Intelligent, Variable-rate Spray Technology Reduces Total Pesticide Output while Controlling Foliar Disease of Shumard Oak¹

Variable-rate spray technology has repeatedly shown the potential to significantly reduce total pesticide output and off-target waste when retrofitted onto sprayers already being used at commercial nurseries. With this technology now commercially available (Smart Apply Inc., Indianapolis, IN), this reduction would not only decrease annual input costs for growers but also decrease environmental and health impacts. However, variable-rate spray technology must also sufficiently control pests in a range of nursery production systems for it to appeal to growers and for large-scale adoption to occur. This research shows that variable-rate spray technology utilized in a dense production system, i.e., a multi-row block of pot-in-pot tree production, can control diseases equally or better than conventional, constant-rate spray applications giving growers confidence to reap the benefits of adopting this technology. Additionally, this study documented substantial off-target spray losses to the nursery floor with both intelligent, variable-rate and conventional applications. Growers with variable-rate spray technology can reduce off-target ground applications and concomitant input cost by adjusting their sprayer detection to a crop height commensurate with the canopy height as opposed to the 0.1 m (0.3 ft) used in this study when making applications to control canopy pests. Prior to spraying, producers without variable-rate technology can manually close lower nozzles to achieve a similar effect. An additional benefit to reducing off-target ground applications is minimizing the opportunity for non-target organisms, including beneficial fungi, e.g., Trichoderma spp., and beneficial insects from being unnecessarily exposed to pesticides, which may contribute to improved soil, and overall ecosystem, health.

Pesticides are an essential part of pest management and prevent pest damage to $40 billion worth of crops in the United States (US) each year (Pimentel and Burgess 2014). While crucial to maintaining saleable crops, pesticides can have detrimental effects on natural ecosystems, beneficial insects, and human health, especially pesticides that reach unintended targets (Kim et al. 2017). Each year, an estimated $9 billion is spent to combat human health and environmental impacts of pesticide use (Koleva et al. 2011). These harmful outcomes, which include an estimated 385 million poisonings and 11,000 deaths per year (Boedeker et al. 2020) as well as surface and groundwater contamination (da Silva Sousa et al. 2021), combined with the rising costs of pesticides (Elkin 2022), provide an incentive for growers to reduce their pesticide output.

Many nurseries utilize air-blast sprayers to apply pesticides because of their relatively easy maintenance and ability to spray a wide range of crop sizes. However, the inefficiency and poor uniformity of coverage from air-blast sprayers in nursery settings is well documented. For example, in a study of crabapples (Malus spp.) sprayed two rows at a time, less than 30% of the total spray volume applied by an air-blast sprayer landed on the target crop and over a third was lost to the ground (Zhu et al. 2008). In addition, significant drift from air-assisted sprayers has been found beyond 46 m (151 ft) away from the intended target crops (Grella et al. 2017). Depending on nozzle type, drift can be detected as much as 61 m downwind (Zhu et al. 2006). Applications that fail to reach or overshoot the target crop can lead to contamination of water, air, and soil (Kim et al. 2017). This off-target movement can contribute to a decrease in beneficial insects and biodiversity as well as resistance to pesticides (Koleva et al. 2011).

Several characteristics of the nursery industry make increasing pesticide application efficiency uniquely challenging. In the US, over 2,000 different nursery species are produced (Yeager et al. 2013), which means that applications must be applied to plants in a variety of shapes, sizes, and canopy densities. Additionally, to increase production without increasing land area, many growers use multi-row blocks with off-set rows (LeBude et al. 2006). Conventional air-blast sprayers are not tailored to small nursery crops planted in this arrangement (Chen et al. 2013a), nor do producers typically make manual adjustments when spraying to account for blocks with smaller and larger-sized crop plants. To ensure adequate spray application within multi-row blocks, growers often spray to “runoff” (Walklate et al. 2006), using a rate such as 1,870 L·ha⁻¹ (200 GPA) even though this can result in excessive application in target trees (Zhu et al. 2008). Collectively, these practices highlight the challenges to making meaningful improvements that reduce off-target movement of pesticides during foliar applications.

