Abstract
Background and goals Pesticide drift and its potential for human and environmental exposure are of significant concern during vineyard airblast spray applications. This study aimed to quantify drift from grape vineyard pesticide applications for the purpose of validating an ongoing drift mechanistic model development.
Methods and key findings A field study was conducted in a commercial Vintage Red table grape vineyard in Del Rey, California. The spray application of a fluorescent tracer dye solution was completed in 20 test runs, with the sprayer making four passes along the third drive lane upwind from the edge of the vineyard. Using a flat plastic card, artificial foliage, and horizontal and vertical polyester string samplers, airborne dye drift and drift deposition were captured, analyzed by fluorometry, and expressed as percentage of the applied rate. Meteorological data during the experiment were recorded inside and outside the vineyard using two meteorological stations with sensors installed at various heights. Airborne drift value significantly reduced (p < 0.05) from a downwind distance of ~8 to ~23 m, where it was measured. For all samplers, drift deposit also generally declined progressively, with downwind distance up to ~165 m under the different meteorological conditions that prevailed.
Conclusions and significance The results and data provide a more complete understanding of potential downwind pesticide exposure caused by drift, which will be utilized in confirming and/or developing new best practices for spray application to promote efficiency, effectiveness, and environmental sustainability.
Introduction
Grape is an important worldwide agricultural commodity, both eaten fresh and processed for juice and wine (Das and Bhattacharjee 2020). California’s grape industry is a multibillion-dollar industry, contributing $6.52 billion to California’s economy in 2023 alone (as found on the website https://www.cdfa.ca.gov/Statistics/). Protecting this crop against pests is central for producers whose livelihoods depend on it (Zheng et al. 2023); applying plant protection products (pesticides) to grapevines is crucial to maintain the viability of wine, table, and raisin grape production. Pesticides help to keep pest pressure below economic thresholds (Goldammer 2018). However, the application of pesticides to grapevines using mostly airblast sprayers (and other application equipment) is prone to drift (Sinha et al. 2019, Kasner et al. 2020). Drift is defined here as the transport of pesticide spray particles to off-site locations during application, not later drift by volatilization.
Pesticide drift can present risks to human health (e.g., pesticide exposure to workers, bystanders, and neighboring communities [Butler Ellis et al. 2010, 2017, Dubuis et al. 2023]) and to the environment (through both indirect and direct contamination [Schönenberger et al. 2022]) (Otto et al. 2015). While of concern to federal and state regulators, other risks can be introduced from pesticide drift regulation (Soheilifard et al. 2020, Schönenberger et al. 2022), such as underestimation of off-site pesticide movement which can lead to unanticipated ramifications to human health and the environment, and overestimation of pesticide exposure, which can result in unnecessary label restrictions (Rathnayake et al. 2021). Such label restrictions can give rise to unwarranted removal of arable lands from production, or to prohibition of chemicals vital for pest control, thus limiting pest control alternatives for common perennial production systems, nonnative pests needing isolation, and/or vectors of human pathogens.
Airblast spray drift has been studied across various situations to establish trends and identify factors that influence its occurrence or non-occurrence. Drift has been evaluated using actual pesticides and tracers (Derksen et al. 2007, Munjanja et al. 2020). Spray deposition and drift are influenced both by application parameters (Pergher and Gubiani 1995, Nuyttens et al. 2007, Grella et al. 2017) and target characteristics (Cross et al. 2001, Bock et al. 2015). Some methods used to minimize drift in vineyard airblast applications involve adjusting the amount of air used (in some situations, shutting the air off) or modifying the spray nozzle configuration (McCoy et al. 2022). To account for the total applied spray, the occurrence of drift from airblast spray applications is balanced with canopy deposition and ground deposition (Salyani et al. 2007, Jensen and Olesen 2014, Soheilifard et al. 2020). This provides a framework for modeling spray drift pathways (Larbi and Salyani 2012a, 2012b).
Several modeling studies have recently been conducted on drift (Holterman et al. 2017, Djouhri et al. 2023), and pesticide application models can roughly be divided into three types: full physics, empirical (statistical), and mechanistic. An overview of pesticide application modeling is presented by Teske et al. (2011).
