Airborne-Spray-Drift Collection Efficiency of Nylon Screens: Measurement and CFD Analysis

Abstract

Pesticide application is essential for improving crop productivity; however, undesirable pesticide drift must be mitigated because of its adverse impacts on humans, the environment and ecosystems. The collection and accurate quantification of airborne droplets are key elements involved in identifying the spatial and temporal dispersion of off-target spray movement. Various types of passive and active collectors have been deployed to measure airborne spray drift; however, the collection efficiencies of only a few samplers have been verified. This study evaluated the collection efficiency of two airborne-spray-drift collectors using an experimental drift wind tunnel. The airborne spray drifts were quantified by a total organic carbon analyser and validated by comparison to measurements using liquid chromatography with tandem mass spectrometry. Computational fluid dynamics (CFD) simulations were used to explore the effects of droplet size and wind speed on the collection performance. It was found that nylon screens, passive samplers, captured 57.9–88.1% of the airborne spray drift. These results are considered reliable and are comparable to those found in the literature. Additionally, the CFD results demonstrated that the collection efficiency increased with droplet diameter. An increase in wind speed improved the collection efficiency of fine droplets (≤100 μm diameter); however, wind speed had no significant influence on the collection of coarse droplets. These measurements, alongside the aerodynamic approach adopted in this study, can provide a comprehensive understanding of the collection performance of nylon screens.

1. Introduction

High-value crop production is crucial for ensuring global food security. The application of pesticides can protect crops from weeds, pests and diseases, which can significantly deteriorate the quality of crop products. However, pesticides deposited off-target can cause damage to neighbouring land, the environment and ecosystems, raising concerns related to environmental contamination, dietary risks and human health [1,2]. Misdirected pesticides can have serious economic consequences for organic farming since organic products exposed to pesticide drift can lose their value or end up being sold as non-organic products, resulting in significant economic losses [3,4].

Spray drift can take the form of gaseous diffusion (vapour drift) or droplet drift. Whilst gaseous diffusion is caused by volatile active substances with a relatively smaller amount of chemicals [5], droplet drift gives rise to greater environmental concerns due to the larger amounts of wind-blown mass involved [6]. Spray drift, generally referred to as droplet drift, according to ISO22866 [7], can be measured by the amount of liquid deposited on the ground, which is used to further assess the risk of contamination of crop fields and surface waters, or airborne collection, which is related to personal exposure and predicting potential drift distances [8,9]. To predict spray drift under unpredictable wind conditions, it is important to identify both the vertical and horizontal distributions of airborne concentrations or fluxes released from the target application area.

Airborne spray drift can be measured by passive collectors, such as pipe cleaners [10,11], ball-shaped collectors [10,12], tubes [13,14], plastic or nylon lines [11,14,15,16,17,18,19,20], and nylon screens [10,21]. The major drawbacks of the aforementioned methods are single-point measurements. Furthermore, these methods require time-consuming analysis and produce time-averaged sampling results, especially when multiple-point measurements are required [15]. To overcome these limitations, the recent technologies, such as light detection and ranging (LiDaR) [15,22,23] and laser particle counters [24], have been employed for remote real-time measurements of airborne spray drift. However, these methods have difficulties in quantifying the mass of spray drift using the signals received [25]. They are also not yet available commercially for practical field applications.

Despite their limitations, conventional passive collectors are simple to use for investigating the horizontal and vertical distribution of spray drift using multi-point observations. They also provide data concerning the mass of drifted pesticide. This is why passive collectors are still widely used and are considered reliable samplers. Standardised protocols, such as ISO22866 [7] and ASAE S561 [26], prescribe the minimum of experimental descriptions required to ensure that the results of different field experiments carried out globally are comparable. However, it is necessary to investigate the performance of the collectors, particularly to determine their collection efficiency, to help improve both the efficiency and reliability of the measurements. There is a lack of studies examining the collection efficiency of collectors [27], and only a few studies [10,12,21,28] have investigated the collection efficiency of collectors based on variable wind speeds. Nevertheless, collection efficiency (in the strict sense) has been examined only by Fox et al. (2004) [21], while others studies have measured the relative collection efficiencies, expressed as a percentage of the application rate or the largest spray deposit. Fox et al. (2004) [21] measured the collection efficiency by comparing the spray deposit on samplers with the total airborne mass passing through the measurement location. Measuring the total airborne mass requires an understanding of the aerodynamic processes around passive collectors, such as air penetration, airflow dispersal and droplet deposit. Active samplers have been used by drawing the air with air pumps [10,12,29,30] or moving battery-powered rotary samplers [28] to overcome the inaccuracies induced due to the lack of aerodynamic knowledge of passive collectors; however, active samplers are less cost-efficient compared to passive ones. Collection efficiency should be investigated regardless of the type of collector used for spray-drift measurement, and in-depth aerodynamic studies can help look into this collection process.