A number of technologies have been developed in an attempt to increase air-blast sprayer application efficiencies. These include the use of ultrasonic sensors and cameras to detect targets (Hočevar et al. 2010, Stajnko et al. 2012), variable air assistance (Hołownicki et al. 2017), and drift reduction nozzles (Hoffmann et al. 2010). In the last decade, intelligent, variable-rate spray technology has been developed that uses a laser to detect target crops and pulse-width modulated (PWM) nozzles to provide customized applications (Chen et al. 2012). This technology has consistently demonstrated an increase in efficiency and decrease in overall spray output. It reduced spray loss to the ground in an experimental orchard by as much as 93% and reduced aerial off-target losses by up to 80% (Chen et al. 2013a, Chen et al. 2013b). When operated in variable-rate mode to spray small trees, this technology consumed as little as 12.1% of the spray volume compared to the constant-rate mode (Shen et al. 2017), yet the spray discharged provided adequate deposition even within multi-row blocks (Zhu et al. 2017a). This technology also proved that it could provide comparable pest management in commercial settings when compared to a conventional, constant-rate pesticide application (Fessler et al. 2021, Zhu et al. 2017b). To further the adoptability of this technology, it was adapted for maximum compatibility (Warner et al. 2022) across a range of models and manufacturers so that it could be retrofitted onto growers’ existing sprayers. These retrofitted sprayers were also able to reduce off-target spray and total spray volume using variable-rate mode. The technology was estimated to reduce spray volume by 59% to 83% compared to the grower’s standard practice at a commercial apple orchard (Fessler et al. 2020) and by 50% at a multi-row nursery (Chen et al. 2019). Off-target losses were reduced using variable-rate mode in an apple orchard and vineyard by up to 40% and 33%, respectively (Nackley et al. 2021).

In order for widespread adoption of this now commercially available technology (Smart Apply Inc, Indianapolis, IN) to occur in the nursery industry, compatibility and relative advantage must be demonstrated (Warner et al. 2022) beyond a reduction in spray volume, i.e., it must be shown that this technology can effectively control pests in a multi-row nursery utilizing growers’ existing spray equipment. Oaks are commonly produced in nurseries and are prone to leaf spot diseases (Dirr 1998). Two diseases known to occur in association on oaks, cylindrosporium leaf spot (Cylindrosporium spp.) (CLS) and Tubakia leaf spot (Tubakia dryina) (TLS) (Munkvold and Neely 1989), were selected for this study. The objectives of this experiment were to evaluate the effectiveness of this technology in a commercial pot-in-pot nursery setting, compare spray volume and characteristics of conventional, constant-rate and intelligent, variable-rate applications, and assess disease control in the two sprayer modes.

For this experiment, an air-assisted trailer sprayer (Storm 2000, Tifone, Porotto, Italy) was retrofitted with a laser-guided, variable-rate control system (Fig. 1). The original sprayer consisted of a 2,000-L (528-gal) spray tank, fan with 80-cm (31-in) propeller, 20 hollow-cone nozzles (10 on each side) (D6-DC25, TeeJet, Glendale Heights, IL), pressure regulator, and continuous agitation. The retrofit kit of the intelligent spray control system included an embedded computer with touch screen, a switch box to control the mode of operation, a high-speed laser scanning sensor (UTM-30LX, Hokuyo Automatic Co., Ltd., Japan), a non-contact Doppler-radar ground travel speed sensor (RVSIII radar velocity sensor, Dickey-John Corp., Auburn, IL), an automatic flow control box, and PWM solenoid valves (115880-1-12, TeeJet, Glendale Heights, IL). The system manipulated the spray output of each nozzle by using PWM solenoid valves to achieve variable rate applications. The output was based on the set spray rate (a designated spray volume per volume of crop), tree characteristics (presence/absence, height, width, density), and tractor speed. The spray rate is selected by the applicator using the touch screen computer. More details on the intelligent sprayer system are described in Shen et al. (2017).

During this trial, the tractor was operated using the grower’s standard practices; it was driven at an average speed of 4.3 km·hr⁻¹ (2.7 mph) with the power take off at 540 RPMs and the nozzles pressurized to 689 kPa (100 psi). The total spray output ranged from 0 to 52 L·min⁻¹ (0 to 14 gal·min⁻¹) for the variable-rate mode based on the calculated spray output, with the maximum flow rate for each nozzle being 2.6 L·min⁻¹ (0.69 gal·min⁻¹). The default spray rate of 0.07 L·m⁻³ (0.07 oz·ft⁻³) (the amount of spray solution applied per the volume of crop detected) was used during this experiment while the sprayer was operated in the variable-rate mode. Additional spray parameters were set as follows: spray width (left and right) = 15.2 m (50 ft), vertical maximum = 20 m (65.6 ft), vertical minimum = 0.1 m (0.3 ft), horizontal maximum = 20 m (65.6 ft), horizontal minimum = 0.5 m (1.6 ft). This was compared to operating the sprayer in constant-rate mode which simulates the output of the sprayer without the retrofitted system and consistently applied 52 L·min⁻¹ (14 gal·min⁻¹).

