moisture indicates that the irrigation water was mixed with soil moisture. As the colour experiment does not indicate a lot of contact with soil moisture, this contact must take place close to the surface and probably in what Gao et al. (2004, 2005) call the exchange layer. In experiments with columns with protected surface, the conductivity of the leaching water often approach the conductivity of the rain water (see i.e. Laegdsmand et al. (2005); Laegdsmand et al. (1999; de Jonge et al. (2000)).

An almost inverse relation between TOC/DOC and pesticide was found on the minimally tilled columns. On the ploughed soil a much better correspondence was observed between the development over time of TOC/DOC and dissolved (<0.02 mm) pendimethalin and colloid-bound glyphosate, respectively. This could indicate that the generation of TOC/DOC took place at the same place as the generation of colloid- and pesticide in the ploughed columns, while this did not seem to be the case in the minimally tilled columns.

-glyphosate

The total leaching of glyphosate varied between 0.8-38.3 mg per column, equal to 0.007-0.32% of the added amount. The average concentrations generated for each event varied between 0.04 and 24.8 mg/l.

The leaching trials with glyphosate showed that in the ploughed soil a large part of the glyphosate (in average 63-74%) was transported on colloids in the effluent. Only a smaller fraction (in average 11-22%) of the glyphosate was transported on colloids in the minimally tilled field.

The difference between the amount of total leached glyphosate from ploughed and minimally tilled columns was not clear. For both treatments one column showed high and one showed low leaching. Also, no general relationship between colloid concentration and glyphosate concentration could be derived. The amount of colloid-associated glyphosate was almost constant or decreased with increasing event number, while the total leached amount of glyphosate increased with number of event on two columns (one of each treatment), decreased on one column and had its maximum during event 2 for the last column.

The correlation between colloid concentration and glyphosate concentration differed between columns and events. The highest R²-value found was 0.95, but values down to 0 were also observed. For some columns and events, there was a clear difference between the first effluent samples and later ones. The R²-values increased considerably for these events when the first effluents were excluded from the analysis. One could argue that the first effluents were dominated by colloids that were not generated in the surface layer, while the colloids later in the event primarily came from the surface, where the pesticide was added. Alternatively it could be caused by a higher fraction of coarse particles (with a smaller surface area) in the first effluent samples. Experimentally it was attempted to avoid this effect by allowing the samples to settle for 1 min before turbidity was measured, but some effect cannot be excluded.

Studies of sorption/desorption kinetics showed that if the effluent contained much colloid-bound glyphosate a relatively fast desorption of glyphosate took place from the colloids (within 30 minutes). This effect was found in effluent from columns sampled from ploughed soil. The effluent from minimally treated soil contained more dissolved glyphosate and little colloid-bound glyphosate. Here sorption of glyphosate to the colloids took place in the effluent (within the same time scale). Measurements of desorption from splashed material showed a similar time dependent desorption, with strong desorption taking place within the first hour. There was no substantial difference between sorption of glyphosate in experiments carried out with or without addition pesticide formulation.

The results indicate that the conditions on the soil surface are very important for the mode of transport of glyphosate. Glyphosate sorbs poorly to stubble and moss (K_d experimentally determined to »0), and it was therefore washed off the organic material during the rain events. In case of strong precipitation relatively close to the time of spraying, it is probable that the glyphosate, due to the sorption kinetics, does not have time to sorb to the soil, as it happened in these trials. When glyphosate was sprayed on bare, newly ploughed soil, it sorbed to the soil particles and was primarily transported bound to soil particles/colloids. This result is also of interest in connection with ploughing of sprayed stubble and straw. Glyphosate on the surface of ploughed back stubble and straw will probably be mobilised rather easily in connection with water flow in the soil, particularly because observations show that the flow often takes place along the plough furrows, where the stubble is situated. Due to the physical distances in the soil, diffusion of glyphosate from the stubble to sorption sites in the soil will take considerable time and the glyphosate can therefore probably be mobilised for a long period of time.

- pendimethalin

The total leaching of pendimethalin varied from 11.0-32.7 mg per column, equal to 0.12-0.43% of the added amount. The average concentrations generated for each event varied between 1.21 and 9.3 mg/l.

The leaching experiments carried out for pendimethalin showed that the largest amount of the radioactively tagged pendimethalin passed through the 0.02 mm-filter. It was therefore was classified as “in solution”. Still, the correlation between the colloid concentration (measured as colloids > 0.02 mm) and the pendimetalin concentration was high (R² = 0.42-0.92). Therefore, further experiments with fractionation of effluent and filtrate by nano-filterning over a 500 Dalton membrane was carried out. Approximately 20-25% of the compound could be found in the fraction < 500 Dalton, which primarily will consist of free pendimethalin in solution. The rest of the pendimethalin was bound to particles, colloids or larger organic molecules and was found in other fractions. The majority of the pendimethalin (75-80%) is therefore, in practical terms, “colloid”-bound.

As for glyphosate, there was no substantial difference in the amount of total leached pendimethalin from ploughed and minimally tilled columns. For both treatments, one column resulted in high and one showed low leaching. No general relationship between colloid concentration and pendimethalin concentration was found. The fraction >0.02 mm was, however, largest in the first event or almost constant through the three events (as for glyphosate). For three out of four columns, the total pendimethalin leaching increased with number of event, while the leached amount reached maximum in the second event for the last column.

