Click on author name to view affiliation information
The main objective of this study was to evaluate the suitability of the DEIAFA drift test bench system (Dipartimento di Economia e Ingegneria Agraria, Forestale e Ambientale; University of Torino, Italy) for assessing drift potential of a citrus (Citrus sp.) herbicide applicator. The study involved testing the effects of spray drift shield, nozzle type, and ground speed on drift potential of the applications. It was carried out in randomized block design within a split-split-plot experiment with five replications. A computational analysis procedure for evaluation of deposit values, measured along the test bench, was developed to compare the treatments in terms of a drift potential index (DPI). The methodology provided repeatable results. Among the treatments, ground speed was the main factor affecting the DPI. Both nozzle types tested [flat fan extended range nozzle (XR) and wide-angle deflector nozzle (TT)] showed higher DPI at faster speed. Decreasing the ground speed from 6.0 to 3.0 km·h−1 decreased the drift potential on average ≈35%. The performance of XR nozzle was improved by the presence of spray drift shield (27% reduction in DPI). However, the shield did not affect the drift potential of the TT nozzle significantly. The results were significantly affected by the wind velocity normalized by its direction relative to the sprayer travel; therefore, the tests should be carried out in relatively calm wind conditions, as much as possible.
Weed management in citrus orchards largely involves the application of nonselective herbicides (e.g., glyphosate and paraquat) for postemergence control (Singh et al., 2005). The off-target movement of these herbicides (drift) can affect citrus tree physiology with consequent impact on orchard profitability. It has been reported that the glyphosate contamination of sweet orange (Citrus sinensis) can result in peel burn and fruit drop (Erickson, 1996).
To minimize spray drift from herbicide applications, efforts have been made to develop less drift-prone equipment and application techniques. The most notable technology advances include: low-drift nozzles (Nuyttens et al., 2009), air-assisted boom sprayers (Pichè et al., 2000), drift reduction adjuvants (Johnson et al., 2006), and sprayer shield (Wolf et al., 1993). While the use of drift-reducing technologies has raised concern about potential reduction in herbicide application effectiveness (Knoche, 1994); several authors have reported no reduction in weed control efficacy of glyphosate applications with low-drift nozzles (Etheridge et al., 2001; Wolf, 2000).
Different approaches have been adopted to quantify the drift generated during pesticide applications. Salyani and Cromwell (1992) used high-volume air samplers and polyester film (Mylar®, DuPont, Wilmington, DE) to, respectively, quantify airborne and fallout spray drift from aerial and ground orchard sprayers. In another study, Salyani and Farooq (2004) used vertically hung polyester strings to sample spray drift from various citrus sprayers. Nevertheless, the most widespread method of drift assessment involves sampling of spray deposit at various downwind distances (Ganzelmeier et al., 1995; McArtney and Obermiller, 2008; Van de Zande et al., 2006). In herbicide applications, drift can be monitored by evaluating the injuries on sentinel plants (Felsot et al., 1996). Overall, the above techniques have shown a high variability of deposition in time and space because of the changes in weather conditions (Stover et al., 2003; Van de Zande et al., 2006). It has been difficult to compare different trials without directly comparable and repeatable conditions. Therefore, laboratory procedures have been developed to evaluate the potential drift of nozzles in wind tunnels (Derksen et al., 1999; Herbst 2001; Nuyttens et al., 2009). Wind tunnel may provide a method to classify drift potential of single or multiple nozzles without the effects of sprayer movement, crop characteristics, and environmental conditions. To simplify the test procedure and to reduce trial costs, Balsari et al., (2007) developed the DEIAFA test bench to evaluate drift potential of field crop sprayers. The methodology involves sampling of spray cloud that lingers on after the sprayer pass. Normally, the larger spray cloud mass and longer lingering time associate with greater drift potential under field conditions (Balsari et al., 2007). The main objective of this study was to evaluate the suitability of the DEIAFA drift test bench system in assessing drift potential of a citrus herbicide applicator. Specific objectives of the study were: 1) to develop a computational procedure for the evaluation of deposit values measured with the test bench and 2) to investigate the effects of drift shield, nozzle type, and ground speed on the drift potential of the sprayer.