The test plot consisted of 10 production blocks of trees in 57-L (#15) containers in a pot-in-pot production system at Hale & Hines Nursery, Inc. (McMinnville, TN). Between blocks were alternating narrow [2.4 m (7.9 ft)] and wide [3.4 m (11.2 ft)] driveways (Fig. 2). Trees were sprayed from the wide driveways only. Each block had 6 rows of trees, with rows being 1.2 m (3.9 ft) apart and trees being 1.2 m (3.9 ft) apart within the row. Block length varied but was approximately 220 m (656 ft). The plot was divided into 2 approximate halves, one for each sprayer mode. In 2018, the constant-rate mode was used in the western half and the variable-rate mode was used in the eastern half. In 2019, this was reversed with constant-rate used in the eastern half and variable-rate in the western half.

Ten Shumard oaks (Quercus shumardii Buckley) from each treatment were used to assess pest control and spray characteristics. From each treatment, five trees in the western exterior row of the block (row 1) adjacent to a wide driveway and five trees in an interior row (row 3) two rows in from the western exterior row were selected to be scouted for pests and used to evaluate target spray characteristics. Increase in tree caliper was calculated from caliper measurements taken at the beginning and end of each season.

Throughout the growing seasons in 2018 and 2019, approximately every three weeks, one half of the plot was sprayed in the intelligent, variable-rate mode and the other half was sprayed in the manual, constant-rate mode. Trees were sprayed with a rotation of insecticides (i.e., B-cyfluthrin, imidacloprid, bifenthrin, and permethrin) and fungicides (i.e., propiconazole, azoxystrobin, and thiophanate-methyl).

In 2018 and 2019, trees were scouted for pests and beneficial insects a total of nine times. In 2018, CLS and TLS were both present. In 2019, only CLS was observed. Both the incidence and severity of pests were recorded on a whole tree and individual branch basis. Three representative branches at approximately 1.5 m (5 ft) from ground level were selected and scouted for each tree. Incidence was measured the same way for both the whole tree and the individual branches as “Yes” or “No” as to whether evidence of the pest was present. Severity for the whole tree was measured on a scale of 0-4 that indicated the percent of the tree affected by the pest with 0 = 0%, 1 = 1-25%, 2 = 26-50%, 3 = 51-75%, and 4 = 76-100% similar to Hagan et al. (1998). Severity of individual branches was measured as the percent (0-100%) of the branch that was affected by the pest. Pest identification was verified by sending foliar samples to the University of Tennessee Soil, Plant, and Pest Center (Nashville, TN). Two categories of beneficial insects were monitored and recorded: natural enemies and pollinators. Natural enemies were recorded if they were present on the same branches that were scouted for pests. Likewise, insects that are known to serve as pollinators for any species were noted if observed on the same branches that were scouted for pests, in addition to the vegetation at the base of each scouted tree, even if they were not necessarily actively foraging. However, vegetation on the nursery floor was generally non-existent, and if present, sparse and consisted of weedy species.

Water sensitive cards [WSCs (Syngenta Crop Protection AG, Basel, Switzerland)], a common method of assessing foliar spray characteristics (Özlüoymak and Bolat 2020, Zhu et al., 2011), were utilized to determine spray characteristics at two time periods in 2019 and in 2020, early season (7 May and 2 June, respectively) and late season (12 August and 9 September, respectively). WSCs change color, turning from yellow to dark blue, after coming in contact with water droplets. WSCs were placed in canopies of five trees each in an exterior row and an interior row and on the ground below each to assess target and off-target spray deposits (Fig. 3). A single electrical clip was placed on a lateral branch on the south side of each selected tree. The clips were set at a height between 1.4 and 1.8 m (4.6 and 5.9 ft) and were between 5.1 and 15.2 cm (2.0 and 6.0 in) from the branch tip. In the target trees, a pair of WSCs were placed back-to-back in the clip facing east and west (i.e., perpendicular to the spray). Additionally, in the block just east of the selected trees, five trees in the western exterior row were selected to characterize spray drift. In the drift trees, the clip was in the same position but only held a single WSC that faced west. At the base of target trees, a board was placed on the ground on the west side of the pot. A WSC was attached to this board and was used to assess non-target ground applications.