Full physics models use basic physics to gain insight into the drift process; among the three modeling types, these models are typically the most realistic, but they are also the most complicated and require the greatest expertise and resources to run. An example of full physics modeling is of aerial pesticide application (Ryan et al. 2013). With empirical models, existing data are compiled into a statistical summary which is then used to predict future outcomes. In the case of pesticide drift, the model is often a curve of deposition versus distance—this curve may be the mean curve of many data collection test runs. If there are sufficient data and it is correctly assembled, empirical models can be powerful, however, some do not fully account for the influence of foliage density (e.g., Holterman et al. 2017). Another limitation of empirical models is that it is difficult to extrapolate beyond the environmental conditions and equipment that was used in the data collection experiments on which the models are based. The mechanistic modeling approach strikes a balance between the complexity of full physics models and the simplicity of empirical models, proposed here to allow for more realistic predictions than the empirical approach.
The empirical approach is often preferred by biological scientists and agronomists that are comfortable with statistically based conclusions; it may be of less utility to engineers that are accustomed to describing mechanical systems, and to meteorologists that continuously work with mathematical models and forecasting. A good example of empirical modeling in pesticide applications is the development efforts of the Spray Drift Task Force (SDTF) that resulted in the AgDrift model (Bird et al. 2002). Using various application scenarios (including vineyard/orchard airblast spraying), data were collected and used to create composite empirical curves. These curves of deposition versus distance are now used by regulators when evaluating the risks associated with vineyard/orchard airblast spraying. Note that these model categories are not discrete, as even full physics models are driven by empirically based relationships and various components of mechanistic models are statistically driven or derived.
In mechanistic modeling, the mechanical and environmental aspects of a given spray scenario are input and the downwind field of spray deposition and the accountancy of the spray material (evaporation, airborne droplets, etc.) are simulated and output. The list of inputs is restricted by what is easily known about the sprayer characteristics—a limitation in this type of simple (relative to full physics modeling) and applied modeling. Mechanistic modeling is exemplified by the AGDISP model (Teske et al. 2003, 2019) and by the citrus airblast model developed by Larbi and Salyani (2012a, 2012b, 2013); both of these models also require the use of field experimental data to validate drift predictions, where the characteristic variables defining the application scenario are used as input to the mechanistic model and the output drift values are compared with the measured values.
Input to mechanistic models can be grouped into categories of machine, material and release characteristics, canopy, and environment. Mechanical inputs may include a description of the droplet size in the spray (or an input that will allow characterization of droplet size from the sprayer), the position of nozzles in relation to the canopy, the amount being sprayed, the forward sprayer speed, the exit velocity or fan speed, the nozzle pressure, etc. The type of crop being sprayed may also be an input; data has been collected in grapes, citrus (Larbi et al. 2022), almonds (Larbi et al. 2025), and in foliated and dormant apples (Rathnayake et al. 2021). Other input regarding canopy architecture may also be included. Finally, basic meteorological variables are entered (such as relative humidity and wind speed). It is imperative that the new model runs quickly and that the required inputs are available. In other mechanistic models, default values are available, but these need to be carefully considered, as they are often used though never preferable to real inputs.
Mechanistic models also typically rely heavily on libraries. It is anticipated that libraries of canopy type and sprayer type might be incorporated into a vineyard/orchard airblast spray model. For instance, a user could retrieve a typical grape canopy from a library and then modify the library entry to reflect the actual row spacing, canopy height, phenology, etc. The primary audience for this kind of modeling is the pesticide regulatory community comprising federal and state agencies, although registrants and users would also benefit. Regulators need simple, generic modeling that will allow rapid evaluation of the risk of a given pesticide application scenario. Model output is used to calculate exposure to humans and the environment.
Developing the mechanistic model entails collecting new data that will help account for actual application parameters used in modern vineyard environments. Accordingly, the principal goal of the current study was to create new data based on a United States Environmental Protection Agency (USEPA)-approved protocol (Larbi et al. 2022), in support of development and validation of the mechanistic model for predicting pesticide drift. The specific objectives were to quantify spray drift from a commercial vineyard airblast application at different downwind distances under varying daytime meteorological conditions and to determine the influence of downwind distance and meteorological variables on drift. The newly generated data together with a newly developed well-founded model will enable the possibility of a more objective assessment of downwind pesticide drift, thus creating an opportunity for better understanding of drift mitigation methods to ultimately support and improve extension education trainings on spray application (Wunderlich et al. 2019). The exact form of the mechanistic model that will be developed using the data collected herein, as well as the data from over 100 other test runs in different crop types, is still under consideration.
Materials and Methods
This study, which focused on tracer dye drift measurements from airblast spray applications, was conducted in California table grapes in accordance with a USEPA-approved data collection protocol (Thistle et al. 2017). The experiment set-up was accomplished between mid-October and early November 2020, involving canopy characteristic (i.e., profile and foliage density) measurements, sprayer calibration, and field set-up of drift sampling structures and meteorological instrumentation. The experiment was performed in mid-November. Further details of the study are provided in the following sections.