The objective of this study was to determine the collection efficiency of airborne-spray-drift collectors. This efficiency was investigated in two steps through a controlled experiment (using a small spray-drift tunnel) and aerodynamic analysis (using a computational fluid dynamics (CFD) model). The studied collectors will be applied to measure spray drift in agricultural fields and quantify the horizontal and vertical distributions of spray drift.

2. Materials and Methods

2.1. Spray Collection Experiment

2.1.1. Drift Tunnel Setup

A small spray-drift tunnel was developed to generate airborne drift of sprayed droplets and test the collection efficiency of the collectors (Figure 1). The tunnel comprised two sections; the spraying and collecting sections. The spraying section was a chamber with dimensions of 1500 mm (H) × 1200 mm (L) × 500 mm (D). A flat-fan nozzle (XR11002, Teejet Technologies, Springfield, IL, USA) was mounted under the ceiling of the chamber to spray droplets downward at the working pressure of 0.3 MPa. The nozzle produced fine-sized droplets at 0.3 MPa based on the manufacturer’s guidelines. The collecting section with dimensions of 350 mm (H) × 1150 mm (L) × 350 mm (D) was connected to the lower part of the spraying section so that some droplets, called airborne droplets, could be blown by the air exiting the spraying section. An air inlet with dimensions of 350 mm (H) × 350 mm (D) was located at the other side of the spraying section. An exhaust fan (SLF-300D, Sung Il E-B Corp., Seoul, Republic of Korea) with a capacity of 3568 m³ h⁻¹ was mounted at the end of the collecting section and aspirated the airborne droplets into the collecting section. In the collecting section, various collectors were placed in the middle of the cross-section. Three 350 mm × 350 mm nylon screens were also mounted at the front of the fan to recover all airborne droplets and stabilise the cross-sectional distributions of pressure and air velocity. The first nylon screen among the three recovery screens could be moved to the test location, where collectors would be placed, to measure the total amount of airborne droplets at the test location, which was then used to calculate the collection efficiency of the collectors.

The wind speed in the collecting section was controlled by an AC controller connected to the exhaust fan. Although the fan created swirling airflows, the three nylon screens contributed towards stabilising the pressure distribution over the cross-section, homogenising, to some extent, the distribution of longitudinal velocity over the cross-section. Air velocity was measured at nine points in a cross-section of the test location where the collectors would be placed. The standard deviation between the equally distributed nine points, which were equally distributed as 3 × 3, was ±0.09 m s⁻¹ for the average wind speeds of 1 m s⁻¹, ±0.26 m s⁻¹ for 2 m s⁻¹ and ±0.48 m s⁻¹ for 3 m s⁻¹, indicating uniform velocity with the coefficient of variation ranging from 9 to 16%.

2.1.2. Airborne-Drift Collectors

Two collectors were selected to evaluate their collection efficiency. Nylon screens are one of the well-known passive collectors [10,21], whose collection efficiency has been tested in a previous study [21]. Since the nylon screens showed different collection efficiencies with respect to the mesh number or air resistance, the effect of aerodynamic characteristics of the nylon screens on the collection performance was investigated through experiments and CFD simulations.

The nylon screen with the mesh number 100 was used as the first collector as well as for three recovery screens in the collecting section. The thread diameter of the nylon screen was 60 μm, and the distance between the thread centroids was 254 μm. Therefore, the porosity was 58.3%. Compared to the nylon screen (mesh number 40 and porosity of 56%) tested by Fox et al. (2004) [21], the used nylon screen had more threads per inch and a little higher porosity. A 10 cm × 10 cm nylon screen was mounted in the middle of the test location, and its collection efficiency was evaluated.

The other collector was the aerosol droplet sampler (ADS, model 312, John W. Hock Company, Gainesville, FL, USA), which was designed to capture insecticide droplets using rotating slides, but we found that the original Teflon-coated slides were inefficient at analysing the deposition described (see Section 2.1.3) and thus used rotating nylon screens instead. Two nylon screens with dimensions of 50 mm × 100 mm were mounted at both ends of a rotating bar. The rotating bar had a length of 200 mm and was rotated at 344 rpm, creating a projected area of 0.02 m² (=100 mm × 200 mm) for airborne sampling. This active collector was expected to show higher collection efficiency than the former static sampler because fast-moving screens would have snatched droplets with even higher rotating speeds compared to the droplet velocity.

2.1.3. Deposit Analysis

It is essential to quantify captured droplets on collectors during spray-drift measurements. Most studies adopting passive collectors have used chemical tracers, especially metallic salts, fluorescent solutions or dye solutions [9,13,14,15,16,17,18,19,20,30,31,32,33,34,35,36,37], instead of pesticide solutions to avoid the risk of toxic concerns. Gil and Sinfort (2005) [38] stressed that spray-drift measurements with passive collectors can only be used by applying chemical tracers. However, farmers can disapprove the use of fluorescent or dye solutions because of crop discoloration or additional spraying responsibilities. To mitigate these issues, spray-drift measurements need to be carried out alongside farmers’ pesticide application. In this regard, the pesticide mass captured by collectors has been analysed by gas chromatography-mass spectrometry (GC-MS) [29] or inductively coupled plasma mass spectrometry (ICP-MS) [11], which quantify the mass of individual pesticide components. However, since these analysis methods are expensive, this study suggests a new method using a total organic carbon (TOC) analyser. This method quantifies the total mass of pesticide rather than the individual components as the TOC mass and is more cost-efficient than the former method. The greatest advantage of this method is that it can be used for quantifying both pesticide and other tracers and is relatively inexpensive.