The sprayer was driven down the driveway adjacent to the selected exterior trees. The WSCs placed in non-target trees to capture spray drift were then collected in labeled envelopes and stored with a desiccant. Next, the sprayer was driven down two additional driveways and the remaining WSCs were collected in the same manner as the drift WSCs. This process was done first in the intelligent, variable-rate mode and then in the manual, constant-rate mode. Following sprays with WSCs, the two halves were reciprocally sprayed in the non-assigned sprayer mode as crop volume could not be assumed to be equal in the two blocks, and a correction factor was created to adjust spray volume output to take into account the difference in crop canopy between the two halves as crop varieties outside of the selected trees varied. The volume output by the sprayer was documented. Average and maximum wind speeds, temperature, and relative humidity were recorded (Table 1). Weather data was collected using a handheld anemometer (Kestrel 3000, Nielsen-Kellerman Company, Boothwyn, PA). On spray dates with WSC, the percent full sun that penetrated through the canopy was also recorded as a measure of canopy density for each of the 10 selected trees per treatment by using a line quantum sensor (LQS706; Apogee Instruments, Logan, UT) connected to a quantum meter (QMSS; Apogee Instruments). Two parallel measurements were made: one in full sun and a second centered in the shadow cast by the canopy. The percent of full sun that was able to penetrate the canopy was then calculated by dividing the under canopy value by the full sun value.

After returning to the lab, the WSCs were scanned at 600 dpi (HP Photosmart Plus All-in-One Printer - B209, Hewlett-Packard, Palo Alto, CA) and saved as jpg files. These files were then analyzed using the DepositScan program (Zhu et al. 2011) for spray coverage (%), deposit density (droplets/cm²), and deposits (μL·cm⁻²).

Tree measurement data (increase in caliper and tree height), percent full sun data, scouting branch data (binary pest presence, branch, binary pollinators, severity) scouting whole tree data (pest presence, severity), and card data (coverage, deposits, deposition) were analyzed using generalized mixed models analysis. Ranked transformation was applied for numeric continuous outcomes when diagnostic analysis on residuals exhibiting violation of normality and equal variance assumptions using Shapiro–Wilk test and Levene’s test. Post hoc multiple comparisons were performed with Tukey’s adjustment. Statistical significance was identified at the level of 0.05. Data were presented as mean ± standard error. Analyses were conducted in SAS 9.4 TS1M7 (SAS institute Inc., Cary, NC).

Because height was influenced by pruning, differences in height between treatments or over time were not considered. On average, trees were 2.45 and 2.16 m (8.04 and 7.09 ft) tall in 2018 and 2019, respectively. There was no difference in the increase in caliper between the variable-rate [9.25±1.80 mm (0.36±0.07 in)] and constant-rate treatments [7.62±1.61 mm (0.30±0.06 in)] (P = 0.22), indicating that tree growth was not affected by sprayer mode.

The average percent of full sun that penetrated the canopy in early season and late season for the variable-rate sprayer mode was 34%±5% and 24%±2%, respectively, while in constant-rate sprayer mode the average percent of full sun in early season and late season was 31%±4% and 23%±2%, respectively. Between the two sprayer modes, there was no difference in percent full sun on selected trees (P = 0.8472). In addition, there was no effect of timing on percent full sun (P = 0.8176) nor a significant interaction. Percent full sun can be an indicator of foliar density (i.e., the sparser the foliage, the greater percentage of full sun that can penetrate the canopy). Since there were no differences between sprayer modes, likely they were subject to the same density of foliage and therefore had the same opportunity to penetrate the canopy and coat the foliage. Because there was no effect of timing, the increase in foliar density between early season and late season sprays was not significant.