Study site characteristics
The study site was a commercial Vintage Red variety table grape vineyard in Del Rey, California, located in the San Joaquin Valley (36°38′N; 119°37′W), with an open gable trellis system. The vineyard had a minimum row length of 152 m (500 ft) and an adjacent open field (bare ground) stretching to a minimum of 183 m (600 ft) downwind. Commercial vineyards with such a large adjacent open field are extremely rare or not readily available because of the very active agricultural production in the San Joaquin Valley, so finding the site for this study was difficult. The open field in this site was an old vineyard block that had been removed and transitionally laid fallow.
Although the vines in the vineyard crossed over the drive lanes to provide complete canopy cover for a great part of the season, the experiment was conducted near harvest time, when a strip of canopy along the mid-row had been removed to allow for disease management, berry color development, and ease of hand harvest (Figure 1). With the removed strip of overhead canopy, the average vine canopy profile was comparable to other grape types. Owing to the grower’s preference for the high-value crop, spraying test runs (as described below) were conducted immediately following harvest, at which point the canopy characteristics were akin to a time point earlier in the season. The mean canopy profile, not including the trunk (Figure 2), and a summary of the vine/vineyard characteristics (Table 1) were determined from a random sample of 10 vines, using a telescopic pole, a measuring tape, and a plant canopy analyzer (LAI-2200C, LI-COR, Inc.).
Table grape vines: A) view along the outer row; B) view from the edge of the row before the experiment; and C) view from the edge during the experiment, showing a removed strip of canopy.
Mean vine canopy profile (not including the trunk), determined from a random sample of 10 vines using a telescopic pole, a measuring tape, and a plant canopy analyzer (LAI-2200C, LI-COR, Inc.).
Attributes of the table grape vineyard used in the study.
Field set-up
In the field set-up, upwind was considered negative direction into the vineyard, while downwind was taken to be positive direction away from the vineyard. All distances were measured from the edge of the vineyard. Spray drift collection stations made up of artificial sampling structures were distributed along four equally spaced transects (18 m [60 ft] between adjacent transects, T1 to T4) from upwind, up to 183 m (600 ft) downwind. The station sampling structures had combinations of flat plastic cards (C), artificial foliage (AF), and string collectors (horizontal string [HS] and vertical string [VS]), with individual collection stations of similar distance taken as subsamples. Towers for the VS samplers were stationed at ~8.0 m (25 ft) and ~23.0 m (75 ft) downwind, and each had a basic station at the base (sampling height of ~0.91 m (3 ft) above ground level [AGL]). Figure 3 presents an oblique aerial view of the field set-up, showing the application and drift sampling areas. Eye-level views of the field set-up showing the different sampling structures are also shown in Supplemental Figure 1, providing a full view of one of the towers used for holding VS samplers. The VS samplers were pre-marked and cut after each test run at 1H, 1.5H, and 2H (where H = vine height) to obtain three samples. Supplemental Figure 2 further provides close-up views of three sampling structures that were used to hold different combinations of horizontal samplers.
Oblique aerial view showing the application area (vineyard), the nearby sampling area (light colored area in open field), and the meteorological stations (Met 1 and Met 2). The white broken line indicates sprayer travel path, and the red flags represent spray start/end points in both directions. Transect sampler locations: C, flat card (11 per transect); AF, artificial foliage (10 per transect); HS, horizontal string (5 per transect); VS, vertical string (2 per transect). Met 1 was located inside the vineyard at 21.9 m (72 ft) upwind; Met 2 was located outside the vineyard at 186.9 m (610 ft) downwind.
Meteorological instrumentation and measurement
To record meteorological data for assessing variation in spray drift, two meteorological stations with sensors installed at different heights were set up, one upwind inside the vineyard (Met 1) and another ~183 m (600 ft) downwind in the adjacent open field (Met 2). Each sensor height had a 3-D sonic anemometer (Young’s 3D Ultrasonic Anemometer, R.M. Young Company) and an all-in-one weather sensor (ATMOS 41, METER Group, Inc.). The instruments were connected to data loggers (Zentra ZL6, METER Group, Inc.; CR1000/CR1000X, Campbell Scientific) and powered by 12-volt power sources. Data were logged at 1-min intervals and retrieved from the sensors at the end of the test runs. Table 2 provides a summary of the sensors used at both meteorological stations.
Instrumentation used to collect meteorological parameters. Sensors were installed at two meteorological stations (Met 1 and Met 2), at different heights. Met 1 was located inside the vineyard at 21.9 m (72 ft) upwind; Met 2 was located outside the vineyard at 186.9 m (610 ft) downwind. Wind direction was monitored at Met 2 and used to trigger the start of spraying for each test run. AGL, above ground level.