The test was carried out in Lake Alfred, FL. It involved a tractor front-mounted off-center herbicide applicator (Chemical Containers, Lake Wales, FL) with a vertical lift for the adjustment of boom height (Futch and Salyani, 2005). The boom had a stainless steel leading edge and a 1-ft-wide flexible back curtain (drift shield) to provide housing for a series of recessed nozzles (Fig. 1). This sprayer is a relatively new design, which has been developed to reduce the drift and contact of herbicides to low-hanging citrus tree limbs. The boom was equipped with seven nozzles at 1-ft spacing and an off-center uncovered nozzle at the boom-end. The total boom length was ≈6.3 ft. The boom was operated at the height of ≈1 ft from deposit samplers (≈26 inches above the ground level) to prevent the curtain from touching the test bench.
Citation: HortTechnology hortte 21, 6; 10.21273/HORTTECH.21.6.745
Drift potential was assessed using a 10-m-long DEIAFA test bench (Salvarani S.r.l., Poviglio, RE, Italy) in accordance with Balsari et al. (2007). It consists of a 10 × 0.5-m modular frame with a sampler trays spaced at 0.5 m. The trays are covered by sheet metal lids that are attached to sliding rails. The methodology involved using a series of 15-cm diameter petri dishes (samplers) at 0.5-m intervals which were covered until being uncovered automatically by a pneumatic triggering mechanism activated by the boom movement. It took ≈0.3 s from the trigger actuation to uncovering the samplers. The trigger distance (1.5 m from the head of the test bench) remained the same for all treatments. A 1-m extension to the bench included two additional samplers that were kept uncovered at all times (UN00) to assess the actual application rate of the boom during each replication. The test bench was positioned on a relatively level ground covered by natural turfgrass (mowed). It was located at ≈1.9 m away from the centerline of the tractor and under the center of the spray boom. The nozzles were opened 10 m before the bench and stopped 10 m after. Spray solutions contained fluorescent dye (Pyranine-10G; Keystone Aniline, Chicago, IL) at ≈1000 mg·L−1. Earlier examination of this dye had shown negligible deposit degradation under short exposure to sunlight during spray tests (Salyani, 2000). Deposit samples were collected 2 min after stopping the sprayer to allow collecting potential droplet fallout.
Experimental factors included the presence/absence of the drift shield, two types of nozzles, and two ground speeds. The eight combinations of these factors (treatments) were applied as a randomized complete block design within a split-split-plot experiment (Table 1). Each treatment was replicated five times (blocks). All the treatments of a block were carried out in a single day. The shield was considered as the whole plot with the nozzle type as subplot and speed as sub-subplot factors. The two nozzle types tested were: 1) flat fan TeeJet® Extended Range XR8002VS [XR (Spraying Systems Co., Wheaton, IL)] and 2) Turbo TeeJet® deflector nozzle TT11002VP [TT (Spraying Systems Co.)]. They were operated at 40 psi pressure throughout the experiment. Those nozzles generated different droplet size spectra, as reported in the manufacturer's catalogue. At the tested pressure, the XR and TT are characterized by fine and coarse droplet size spectra, respectively, according to the British Crop Protection Council classification (Southcombe et al., 1997). The off-center nozzle [1.51 L·min−1 at 40 psi, TeeJet OC-04-AL (Spraying Systems Co.)] on the boom end remained constant for all the treatments. Nominal ground speeds were 3.0 km·h−1 (slow) and 6.0 km·h−1 (fast); however, the actual speed of each pass was recorded by timing with a stopwatch between two markers 31 m apart. Before starting the field test, the sprayer was calibrated and flow rates of different nozzles were measured. For simplicity, hereafter the treatments are identified with combinations of Shield_Nozzle_Speed (e.g., No-Shield_XR_Slow).
During the test, weather data including temperature, relative humidity, wind velocity, and wind direction were recorded at frequency of 1 Hz (Table 2). The weather station (Campbell Scientific, Logan, UT) was located close to test bench position, ≈15 m southwest of the application line. It consisted of a data logger (CR-10 WP), dry/wet bulb thermocouples (CR-10 TCR) located at 1.7-m height, and a wind sentry set (03101–5R.M Young anemometer and vane) at 3.2 m height above the ground level. Test runs were applied early morning (0600 to 0730 HR) when atmospheric conditions were near neutral (slightly unstable to slightly stable). Atmospheric stability, expressed by the stability ratio (Munn, 1966), was calculated using the meteorological data (15-min averages) obtained from the nearby Florida Agricultural Weather Network meteorological station. The meteorological data were used as covariance factors in single way analysis of variance (ANOVA) of DPI values and in the analysis of UN00 samplers deposits. In addition, wind velocity normalized by wind direction (Wnor) was also used to more correctly investigate the effect of wind on the test bench results. The parameter Wnor was calculated with the following formula:
where, Wdir is the wind direction (degrees) relative to the sprayer travel (headwind positive) and Wvel is the wind velocity (meters per second).