Operating in variable-rate mode released an average of 247.3±9.5 L·ha⁻¹ (65.3±2.5 GPA), nearly half the pesticide volume compared to the constant-rate mode which released an average of 432.4±16.6 L·ha⁻¹ (114.2±4.4 GPA) (P <0.0001). This substantial reduction in pesticide volume occurred despite the use of dense, multi-row production blocks with few gaps between plants. Similarly, Chen et al. (2019) found that intelligent, variable-rate spray technology reduced pesticide use by over 50% at two ornamental nurseries utilizing multi-row blocks in either field or pot-in-pot production. In addition, intelligent, variable-rate spray application consistently reduced pesticide and foliar fertilizer applications over a three-year period at two nurseries when compared to a conventional spray (Chen et al. 2021). The prototype version of this technology also reduced pesticide volume in a multi-row field production nursery by 54% (Fessler et al. 2021). In a study looking at multiple ornamental nurseries, Zhu et al. (2017b) found that using the variable-rate mode could save growers as much as $519 per hectare ($210 per acre) each season. Implementation of this technology would equate to a significant annual reduction in input costs for nursery producers.

For both whole tree and branch incidence of CLS, the constant-rate treatment was more likely to have CLS present than the variable-rate treatment (P <0.0001). For branch incidence of CLS, there was also an interaction between sprayer mode and row position (P = 0.0091). Within the constant-rate treatment, there was greater CLS incidence in the interior rows than the exterior rows with the interior row being 1.4 times as likely to have CLS present than the exterior rows. Similarly, higher disease ratings were found in the interior rows of multi-row blocks of flowering dogwood (Cornus florida L.) at another commercial nursery (Fessler et al. 2021). Within the variable-rate treatment there was no difference in CLS branch incidence between the interior and exterior rows. For both the interior and exterior rows, there was greater incidence of CLS in the constant-rate treatment than the variable-rate treatment with the likelihood to observe CLS nearly 6 times greater in the constant-rate treatment than the variable-rate treatment. Other studies have also reported instances in which the intelligent, variable-rate treatment had better pest control than the conventional, constant-rate treatment (Chen et al. 2019, 2021).

Sprayer mode affected whole tree severity of CLS; the variable-rate treatment had a lower CLS disease rating on average (0.38±0.05) than the constant-rate treatment (0.7±0.06) (P <0.0001, Fig. 4a). For branch severity, there was an interaction between sprayer mode and row position (P <0.0001, Fig. 4b). Within the variable-rate treatment, CLS disease ratings were greater in trees in the exterior row (3.6±0.7) than those in the interior row (0.9±0.2); however, in the constant-rate treatment, CLS disease ratings were greater in the interior rows (6.7±0.05) than the exterior rows (6.1±0.04). For both the exterior and the interior rows, the constant-rate treatment had greater CLS disease ratings than the variable-rate treatment. While statistically significant, given the low severity and small differences (i.e. less than 3% and 1% for the constant- and variable-rates, respectively), these results may have limited practical importance.

Whole tree incidence of TLS data did not converge and therefore could not be analyzed. Sprayer mode, row position, as well as sprayer mode and row position interaction did not affect branch incidence of TLS (P >0.05). For TLS whole tree severity (Fig. 4c) and branch severity (Fig. 4d), there was no effect of sprayer mode, row position, or their interaction (P >0.05).

Overall, both treatments were effective at controlling disease, i.e., maintaining relatively low levels of severity of both diseases throughout the season, with the highest recorded values not exceeding 1.2 for whole tree severity and 27.7% for branch severity. This is further reflected in the average branch severity, which remained below 10% for both diseases and both treatments.

For each scouting metric for both diseases, the variable-rate provided equal or greater pest protection compared to the constant-rate. Similarly, Chen et al. (2020, 2019) concluded that in ornamental tree nurseries, variable-rate sprayers achieved pest control that was equivalent or greater than that of conventional, constant-rate sprayers. For example, they found that apple scab [Venturia inaequalis (Cooke) Winter] and dogwood powdery mildew [Erysiphe pulchra (Cooke & Peck) U. Braun & S. Takam. and Phyllactinia guttata (Wallr.:Fr.) Lev.] severity were reduced when sprayed using variable-rate mode compared to constant-rate. Likewise, spraying in variable-rate mode yielded equivalent or better control of leafhoppers (Empoasca fabae Harris) and aphids (superfamily Aphidoidea) on maples (Acer spp.) and birch (Betula spp.), respectively, compared to the conventional constant-rate mode. Additionally, control of powdery mildew in a multi-row nursery by a variable-rate application was deemed sufficient and at times greater than control in the constant-rate treatment across the season (Fessler et al. 2021). The ability of this technology to adequately control pests is further supported by Chen et al. (2021), who found that this technology provided comparable or more effective control of insects and diseases at a farm producing orchard crops and small fruits and two ornamental nurseries across a 3-year period.