After the spraying experiment, the pesticide spray deposited on the collector was extracted by washing out with 100 mL of deionised ultrapure water. After 10 min of horizontal shaking at 150 rpm, the washing water was sampled into a 40 mL vial for TOC analysis. The concentration of the TOC in the washing water dilution was measured in a TOC analyser (Multi N/C^® 3100, Analytik Jena, Jena, Germany) by the non-purgeable organic carbon (NPOC) method. The TOC concentration of the blank nylon screens was simultaneously measured and deducted from the calculation. The TOC concentration of the spraying liquid in the tank was also measured, and the amount of pesticide deposited on the collector was determined using Equations (1) and (2) by considering the tank concentrations:

$${\mathrm{TOC}}_{\mathrm{collector}}=\frac{\left({\mathrm{TOC}}_{\mathrm{sample}}-{\mathrm{TOC}}_{\mathrm{blank}}\right)\times {\mathrm{V}}_{\mathrm{sample}}}{{\mathrm{A}}_{\mathrm{collector}}}$$

(1)

$${\mathrm{D}}_{\mathrm{collector}}=\frac{{\mathrm{TOC}}_{\mathrm{collector}}}{{\mathrm{TOC}}_{\mathrm{tank}}}$$

(2)

where TOC_collector is the TOC concentration deposited on the collector (mg cm⁻²); TOC_sample and TOC_blank are the TOC concentrations of the washing water for the tested nylon screen and blank nylon screen, respectively (mg L⁻¹); V_sample is the volume of the test liquid, including the volume of the added ultrapure water for the wash out (L); A_collector is the projection area of the collector onto a plane perpendicular to wind direction (cm²); TOC_tank is the TOC concentration of the spraying liquid in the tank (mg mL⁻¹); and D_collector is the amount of pesticide solution deposited on the collector (mL cm⁻²).

To ensure the validity of using TOC for spray-drift measurement, some airborne samples were taken by nylon screens. The sampling time varied between nylon screens to make all screens had different pesticide deposits. Airborne samples were analysed by both the above method and through liquid chromatography with tandem mass spectrometry (LC-MS/MS, Exion LC™ and Triple Quad™ 5000, AB SCIEX, Framingham, MA, USA). The experimental details are described in Kim et al. (2020) [39]. LC-MS/MS is one of the most widely used techniques for accurately quantifying pesticide residue and was, therefore, used in this study for comparison.

2.1.4. Experiment for Measuring Collection Efficiency

The collection efficiency of the two airborne-drift collectors being assessed was measured in the drift tunnel based on various wind-speed conditions. Three wind speeds of 1, 2 and 3 m s⁻¹ were set up. A wind speed of 3 m s⁻¹ is typically the maximum permitted wind speed for spray application to prevent damage from spray drift [40].

After setting up the collector at the test location and adjusting the exhaust fan, the nozzle sprayed a pesticide solution for 10 s. Fan operation without spraying was continued for another 10 s to ensure that all the airborne droplets had escaped from the drift tunnel, following which the collector was picked up and analysed for any deposit. The pesticide product used was a suspension concentrate formulation (SC) (Farm Hannong Inc., Seoul, Republic of Korea) mainly used as an aerial and ground spraying formulation in Korea but without its active ingredient. The product was therefore specially made for the experiment by eliminating active substances. The spray liquid consisted of 250 mL of the pesticide product mixed in 4 L of water. The collection efficiency was calculated as the amount of pesticide solution deposited on the collector divided by the total airborne drift.

$$\mathsf{ϵ}=\frac{{\mathrm{D}}_{\mathrm{collector}}}{{\mathrm{D}}_{\mathrm{ta}}}$$

(3)

where ε is the collection efficiency of the collectors; and D_ta is the total airborne drift (mL cm⁻²).

The total airborne drift was measured by the three 350 mm × 350 mm recovery screens. The total airborne-drift measurements were performed before setting up the collectors at the test location.

All the tests were conducted in triplicate, and the average and standard deviation were calculated. The statistical analysis, the non-parametric Kruskal-Wallis test [41], was carried out to identify the variability of the measured collection efficiency under different wind speeds. Since the triplicate data of each group could not be assumed as a normal distribution, a non-parametric analysis of variance was used. A calculated p-value smaller than 0.05 indicated the collection efficiency would vary significantly by wind speeds.

2.1.5. Miscellaneous Measurements

In addition to measuring the collection efficiency, several measurements were taken to help understand the collection performance and prepare the input data for CFD simulations of the process.