More detailed comparisons between turbidity and pendimethalin concentration showed a substantially different pattern than for glyphosate. The first sample always had a high turbidity and a high concentration of pendimethalin. For all of the following measurements, the pendimethalin concentration was decreasing slowly over time or it remained almost constant. Similarly, the turbidity decreased with time.

The experiments carried out to evaluate the sorption/desorption kinetics for pendimethalin showed that no or a very week sorption/desorption of pendimethalin took place to/from the colloids (>0.02mm) after the effluent had been sampled. The phases therefore appeared to be at equilibrium. Samples of the splash-eroded material desorbed pendimethalin. The interpretation of the results is difficult because the filtered samples contained both sorbed and free pendimethalin, and because sorption might have taken place to the bottle lids used in the experiment.

The chemical analyses of pendimethalin in effluents from columns showed lower concentrations than the determinations based on radioactivity. It was tested whether poor mixing at the application of pendimethalin on the columns caused the disagreement. The results showed that mixing could be of importance, but a final conclusion could not be drawn. The distribution of pendimethalin between the large fractions differed depending on whether the distribution was based on chemical or radioactive analysis, and this difference cannot be explained on the basis of the experiments carried out. Both the chemical and radioactive analyses showed that less than a third of the pendimethalin was present in soluble form.

- modeling

The modeling showed that the improvements in process descriptions made during the project resulted in good simulations of the water flow and bromide transport. There were, however, considerable differences between the different columns, and they had to be parameterized individually. When the observed variation between the columns was taken into account via the flow model, the variation between parameters for columns with different soil treatment was considerably larger than the variation between columns with the same treatment. In spite of the large variation between columns, systematical variation could thus be observed.

The observed patterns of colloids and pesticide transport could be generated by the model using different combinations of parameters for colloid generation and kinetics, and a unique understanding of how the processes complement each other could not be reached. It was expected that there could be temporal differences as to when colloid generated by different processes appeared in the leachate, and while some of the results from the columns also indicated so, this did not appear in the simulations.

To be able to describe the leaching of glyphosate on the minimally tilled soils, it was necessary to assume that the K_d-value for the topmost part of the column was much lower than the K_d-value for soil. This is in correspondence with the K_d-measurements for straw, and the simulations also generated the observed distribution between colloid bound and soluble glyphosate at the bottom of the column. There is thus good correspondence between the earlier described interpretation of the glyphosate trials and the model simulations.

- perspectives

The project gave rise to further investigations and studies:

- A verification of whether glyphosate is mobilised as described from incorporated stubble and straw, and whether this practice therefore is appropriate or can be modified to minimise leaching risk.

- a verification of whether the transport of pendimethalin with colloids of less than 20 nm is an artefact of the use of radioactively tagged pendimethalin or whether it is a phenomenon that is valid for other pesticides sorbing to organic material. This is of importance also for future trials because the size fractionation depends on the sizes of the transporting colloids.

- Adsorbing agents along the transport pathways are shown to play an important role and methods should be developed to characterise these agents in immediate vicinity of the transport pathways.

- All presently used models for pesticide registration treat vegetation cover on the surface as crop. None of them treat vegetation cover in the form of weeds, straw and moss. The fact that glyphosate is caught on but does not sorb to, at least straw and moss, and therefore can be mobilised in soluble form for a period after spraying, is thus not taken into account. It should be investigated whether this phenomenon is of more general validity and it should be determined how it then should be treated modelingwise.

- Colloids are not treated in existing registration models. The possibility of mobilisation of glyphosate in soluble form from incorporated straw does not exist either in the existing (FOCUS-) registration models. Th results imply that an evaluation of glyphosate must take into account wash-off from plant material, the uneven distribution of plant material containing glyphosate in the soil (why ordinary assumptions about (equilibrium-) sorption does not hold true) and the kinetics of the sorption process for glyphosate in solution. Colloid transport is important when glyphosate is sprayed on soil and should be taken into account when evaluating the leaching potential of glyphosate. Colloid transport should also be included in an assessment of the leaching potential of pendimethalin.

- The developed process descriptions are clear improvements with respect to the the description of macropore flow and bromide transport compared to the PestSurf model. The ability to generate colloids by two different processes is also a clear improvement, as is the introduction of process kinetics. The main problems in connection with the developed descriptions are expected to lie in

i. Experience with parameterisation for different soil types and conditions at plot/field scale. The very large variation observed between the columns will, to a certain extent, level out on a larger scale.

ii. Description of the development of the model parameters over time.

iii. Filtering and processes deeper in the profile. The columns investigated here are mainly concerned with “cut through” macropores. In nature, the macropores can end in a drain but the can also stop in the soil or in cracks with a smaller conductivity than the macropore. These processes are being investigated in an ongoing project, ”Multi dimensional modelling of water flow and solute transport”, in the upper 1-2 m of the soil in systems with field drains.

iv. The possibility of describing mobilisation of glyphosate from incorporated straw is not developed in the project and its relevance will depend on whether the process can be proven to exist in practise.

- Before it is considered to expand the colloid model with dissoved organic carbon it should be proven that this is a generally important factor for transport. If it can be proven that DOC is important for transport of strongly sorbing pesticides, the starting point may be the description implemented in the Daisy model or it may possible to treat it as desorption from an infinite pool.