Application equipment and parameters
The application equipment used in the study was a power take-off powered conventional airblast sprayer (GB-2/32, Air-O-Fan) with five open nozzles per side attached to a T4.95F New Holland tractor (New Holland Agriculture). The sprayer was calibrated by measuring and determining travel speed, assessing the air profile, measuring spray nozzle flow rates, and adjusting sprayer settings to obtain a target application rate of 561 L/ha (60 gpa) at a chosen operating pressure. The selected application rate was based on a grower standard and was typical of the region. The number of open nozzles used to apply the spray was chosen so as to direct the spray onto the target canopy. A summary of the calibrated parameters used in the study is shown in Table 3 and details of the nozzle type and configuration are also provided in Table 4.
Spray application parameters used in the study. Wrt, with respect to; gpm, gallons per minute; gpa, gallons per acre.
Nozzle configuration of the sprayer used in the study. A dash (-) indicates that the nozzle was nonexistent by design; an “X” indicates that the nozzle was closed (i.e., not used). Wrt, with respect to; gpm, gallons per minute.
Spray tests
Spray tests were performed between 0800 and 1800 hr local time. Twenty spray test runs (replications) were conducted using the spray equipment described above, by traversing through the third drive lane upwind (Row 3), as shown in Figure 3. For each replication, a complete set of samplers were mounted prior to spraying, then collected after spraying. The VS samplers were meticulously retrieved from the towers and cut into three predetermined sections of strings. To determine when to start spraying, the data from the ATMOS 41 instrument at 1.8 m (6 ft) at the Met 2 station were monitored in real time. Over a distance of 152 m (500 ft), the spray was applied from both sides of the sprayer to the immediate vine row on each side, in four passes (two passes per direction of row). Of the 20 test runs, 17 (2 to 11 and 13 to 19) involved applying a solution of pyranine fluorescent tracer dye (sodium 8-hydroxypyrene-1,3,6-trisulfonate) at a target concentration of 2 g/L (2000 ppm), whereas three runs (1, 12, and 20) were blank applications (clean water with no dye). Drift samples were collected into prelabeled zipper bags after each run and stored in a cooler underlain with ice packs. Tank samples were also obtained at different stages of the experiment to normalize the drift data and to be used for quality control/quality assurance assessment. The samples were transported to the lab and stored in a refrigerator while awaiting sample analysis (Salyani 2000). The meteorological data were retrieved at the end of each day of the experiment.
Sample analysis
The spray drift samplers and sprayer tank samples were analyzed by fluorometry (Salyani 2000). The drift sample analysis was accomplished by transect and by replication, meaning that all samples for Transect 1 were analyzed by replication before Transects 2, 3, and 4. This procedure was part of the quality control measures to prevent errors during sample analysis. The fluorescence data obtained were converted to dye concentration using a calibration curve for the dye, and then to dye deposition (Larbi 2022). The actual spray application rate for each test run was determined and the data were combined with the tank concentration data to normalize the dye deposition data. The dye deposition data were expressed in ng/cm2 and as a percentage of the applied rate.
Statistical analysis
The study data, including the meteorological data retrieved from all the sensors, were organized and processed in Microsoft Excel (Microsoft Corporation). The processed deposition data were used to produce scatter plots of airborne drift captured with the VS samplers and downwind drift deposition (sedimentation) evaluated with the C, AF, and HS samplers. Mean plots of the data were created and visually compared. It was not possible to reliably fit the C drift deposition data, however, logarithmic curves were fitted to the AF and HS data and the fit equations were utilized in estimating drift termination distance and final deposition value. Airborne drift comparison was made between the two downwind distances. Visualization of downwind drift deposition was accomplished by considering the trend of the overall mean, the comparison between transects, and the comparison between upwind and downwind mean wind direction (upwind implying wind had a zero or negative downwind component; downwind implying wind had a positive downwind component).
All statistical analyses, including post-hoc pairwise multiple comparisons where necessary, were conducted with a significance level of 0.05 in SigmaPlot 12.5 (Grafiti LLC). A two-way analysis of variance (ANOVA) of downwind distance and sampling height was performed on the airborne drift data. A variety of two- and three-way ANOVAs were also conducted on the downwind drift deposition data. Collection efficiencies at the five sampling locations (15, 30, 61, 122, and 183 m) common to the horizontal samplers were compared using one-way ANOVAs. The main effects of the meteorological variables on downwind drift deposition were analyzed by linear regressions also at those common locations. The collective effect of meteorological parameters (solar radiation, wind direction, wind speed, air temperature, atmospheric pressure, and relative humidity) on mean drift deposition was analyzed by multiple linear regression (MLR) to identify redundant and non-influential factors. Interactions among the factors were not analyzed.