The size distribution of the droplets sprayed from the nozzle was measured using a laser diffraction size analyser (Spraytec, Malvern, UK). The measurements were taken in triplicate for 5 s at two distances from the nozzle (300 mm and 500 mm). The size distribution of the airborne droplets drifting in the collecting section was also measured and compared to the size of the sprayed droplets. Since the laser diffraction size analyser could not be mounted in the collecting section, the measurement was carried out by following the method adopted by Zhu et al. (2011) [42] using water-sensitive papers and image analysis. The size distribution of the airborne droplets created by various wind speeds (1, 2 and 3 m s⁻¹) was analysed to identify the driftable size of droplets with respect to the wind speed.

The aerodynamic resistance of the nylon screens is a key factor for airflow modelling around the screen in CFD simulations. It was measured using the pressure-drop measurement system [43]. The viscous resistance coefficient (α) and pressure loss coefficient (C_ir) of the Darcy-Forchheimer equation (Equation (4)) were also measured.

$$\Delta {\mathrm{p}}_{\mathrm{s}}=-\left(\frac{\mathsf{\mu }}{\mathsf{\alpha }}\mathrm{u}+{\mathrm{C}}_{\mathrm{ir}}\frac{1}{2}{\mathsf{\rho }\mathrm{u}}^2 \right)$$

(4)

where Δp_s is the pressure loss caused by the screens (kg m⁻¹ s⁻²); μ is the fluid viscosity (kg m⁻¹ s⁻¹); ρ is the fluid density (kg m⁻³); u is the fluid velocity in the pipe (m s⁻¹); α is the viscous resistance coefficient or permeability of the screens (m⁻¹); and C_ir is the pressure loss coefficient or inertial resistance coefficient of the screens.

2.2. Computational Fluid Dynamics Simulations of the Collectors

2.2.1. Theoretical Basis for Droplet Collection

The collection performance of nylon screens can be modelled in two stages. The first stage is the impact process in the collecting section of the drift tunnel. While the wind flow decelerates due to the aerodynamic resistance offered by the nylon screen, some of the air streams diffuse and detour around the edges of the screen disturbing the impact of the airborne droplets on the screen. Therefore, the probability of droplet impact on the nylon screen is generally smaller than 1.

The second stage is the deposition process at the nylon screen. A droplet approaching the nylon screen is destined for one of three ends: (1) pass through the screen openings; (2) collect on the screen threads; or (3) split into several fractions, some of which collect on the threads. Droplet behaviour with screen threads is highly complicated and influenced mainly by the eccentricity between the droplet and the thread, the radii of both, the surface tension of the liquid and the hydrophilicity of the thread [44,45,46,47].

The maximum size for a droplet hanging on a circular thread was given by [45]:

$${\mathrm{R}}_{\mathrm{c}}=\sqrt[3]{3{\mathrm{b}\mathsf{\delta }}^2}$$

(5)

$$\mathsf{\delta }=\sqrt{\frac{\mathsf{\sigma }}{\mathsf{\rho }\mathrm{g}}}$$

(6)

where ${\mathrm{R}}_{\mathrm{c}}$ is the critical radius of the droplet; $\mathrm{b}$ is the fibre radius; $\mathsf{\delta }$ is the characteristic capillary length for which the surface tension and gravity balance each other; $\mathsf{\sigma }$ is the surface tension; $\mathsf{\rho }$ is the density of a droplet; and $\mathrm{g}$ is the gravitational acceleration.

The critical radius of the droplet was calculated to be 870 μm using Equations (5) and (6), and this meant almost all the droplets considered in this study could be captured on the threads.

The threshold velocity of capture for the off-centre impact of a droplet on fibre is a key variable in evaluating droplet capture on threads. Moreover, the threshold capture velocity is dependent on the surface wettability, droplet radius, eccentricity, fibre diameter and capillary length [46]. Safavi and Nourazar (2019) suggested various formulae according to the advancing and receding angles of droplet impact [46]. The authors showed that when the advancing angle (θ_a) for water on nylon fibres is approximately 71° [48], the threshold capture velocity for θ_a < 90° can be obtained by Equation (7) [46]:

$${\mathrm{u}}_{\mathrm{c}}=2\sqrt{\left(\frac{{\mathsf{\sigma }\mathrm{pR}}_{\mathrm{d}}}{\mathsf{\rho }}\cos {\mathsf{\theta }}_{\mathrm{r}}-{\mathrm{V}}_{\mathrm{d}}{\mathrm{R}}_{\mathrm{d}}\mathrm{g}\right)/\left({\mathrm{V}}_{\mathrm{d}}-\mathrm{AL}\right)}$$

(7)

where ${\mathrm{u}}_{\mathrm{c}}$ is the threshold capture velocity; $\mathrm{p}$ is the wetted perimeter of the thread; ${\mathrm{R}}_{\mathrm{d}}$ is the radius of the droplet; ${\mathsf{\theta }}_{\mathrm{r}}$ is the receding contact angle; ${\mathrm{V}}_{\mathrm{d}}$ is the droplet volume; $\mathrm{A}$ is the 2D projected contact area of the droplet on the thread; and $\mathrm{L}$ is the length of the contact line.