The temperature varied between 16 and 26 °C while relative humidity was above 82% (for 27 of the 40 runs, it was >95%). High relative humidity involved the condensation of water on test bench, which could cause the contamination of samples. Thus before each spray, both top and bottom surfaces of the test bench sliding cover were wiped with paper towel. Although the measured weather data showed relatively calm wind conditions, there were some movements in the ambient air as indicated by variable wind directions.
The petri dishes containing spray deposits were stored in a refrigerator (at ≈4 °C) until analysis (up to 48 h). Tracer dye was analyzed by adding 20–140 mL of deionized water into each petri dish, shaking the samples for ≈2 min on a shaker platform at 10–40 rpm, taking two 5-mL cuvette samples, and measuring the solution fluorescence with a fluorometer (AU-10; Turner Designs, Sunnyvale, CA). The fluorometer primary (excitation) and secondary (emission) filters had wavelengths of 340–500 nm and <520 nm, respectively. The range of wash volume accommodated the readable range of the fluorometer (0.001–2 ppm v/v). Some samplers containing higher deposits as well as the ones that were left uncovered before sprayer pass (UN00) needed further dilution (five times).
Deposition values of each spray run, measured along the test bench, were normalized to the intended nominal ground speeds to remove the variability in actual measured speeds. After that, they were reported as percentage of the nominal spray discharged from the sprayer [fraction of theoretical deposition (FTD)]. These values were transformed into natural logarithm [LN(FTD × 10,000)] to increase the weight of deposits collected at the tail of the test bench.
Two approaches were used to compute the DPI for the treatments. In the first method, deposits were evaluated based on the distance from the first fully uncovered sampler after trigger (FU). This method is based on Balsari et al., (2007) approach but with log transformation of deposit data. The FU point was determined considering the distance traveled by the boom sprayer after the actuation of the trigger system. Considering the uncovering system delay of 0.3 s, the FU points were 0.25 and 0.5 m beyond the trigger position for 3.0 and 6.0 km·h−1 speeds, respectively. The DPI was calculated as the sum of log-transformed percent deposit values measured along the test bench (DPId), from the FU point for a total length of 9 m, as described by the following formula:
where, d is the distance between two consecutive samplers.
In the second computational approach, the DPI was assessed in terms of the elapsed time for uncovering a sampler after passing of the sprayer over its position (DPIt). The DPIt was calculated as the sum of log-transformed percent deposits for a total interval of 5.4 s from the FU point. The 5.4 s value was the time required by the sprayer to cover the distance between the test bench end and the FU position at 6.0 km·h−1. This time interval corresponds to a length of 9 m at 6.0 km·h−1 and 4.5 m at 3.0 km·h−1. The following is the formula used for DPIt calculation:
where, t is the time (seconds) required to cover the distance between two samplers at the given ground speed.
The DPI values for different treatments were analyzed with the ANOVA procedure. Mean separation of the DPI data were carried out with the Tukey's honestly significant difference multiple comparison test. All the statistical analyses were evaluated at the 5% significance level. Statistical procedures were carried out with SPSS® (version 17.0; IBM, Somers, NY).
The percentage of tracer dye sampled on uncovered petri dishes (UN00) ranged from 85.9% to 96.8% of the actual applied volume rate. These base deposits were unaffected significantly by the experimental treatments. Nevertheless, the covariance analysis of the deposit values revealed significant effects of air temperature and relative humidity on UN00 values. At higher temperature and lower humidity conditions, the amount of spray deposited on uncovered samplers decreased.
Illustration of deposit values detected along the test bench (Fig. 2A) features two subgroups characterized by two ground speeds. At 6.0 km·h−1 (solid lines), the uncovering delay effect resulted in fewer samplers exposed to the spray after the trigger actuation than those at 3.0 km·h−1 (dashed lines). This effect could be misunderstood with an earlier decrease of deposits at higher ground speed, which in turn characterizes less drift-prone sprays. As it is shown in Fig. 2B, shifting the deposit curves for the distance traveled by the sprayer during the samplers uncovering actuation delay brings all the curves to the same starting point. Since the principle of drift potential study with the DEIAFA test bench involves sampling the spray cloud that lingers on after the spray pass, the results were also reported as a function of the elapsed time from the FU point (DPIt). The DPIt limited the calculation interval to the elapsed time for traversing the length of the test bench at 6.0 km·h−1 and a subsequent reduction of the sampling line considered for the assessment of drift potential at lower speed (4.5 m) (Fig. 2C).