There was no effect of sprayer mode, row position, or their interaction (P >0.05) on natural enemies observed on branches. Pollinator data did not converge and therefore could not be analyzed. In another study evaluating spray technology at this location, no pollinators were recorded and fewer natural enemies were recorded than at a nearby field nursery, possibly due to the lack of ground vegetation at the pot-in-pot nursery (Fessler et al. 2022). More research is needed to further investigate the influence of spray technology on beneficial insects and other non-target organisms including nursery systems with a range of nursery floor vegetation.

There was no difference between the variable- and constant-rate modes in coverage on WSCs (P = 0.0785). There was also no interaction between sprayer mode and card position (P = 0.5580) on coverage (Fig. 5a). There was an interaction between timing and card position (P <0.0001, Fig. 5b). Among the WSCs in the canopy, for those in the exterior row facing towards and away from the sprayer and those in the interior row facing away from the sprayer, there was no difference in coverage between the early season and late season sprays, however, for those in the interior row facing towards the sprayer there was greater coverage applied during the early season spray than the late season spray, 42.1%±5.7% versus 25.9%±7.6%, respectively. During both the early season and the late season sprays, both canopy WSCs had greater coverage in the exterior row than the interior row.

For the ground WSCs, those in both the exterior row and the interior row had greater coverage in the late season sprays than in the early season sprays (P <0.0001). This could be due to mid-season pruning decreasing the width of the canopy, creating larger gaps between trees. With less canopy to intercept spray, more of the spray droplets would fall to the nursery floor. During the early season spray, the interior row ground WSC had greater coverage than the exterior row ground WSC, 24.4%±2.7% versus 8.0%±1.9%, respectively, but there was no difference in coverage at this location between the two rows during the late season spray. The intelligent, variable-rate mode had comparable ground spray coverage to the conventional, constant-rate mode. This is likely because the protruded part of the 2-container system was detected by the laser scanning sensor and was sprayed in the intelligent mode. To minimize the spray loss to the ground with the intelligent mode, the vertical detection range of the laser scanning sensor could be adjusted prior to making spray applications.

For the drift WSCs, coverage was greater in the early season sprays than the late season sprays (P <0.0001), 3.9%±1.1% versus 0.9%±0.4%, respectively. The high maximum wind speeds during the early season spray in 2020 [3.8 and 5.36 m·sec⁻¹ (8.5 and 12 mph) during the constant- and variable-rates, respectively, compared to less than 1.9 m·sec⁻¹ (4.3 mph) for all other dates and treatments] might have contributed to this difference. Overall, coverage on drift cards was low, less than 5%, regardless of sprayer mode or timing.

The lack of differences in coverage due to sprayer mode (Fig. 5a) indicates that the variable-rate mode was able to provide similar coverage when compared to the constant-rate mode. While there is no established coverage threshold to achieve protection from a foliar spray application, Chen et al. (2013a) established that an overspray threshold of 30% coverage exceeds what is deemed necessary for pest control. The canopy WSCs facing towards the sprayer in the exterior row were more than double this threshold, suggesting an opportunity for further refining spray rates. Furthermore, even though the variable-rate mode was able to greatly reduce the volume output, parameters could be further refined by increasing the lower vertical height in variable-rate mode or physically closing lower nozzles of conventional sprayers to increase sprayer efficiency when targeting canopy pests. The opportunity to refine the application parameters was further supported by the substantial off-target movement to the ground as documented by coverage on the ground WSCs exceeding the overspray threshold devised for intended targets in both positions for the late season spray.

There was a three-way interaction among timing, sprayer mode, and card position (P = 0.0435) on deposit density on WSCs (Fig. 6). However, for both the early season and the late season spray application, there were no differences in deposit density between the two sprayer modes for any given WSC position. This observation highlights the ability of the variable-rate mode to provide comparable application to intended targets while reducing total pesticide output compared to the constant-rate mode.

Within each specific canopy WSC position (i.e., exterior row facing toward the sprayer, interior row facing toward the sprayer, exterior row facing away the sprayer, interior row facing away from the sprayer), there was no difference in deposit density among the four combinations of sprayer mode and timing (i.e., variable-rate early season, variable-rate late season, constant-rate early season, constant-rate late season). This trend was also true for the ground WSC in the interior row and the drift WSC as well. However, for the ground WSC in the exterior row, there was greater deposit density during the late season application in the variable-rate mode than during the early season application in both modes (P = 0.0435).