We assumed that if the droplet velocity was smaller than the threshold capture velocity, the droplet would be entirely captured on the thread; on the contrary, if the droplet velocity was higher than the threshold capture velocity, the droplet would be split into two fragments by the thread, with the smaller fragment representing the captured volume [47]:

$${\mathrm{V}}_{\mathrm{cap}}=\frac{\mathsf{\pi }}{3}\left({\mathrm{R}}_{\mathrm{d}}-\mathrm{e}\right)\left(2{\mathrm{R}}_{\mathrm{d}}^2+{\mathrm{R}}_{\mathrm{d}}\mathrm{e}-{\mathrm{e}}^2\right)$$

(8)

where $\mathrm{e}$ is the impact eccentricity; and ${\mathrm{V}}_{\mathrm{cap}}$ is the smaller fragment volume captured by the thread. For screens, a droplet can be captured by four threads of a square mesh simultaneously; thus, the captured volume was examined for four threads.

2.2.2. Computational Grid

Since the second stage needed micrometre-order precision, the two stages were modelled with different computational grids. The first stage was modelled as a cuboidal volume with dimensions of 1500 mm (X) × 350 mm (Y) × 350 mm (Z), the same as the collecting section of the drift tunnel (Figure 2). The nylon screen was located at X = 500 mm, allowing a large leeward space of 1000 mm behind it. Structured meshes were created with a minimum size of 3 mm near the nylon screen and an increasing ratio of 1.01 in the direction of +/−X. The number of cells was 2,809,856. For the second stage, the cuboidal volumes were modelled as both the windward and leeward sections of the nylon screen and also included an open hole and four surrounding half-diameter threads. The dimensions of the computational domain were 7.62 mm (X) × 254 μm (Y) × 254 μm (Z). Meshes were created with 469,349 tetrahedral cells and 12,352 triangular prism cells. The mesh size was approximately 5 μm near the screen threads and was increased to 20 μm at both ends in the X direction.

The mesh independence analysis was performed only for the first-stage model with three different grid resolution levels namely, coarse with 1,277,218 cells, medium with 2,809,856 cells and fine with 6462,126 cells. The minimum sizes of the cells near the nylon screen were 4.5, 3 and 1.5 mm, respectively. The air velocities at four distances behind the nylon screen (100, 200, 300 and 400 mm) were compared to each other.

2.2.3. Governing Equations and Boundary Conditions

The CFD models were developed to predict the airflows and droplet trajectory around the nylon screen. The airflows were solved using the Reynolds-averaged Navier-Stokes (RANS) equations with the SST k-ω turbulence model. The pressure-based solver was used to solve incompressible flows, and the SIMPLE algorithm was used to calculate the pressure-velocity coupling. The pressure, momentum, energy and turbulent quantities were spatially discretised using the second-order schemes. In the first stage, the aerodynamic resistance of the nylon screen was modelled by adding a source term for the pressure drop, which was calculated by the Darcy-Forchheimer equation and the measured parameters. The trajectory of the spray droplets was modelled through the Lagrangian discrete phase model. Moreover, the discrete random walk (DRW) model was used to consider the effects of turbulence on droplet movement. The details of the governing equations can be found in previous CFD work [49].

In the computational domains for two stages of simulation, the airflows were created along the X direction. The two ends in the X direction were set as the velocity-inlet and pressure-outlet. The other boundaries for the sides, top, bottom and threads were set as a no-slip wall. In the first-stage model, the air velocities at the inlet were set as 1, 2 and 3 m s⁻¹, which were identical to the experimental condition. The spray droplets were released at the inlet boundary, and their initial velocity was set to be identical to the air velocity.

However, since the speed of the air streams decreased at the nylon screen, the air velocity for the inlet of the second-stage model was set as the air velocity obtained at the nylon screen from the result of the first-stage model. The spray droplets were also released at the inlet boundary, and their initial velocity was set as the average droplet velocity obtained near the nylon screen from the first-stage model.

2.2.4. CFD Analysis to Determine Collection Efficiency

CFD simulations were performed using a commercial code ANSYS Fluent (version 2020 R1, ANSYS, Inc., Canonsburg, PA, USA) and an AMD Ryzen 3970X workstation (3.70 GHz, 32-core, 128 GB RAM). Droplet collection efficiency was predicted through two-stage simulations with respect to three wind speeds (1, 2 and 3 m s⁻¹) and 17 droplet diameters (ranging from 40 μm to 360 μm at intervals of 20 μm). In the first-stage simulations, the target wind speeds were set at the inlet, and 12,321 (111 × 111, uniformly distributed) airborne droplets were also released at the inlet. The air-stream velocity and the velocity of the droplets at the corresponding location of the nylon screen were recorded and then used as the inlet boundary conditions of the second-stage simulations. In the second-stage micro-scaled simulations, 255 droplets were released at the inlet.