Citation: HortTechnology hortte 21, 6; 10.21273/HORTTECH.21.6.745
The average coefficients of variation (CV) of DPId and DPIt values of all the replications of each treatment were 23.1% and 18.0%, respectively.
Analysis of variance of both indices showed significant effects of nozzle type and ground speed on drift potential of the boom sprayer (Table 3). However, at the tested pressure of 40 psi, the effect of shield was not significant, although with the DPIt procedure there was a significant interaction between nozzle and shield. A separate ANOVA of each nozzle type showed that the shield significantly reduced the drift potential (DPId or DPIt) of XR nozzle but not that of TT nozzle. The difference in the effect of shield may be attributed to the difference in design of nozzles and their droplet size spectra.
Since the block factor did not have any significant effect on DPI, the one-way ANOVA was used to analyze mean separation of the treatments (Table 4). The mean separation of DPId and DPIt values resulted in comparable treatment discriminations for the two computational procedures adopted. The worst-case scenario in terms of drift potential was represented by the extended range nozzle used at higher speed without the spray shield, while the best case obtained with the shielded TT nozzle at lower speed. The latter showed ≈56% or 67% reduction in the DPIt or DPId, respectively (Table 4). Decreasing the ground speed from 6.0 to 3.0 km·h−1, the DPIt and DPId decreased on average about 35% and 49%, respectively (Table 5). Considering that the boom height was maintained at 1 ft above the target, this result is in accordance with Miller and Smith (1997), who reported a drift increase of ≈51% when the ground speed increased from 4.0 to 8.0 km·h−1 for a boom sprayer operating at 0.8 m from a grass stubble surface. At the tested pressure of 40 psi, the DPIt and DPId of XR nozzle was ≈18% and 22% (on average) higher than that of the TT nozzle. The shield affected mainly the extended range nozzle with about 27% reduction in DPIt, while the TT nozzle drift potential decreased only for about 10% of its unshielded value.
Although individual weather parameters did not affect the DPI values, the combined effect of wind direction and wind velocity identified as normalized wind (Wnor) significantly affected the results (Table 6). This means that the influence of the wind is associated with its intensity and direction. The headwind carries the droplets downwind on the test bench, increasing the amount of deposits at farther locations, and it results in higher DPI value. On the other hand, the backwind pushes the droplet cloud forward and results in a lower DPI. Therefore, to reduce the effect of wind conditions in the assessment of DPI with the DEIAFA test bench, the tests should be carried out in relatively calm wind conditions, as much as possible. Weather stability index did not have a significant effect on DPI; nevertheless, a meteorological station with higher sampling frequency may give different result.
The DEIAFA drift test bench was useful in assessing drift potential of a citrus herbicide applicator when used in different operating configurations. Both DPI indices were practical in estimating the relative drift potential reduction; nonetheless, the DPIt was more functional when speed was a factor. Formulating a DPI in elapsed time basis allowed determining the effect of ground speed and its interaction with the flat fan XR and the wide-angle deflector TT nozzles when used with or without the spray drift shield. The methodology allowed the discrimination between different herbicide application techniques in terms of their drift potential. At the tested pressure (40 psi), a 56% reduction of the DPIt was observed between the worst (XR nozzle at 6.0 km·h−1 without spray shield) and the best (TT nozzle at 3.0 km·h−1 with spray shield) treatments. Ground speed was the main factor to affect the DPI. Both nozzle types showed higher DPI at faster speed. Decreasing the ground speed from 6.0 to 3.0 km·h−1 decreased the drift potential about 35% and 49% using DPIt and DPId, respectively. Flat fan XR nozzle showed higher drift potential compared with the TT nozzle, but its performance was improved by the presence of spray drift shield. However, the shield did not affect the drift potential of the TT nozzle noticeably. The results of DPIt procedure showed a good level of repeatability (CV = 18%) for an open field test; while it increased to 23% with the DPId computation. Longer exposure of droplet cloud to weather conditions resulted in higher variability of deposit samples. This was the case for samplers located at farther distances from the triggering point and at lower ground speed. The method needs relatively stable atmospheric conditions, as the combined effects of wind velocity and direction, expressed collectively by the normalized wind factor (Wnor), showed a significant effect on drift potential of the sprayer.
Contributor Notes
Mention of a trademark or vendor name does not imply an endorsement or recommendation by the University of Florida or University of Torino.
Corresponding author. E-mail: msi@ufl.edu.
