CFD Modelling of Pesticide Dust Drift from Seed Treatments During Planting

The seed of most agricultural crops is commonly treated with pesticides before sowing to protect the germinating seed and emerging seedling from pests and diseases. Seed treatment is a very focused chemical crop protection method, since the pesticide is applied directly onto the target, as opposed to spray treatment of the whole field and granule treatment in the furrows. Consequently, the application rates per hectare of seed treatments are relatively low. Furthermore, since sowing and crop protection are combined into one operation, the need for postemergence spraying is reduced. Despite these important advantages, however, a drawback of seed treatments has become evident in recent years. Dust particles containing the active ingredients may be unintentionally abraded from the seed coating layer after the application. This can occur during bagging, transport and handling, but the dust is mainly abraded during sowing due to friction in the planter. When vacuum-based precision planters are used, which employ a centrifugal fan to create a depression in the sowing elements for seed singulation, the abraded dust is emitted along with the exhaust airflow from the fan into the environment, where the pesticide dust can harm nontarget organisms. This phenomenon is known as dust drift. Various acute honeybee poisoning incidents have been linked with dust drift episodes during sowing of treated seed.

The main objective of this research was to improve the understanding of the drift of pesticide-laden dust released during sowing, by means of a combination of experimental work and a modelling approach. First, the physicochemical properties of seed treatment dust were fully characterized. Then, a CFD model of the dust aerodynamics in controlled wind tunnel conditions was developed and validated. Ultimately, this model was elaborated for the simulation of dust drift during planting in the field, in order to assess the relative importance of the impact of planter design, dust properties and wind conditions on realistic dust drift patterns.

In the first research chapter, the size, shape and internal porosity of abraded dust from treated seeds of various crops (maize, oilseed rape, wheat, rye, barley and pea) were studied by means of 3D X-ray microtomography (micro-CT). The study concluded that these dust properties are crop-dependent, likely because of differences in seed morphology and in the treatment process and recipe. Dust from maize, barley and oilseed rape seed was rather spheroid or disk-shaped, wheat dust particles were needle-shaped, and rye dust particles resembled thin flakes. Porosity increased with particle size and ranged from 0 to 80 %. In addition to the analysis of dust particle shape and porosity with micro-CT, the apparent density of the dust samples was measured with gas pycnometry. The apparent density was corrected with the porosity measurements to calculate the envelope density. The particle size distribution was determined with laser diffraction and was typically in the order of magnitude of tens to hundreds of micrometer. Ultimately, the active ingredient content was quantified with LC-MS/MS. It varied strongly between dust samples and generally decreased with particle size. The quantification of these physicochemical dust properties allowed the development of a CFD model of dust drift in controlled wind tunnel conditions.

A 3D CFD model was developed that simulates the trajectories of individual seed dust particles in an airflow by continuously calculating the magnitude of the drag force and the gravitational force that act on them, based on their previously measured size, density and shape. This method is called Lagrangian particle tracking. The model was validated with wind tunnel data from the Julius Kühn-Institut (JKI). In the wind tunnel trials, three size fractions of maize seed dust were released from a point source at three different uniform wind velocities. The deposition of the dust on the ground was measured at six distances from the dust source. The maize seed dust that was used in the wind tunnel experiment was completely characterized according to the previously described methodology. Subsequently, the configurations of the wind tunnel trials were replicated in CFD and the dust deposition patterns were simulated. The agreement between the simulated dust deposition and the experimental results was sufficient for the model to be elaborated in order to simulate dust drift in realistic field conditions.

In the final research chapter, 3D CAD models of seed drills were imported into the CFD model. The machine geometries were simplified and experimental air velocity measurements were applied as boundary conditions on the dust outlets. Wind was modelled as a horizontally homogeneous atmospheric boundary layer. A series of steady-state simulations with a stationary planter in the field was performed, with different wind conditions, planter designs and operating parameters, dust emission rates and physicochemical dust properties. The trajectories of the dust particles were calculated from the machine outlets to wherever the particles settled on the soil or left the computational domain. Dust particles were assumed to stick to the soil once they hit it, so secondary dust drift was neglected. Dust deposition in the field edges was calculated in postprocessing by integrating the dust mass flow rates on the soil in the driving direction, dividing the integration by the driving velocity, and considering the multiple passages of the planter in the field.

The model simulates pesticide concentrations in the air during drift, and on the soil or on the neighbouring vegetation after deposition. This makes the model particularly suitable for risk assessment studies. The input parameters of the CFD model can be changed in order to compare dust drift patterns in different sowing scenarios. Environmental pesticide concentrations can be simulated in typical sowing conditions and in worst-case conditions. By taking the behaviour of honeybees in the field into account, realistic pesticide exposure levels were estimated from these environmental concentrations. Furthermore, the model can be applied to design dust drift reducing measures, such as seed drill modifications, and to evaluate their effectiveness.