From the two-stage simulations, the collection efficiency was calculated as follows:

$${\mathsf{\epsilon }}_{\mathrm{c}}={\mathrm{r}}_{\mathrm{I}}\times {\mathrm{r}}_{\mathrm{C}}$$

(9)

where ${\mathsf{\epsilon }}_{\mathrm{c}}$ is the collection efficiency; ${\mathrm{r}}_{\mathrm{I}}$ is the impact rate; and ${\mathrm{r}}_{\mathrm{C}}$ is the collection rate.

The impact rate, which was obtained from the first-stage simulations, is the ratio of the droplets that actually hit or pass through the nylon screen to the droplets heading toward the nylon screen. It is generally smaller than 1 because the air stream is diffused due to the aerodynamic resistance offered by the nylon screen. The collection rate, which was obtained from the second-stage simulations, is the ratio of the mass of liquid collected on the threads to the total mass of the 255 released droplets. The amount of collected liquid was determined by considering the threshold capture velocity of the droplets and the capturable portion.

3. Results

3.1. Method Validation

The airborne spray drifts quantified by the TOC showed good agreement with those measured by LC-MS/MS, with a R² of 0.9884 (Figure 3). Assuming that the LC-MS/MS analysis is one of the most accurate methods for the quantification of pesticides, the method proposed in this study can be proved both reasonable and feasible in terms of measuring the amount of spray drift collected on the nylon screen collector.

3.2. Airborne Spray Droplet

The size distribution of the droplets sprayed from the flat-fan nozzle, XR11002, at the working pressure of 0.3 MPa was measured. These measurement results are shown in

Table 1. Size distribution of droplets sprayed from the nozzle (measured by a laser-based analyser) and airborne droplets drifting in the collecting section (measured by water-sensitive papers).

Conditions	Dv10 1 (µm)	Dv50 1 (µm)	Dv90 1 (µm)
0.3 m distance from the nozzle	78	148	268
0.5 m distance from the nozzle	88	166	296
Airborne drifting in the collecting section (1 m s−1 wind speed)	102	154	206
Airborne drifting in the collecting section (2 m s−1 wind speed)	132	203	281
Airborne drifting in the collecting section (3 m s−1 wind speed)	152	229	303

¹ DvXX is the XXth percentile of the cumulative volume distribution.

. The volume median diameter (Dv50) was 148 μm at a distance of 0.3 m and increased to 166 μm at a distance of 0.5 m from the nozzle. Measured droplet size is known to vary according to the measurement distance from the nozzle and ambient air velocity [50]. The slight increase at 0.5 m distance was a result of spatial bias and can be probably attributed to the presence of ligaments in the spray cloud and deceleration and collision of droplets in the still air streams [50].

The Dv50 sizes of the driftable droplets by means of wind streams were 154, 203 and 229 μm for wind speeds of 1, 2 and 3 m s⁻¹, respectively. It was evident that higher wind speeds offered greater momentum to carry the droplets into the air. However, we expected that the size of the airborne droplets would be smaller than that of the sprayed droplets because some of the coarse droplets would have deposited onto the floor due to gravity. This was shown for the wind speeds of 1 and 2 m s⁻¹ because Dv90 decreased, whilst the measurements with higher wind speeds produced conflicting results. There could be three possible reasons for this. The first is that there might have been coalescence between droplets [51]. Fine droplets can easily change their path, especially due to turbulence, while there could be relative differences in droplet velocity depending on the droplet size. Therefore, the collision of small droplets with larger ones can occur more frequently with an increase in travel distance [52]. The second is the evaporation of small droplets. Fine spray can be easily vaporised, increasing the average size of the existing droplets. The third is that the water-sensitive papers might have had less collection efficiency for small droplets. The probability of the droplets impacting the surface increases with the Stokes’ number of the droplets, which is proportional to the square of the droplet diameter [53]. Therefore, small droplets have a lower probability of impacting a water-sensitive paper compared to larger droplets.

The amounts of the airborne droplets were 1.68, 3.17 and 5.40 mL for wind speeds of 1, 2 and 3 m s⁻¹, respectively. Considering that the amount of sprayed liquid was 116 mL, the ratios of the airborne droplets to the sprayed droplets were 1.44%, 2.73% and 4.65% for wind speeds of 1, 2 and 3 m s⁻¹, respectively. Certainly, with an increase in wind speed, more droplets could be blown into the air by the wind. The low conversion ratios were because the flat-fan nozzle was installed so that its fan-shaped spraying pattern would be perpendicular to the airflow direction. Thus, a large amount of the sprayed liquid was shot at the chamber walls to minimise the generation of airborne drift. Based on our experience, excessive generation of airborne droplets soaked the nylon screens quickly beyond their collecting capacity.

3.3. Collection Efficiency

The collection efficiencies of two airborne-drift collectors were calculated and are shown in

Table 2. Collection efficiencies of the two airborne drift collectors with respect to the wind speed.