Within the variable-rate treatment during both the early season and late season applications, there were no differences in deposit density on WSCs in the same location between the interior and exterior rows. Within the constant-rate treatment during the early season application, there was greater deposit density on the canopy WSC facing towards the sprayer in the interior row than the exterior row, 98.5±12.3 versus 34.5±8.2 droplets/cm², respectively, while during the late season application there was greater deposit density on the canopy WSC facing away from the sprayer in the exterior row than the interior row, 76.8±14.7 versus 17.5±3.0 droplets/cm², respectively. However, there were no other differences for WSCs within the same location between the rows during a given application.

Deposit density target ranges for pesticide applications are 20 to 30 droplets per square centimeter for insecticides and 50 to 70 droplets per square centimeter for fungicides (Syngenta AG 2002). These values need to be interpreted alongside coverage values though as there is a known phenomenon in which high coverage leads to droplets coalescing and being detected as a single, large droplet which in turn yields artificially low deposit density values (Nackley et al. 2021). For example, in the early season application, coverage on the canopy WSC in the exterior row facing towards the sprayer had greater coverage than the same WSC in the interior row yet the opposite was true for deposit density for the early season spray in constant-rate mode where the interior row had greater deposit density than the exterior row. Even with these artificially low deposit density values, the insecticide target range was met or exceeded at all target locations by both treatments during the early season spray. This observation was consistent with other research in apple orchards and grape vineyards that showed that the variable-rate mode could reach or exceed these thresholds (Fessler et al. 2020, Nackley et al. 2021, Salcedo et al. 2020).

There was an interaction between timing (early season or late season) and card position (P <0.0001) on deposits (Fig. 7). That is, there were greater deposits on the ground WSC in both the interior and exterior rows in the late season application than the early season application, but there were greater deposits on the drift WSC during the early season application than the late season application. This observation was consistent with the coverage results and likely occurred for the same reasons as listed above, pruning and wind, respectively. Like with coverage, deposits were low on drift cards, less than 1 μL·cm⁻². For all other WSC positions, there was no difference in deposits between the early season and late season applications.

Within both application timings, there were greater deposits on both canopy WSCs in the exterior row than the interior row (P <0.0001). For example, there were deposits of 105.9±15.4 and 101.9±19.7 μL·cm⁻² in the early season and late season sprays, respectively, for canopy cards facing towards the sprayer in the exterior row, and 25.4±11.8 and 20.3±9.9 μL·cm⁻² in the early season and late season sprays, respectively, in the interior row. However, there were greater deposits on the ground WSC in the interior row (4.2±1.3 and 57.4±10.5 μL·cm⁻² for the early season and late season sprays, respectively) than the exterior row (0.6±0.2 and 23.1±8.4 μL·cm⁻² for the early season and late season sprays, respectively). Greater deposits on the ground may be because nozzles at canopy height were directly aligned with the dense foliage of trees in the exterior row, allowing maximum interception of droplets by the canopy, whereas the lower nozzles likely overshot the exterior ground card that was located just in front of the exterior row of trees, even though these cards were in close proximity to the sprayer.

Deposit data collected from WSCs has been shown to overestimate the amount of product applied; however, the trends shown in WSC deposits have followed the same trends as deposits collected through more traditional residue analysis (Witton et al. 2018). Accordingly, the relative trends demonstrated in this data are deemed reliable, especially as they are supported by similar trends in coverage, but should not be considered absolute values. Having greater deposits in trees located in the exterior row than the interior row is consistent with research conducted at a multi-row nursery that showed that trees proximal to the sprayer had greater deposits than those distal to the sprayer (Fessler et al. 2021) and research at an orchard that found that deposits decreased deeper into the canopy (Fessler et al. 2020).

This experiment demonstrates that this intelligent, variable-rate technology can provide sufficient control that is comparable, if not better, for two fungal foliar diseases compared to conventional, constant-rate applications while substantially reducing total pesticide volume as has been seen in studies conducted in other nursery and fruit-production settings (Chen et al. 2019, Chen et al. 2021, Fessler et al. 2021). By tailoring pesticide output with this technology, not only can input costs be reduced but potential impacts to human and environmental health can also be mitigated.