Wind Speeds (m s−1)	Total Airborne Drift (μL cm−2)	Collector Deposition (μL cm−2)		Collection Efficiency (%)
		Nylon Screens	ADS	Nylon Screens	ADS
1	1.368	0.792 ± 0.086	0.918 ± 0.009	57.9 ± 6.3	67.1 ± 0.7
2	2.587	2.034 ± 0.194	1.289 ± 0.341	78.6 ± 7.5	49.8 ± 13.2
3	4.405	3.882 ± 0.175	1.034 ± 0.541	88.1 ± 4.0	23.5 ± 12.3

. The nylon screens and ADS captured 57.9–88.1% and 23.5–67.1% of the airborne droplets, respectively. The nylon screens showed improved performance with an increase in wind speed, while the collection efficiency of the ADS decreased gradually. The Kruskal–Wallis test produced p-values of 0.039 for the nylon screens and 0.027 for the ADS and revealed the statistically significant (p < 0.05) difference of the collection efficiency among different wind speed conditions.

Airborne-droplet collection on both the collectors was based on the impaction of the droplets onto the screens; thus, the relative velocity between the droplets and screens is a key factor influencing the collection efficiency. For the nylon screens, the passive collector, the only driving force creating this relative velocity was the wind; therefore, the collection efficiency increased with the wind speed. However, for the ADS, an active collector, the mounted screens were rotating at 344 rpm resulting in a maximum speed of 3.6 m s⁻¹ at the outermost edge of the screens. The active motion of the collector created appropriate relative velocities and high collection efficiencies under low-wind-speed conditions; however, with higher wind speeds, the rather high relative velocities might bounce off the droplets. Even some of the collected droplets can escape from the screens due to the centrifugal force. Therefore, the collection efficiency of the ADS decreased with the wind speed.

The collection efficiency of the nylon screens measured in this study was similar to that obtained by Fox et al. (2004) [21], which ranged from 50% to 70% under wind speeds ranging from 0 to 6 m s⁻¹. It is difficult to compare the results of this study with the literature because the nylon screens used were different. Even the droplets released from the nozzle in Fox et al. (2004) [21] were mostly smaller than 100 μm in diameter, but we used a commercial nozzle that produced a wider range of droplet sizes. However, a comparison with the literature demonstrates that the nylon screen used in this study can be practically available with a reliable collection efficiency.

3.4. Aerodynamic Analysis of the Nylon Screen Collectors

The comparison of air velocity between three grid resolution levels showed that the relative errors compared to the result of the fine grid condition were 0.53% and 0.11% for the coarse and medium conditions, respectively. The mesh sizes of the three grid conditions were fine enough to obtain reliable simulation results; however, the medium grid condition was chosen to ensure the grid independence of the computed results and reduce the number of computational cells.

The air velocity near the nylon screen and the droplet trajectory at various diameters were predicted through CFD simulations. In the first-stage simulations shown in Figure 4, the air velocity decreased slightly in front of the screen and significantly at their rear, resulting in flow speed-up near the edges of the screen. This meant that a part of the air supposed to reach the screen actually made a detour around the screen, entraining the droplets in the same way. As more droplets changed course to skirt the screen, the impact rate decreased likewise. Meanwhile, the drift paths of the droplets suspended in the air were determined by the airflow streams and the droplets’ inertia. As the droplet size and wind speed increased, the droplets had a higher inertial force. This meant that larger droplets with higher wind speeds were more likely to hit the screen rather than skirt the screen along with the detour airflow; however, smaller droplets with lower wind speeds were blown easily by the curved airflow. The results in

Table 3. CFD-computed impact rate, collection rate and collection efficiency according to wind speeds and droplet diameters.

Droplet Diameter (μm)	$\mathbf{Impact} \mathbf{Rate} \left({\mathbf{r}}_{\mathbf{I}}\right)$			$\mathbf{Collection} \mathbf{Rate} \left({\mathbf{r}}_{\mathbf{C}}\right)$			$\mathbf{Collection} \mathbf{Efficiency} \left({\mathsf{\epsilon }}_{\mathbf{c}}\right)$
	1 m s−1	2 m s−1	3 m s−1	1 m s−1	2 m s−1	3 m s−1	1 m s−1	2 m s−1	3 m s−1
40	0.717	0.826	0.865	0.424	0.459	0.451	0.304	0.379	0.390
60	0.770	0.862	0.927	0.435	0.439	0.459	0.335	0.379	0.425
80	0.821	0.893	0.958	0.510	0.553	0.514	0.419	0.494	0.492
100	0.830	0.925	0.978	0.612	0.635	0.559	0.508	0.588	0.547
120	0.867	0.943	0.994	0.706	0.694	0.568	0.612	0.655	0.565
140	0.899	0.958	1.000	0.788	0.692	0.607	0.708	0.662	0.607
160	0.922	0.969	1.000	0.863	0.711	0.662	0.795	0.689	0.662
180	0.942	0.975	1.000	0.925	0.757	0.745	0.872	0.738	0.745
200	0.949	0.980	1.000	0.953	0.800	0.800	0.904	0.784	0.800
220	0.955	0.986	1.000	0.969	0.866	0.869	0.925	0.854	0.869
240	0.964	0.989	1.000	0.977	0.917	0.920	0.942	0.907	0.920
260	0.969	0.992	1.000	0.993	0.960	0.960	0.962	0.952	0.960
280	0.982	1.000	1.000	1.000	0.994	0.994	0.982	0.994	0.994
300	0.991	1.000	1.000	1.000	1.000	1.000	0.991	1.000	1.000
320	0.999	1.000	1.000	1.000	1.000	1.000	0.999	1.000	1.000
340	1.000	1.000	1.000	1.000	1.000	1.000	1.000	1.000	1.000
360	1.000	1.000	1.000	1.000	1.000	1.000	1.000	1.000	1.000

indicate that the impact rate also increased with the droplet size and wind speed.

The trajectory of the droplets near the screen threads and droplets’ collection rate are presented in Figure 5 and Table 3, respectively. The collection of individual droplets on the threads was mainly influenced by the droplet’s eccentricity, radius and velocity. Due to its complexity, the effect of wind speed on the collection rate was not clear. For small droplets with diameters between 40 and 100 μm, wind speeds of 2 or 3 m s⁻¹ rendered higher collection rates, which, however, did not make a significant difference from a lower wind speed; in contrast, for larger droplets, the wind speed of 1 m s⁻¹ led to higher collection rates. For all the wind speeds, the collection rate increased with droplet diameter because there was a larger possibility for larger droplets to get physically trapped in the screen threads. Since the size of the screen openings was 194 μm, droplets greater than 200 μm in diameter were almost caught in the screen under the low wind speed of 1 m s⁻¹. For all wind speeds, all droplets greater than 300 μm in diameter were captured by the screen threads. As shown in Figure 5, while the air was passing through the screen, the wind speed increased between the threads. It also accelerated the droplets’ velocities and decreased their threshold capture velocity. That is the reason why the droplets with diameters between 200 and 300 μm were greater than the size of the opening and not totally caught in the threads.

The collection efficiency was finally determined by multiplying the impact rate with the collection rate. Table 3 presents the collection efficiency according to the wind speeds and droplet diameter, calculated through CFD simulations. The collection efficiency increased with the droplet diameter as in the impact rate and collection rate. The effect of wind speed on collection efficiency was positive for droplets smaller than 100 μm in diameter but negative for large droplets greater than 120 μm in diameter. This meant that small droplets were more likely to be caught in the screen under high wind speeds, while the collection of large droplets was slightly more conducive at low wind speeds. Very large droplets with diameters of 260 μm and above had high collection efficiencies close to 1.0 regardless of the wind speed.

To compare the CFD-computed collection efficiencies with the measured ones, the overall collection efficiencies were calculated based on the measured size distribution of the airborne droplets according to the wind speeds. Dv10, Dv50 and Dv90 of the airborne droplets (see Table 1) were converted into the upper-limit lognormal (ULLN) distribution [54] as shown in Figure 6. Subsequently, the whole-size distributions were multiplied by the collection efficiency of every droplet size to calculate the overall collection efficiency. The computed collection efficiencies of the nylon screen were 0.705, 0.785 and 0.856 for wind speeds of 1, 2 and 3 m s⁻¹, respectively. Compared to the measured collection efficiency (0.579, 0.786 and 0.881), the CFD predictions under wind speeds of 2 and 3 m s⁻¹ showed sound agreement with relative errors of −0.1% and −2.8%, respectively. At the low wind speed of 1 m s⁻¹, the difference got bigger with a relative error of 21.7%, indicating that the CFD predictions overestimated the collection efficiency. As mentioned earlier, the droplet size measured by water-sensitive papers can include experimental errors; therefore, the CFD analysis and measurements can reveal big differences in some cases. However, the CFD results demonstrated the effect of wind speed and droplet diameter on the collection efficiency of the nylon screens as an airborne-drift collector.

4. Conclusions

Accurate measurement of airborne spray drift is crucial to investigate spray-drift distances and identify the risk of pesticide drift. This study evaluated the collection efficiency of two airborne-spray-drift collectors using a small drift tunnel. The airborne spray drifts were quantified by a total organic carbon analyser and validated by comparison to measurements using liquid chromatography with tandem mass spectrometry. The collection efficiency of the nylon screen was 57.9–88.1% and increased with the wind speed. An ADS, an active sampler, captured 23.5–67.1% of the airborne droplets and showed reduced performance under higher wind speeds due to its rotary motion. Our measurements are in sound agreement with those obtained in the literature for nylon screens, which, despite their various types, can be considered for practical applications and show a reliable collection efficiency.

The collection performance of the nylon screen was also analysed through CFD simulations. The collection efficiency increased with the droplet diameter. Higher wind speeds improved the capture rate of small droplets (≤100 μm) but were not influential when it came to large droplets. The CFD-computed collection efficiency of the nylon screen was 70.5–85.6%, showing sound agreement with the measurements at wind speeds of 2 and 3 m s⁻¹. However, the CFD simulations overestimated the collection efficiency at the low wind speed of 1 m s⁻¹. Nevertheless, this study demonstrates the benefits of employing the CFD approach to understand the effects of droplet size and wind speed on the collection performance of nylon screens.