Results
Meteorological conditions
A summary representation of the meteorological conditions during the spray experiment, as recorded by the all-in-one meteorological instrument mounted at 1.8 m AGL at the Met 2 station, is shown (Supplemental Figure 3). The data represent meteorological conditions from 2 min before the start of spraying through the end of sample retrieval following spraying in each test run. The end of sample retrieval was adjusted to 10 min post-spraying for consistency among test runs. The meteorological variables showed different levels of temporal variation during the 20 spraying test runs. Solar radiation varied with varying amounts of sunshine and cloud cover. Lower solar radiation values resulted from lower ultraviolet light intensity and greater cloud cover. Wind speed was mostly highly dynamic in varying directions within each test run. Air temperature and atmospheric pressure were highly variable among test runs, but mostly stable within each run. Vapor pressure and relative humidity, which are directly related, were also mostly stable and under 100%. Compass rose summaries comprising all wind speeds and directions during all the spray test runs (Supplemental Figure 4) demonstrate differences in meteorological observations among the sensors at different heights AGL.
The meteorological data logged during the experiment are summarized in Table 5. Averaging the data collected from the Met 1 and Met 2 stations, the minimum and maximum values for each application parameter ranged from 0.0 to 590.4 W/m2 for solar radiation, E to W for wind direction, 0.3 to 2.2 m/sec for wind speed, 9.5 to 23.0°C for air temperature, 0.7 to 1.6 kPa for vapor pressure, 100.1 to 101.4 kPa for atmospheric pressure, and 34.0 to 87.8% for relative humidity.
Summary meteorological data for the treatment test runs based on ATMOS 41 all-in-one meteorological sensors. Sensors were installed at two meteorological stations (Met 1 and Met 2), at different heights: 2.4 m (8 ft) at Met 1 and 1.8 m (6 ft) at Met 2. Met 1 was located inside the vineyard at 21.9 m (72 ft) upwind; Met 2 was located outside the vineyard at 186.9 m (610 ft) downwind.
Compass roses summarizing all wind speeds and directions during the treatment test runs are provided (Supplemental Figure 5). South wind conditions (i.e., northbound winds) generally persisted during the spray test, as desired (Supplemental Figure 4B). A two-way ANOVA at 5% significance level with inputs of meteorological station and meteorological variable (not including wind direction) indicated no significant difference (p > 0.05) between the Met 1 and Met 2 stations. Therefore, only Met 2 meteorological data were employed to establish the effect of meteorological conditions on drift, as described later.
Off-site spray drift
Vertical string (VS) samplers
VS samplers were used in each spray test run to capture the drifting spray (i.e., airborne spray drift) at the downwind distances of ~8.0 and ~23.0 m. The drifting spray captured at the two downwind distances was compared, showing a significant reduction in the amount captured at the distance further from the spray release, than that captured at the closer distance (p < 0.05) (Figure 4). In addition, the airborne drift captured at the bottom sampling section was significantly greater than both the midsection (p < 0.05) and the topmost section (p < 0.05), but the midsection and the topmost section failed to show significant difference (p > 0.05).
Spray drift deposit representing airborne drift collected at downwind distances of ~8.0 m (25 ft) and ~23.0 m (75 ft) on vertical strings, at string midpoints of 1H = 0 to 2.4 m, 1.5H = 2.4 to 3.6 m, and 2H = 3.6 to 4.8 m (where H is vine height).
Flat card (C) samplers
C samplers were used to capture spray deposition both within the vineyard (upwind, which is not drift) and downwind beyond the edge of the vineyard (drift deposition). The mean data in linear (Figure 5A) and log-linear plots (Figure 5B to 5D) are shown. Generally, the plots show downwind deposit decay from the 0 m distance, which is muffled in the linear plot because of high deposition at the −9.1-m (−30-ft) downwind distance, i.e., Row 2. Visualization of the overall mean data (Figure 5A) is enhanced in Figure 5B. The variation among sampling transects is shown (Figure 5C), and a reduction in drift deposition is indicated, from upwind compared to downwind mean wind direction, starting from 61 m downwind (Figure 5D). It is suggested in Figure 5D that drift occurred regardless of the average wind direction, however, drift amount decreased when the wind was blowing in the upwind direction.
Spray dye deposit collected on plastic card samplers versus downwind distance (0 m = edge of vineyard) inside and outside of the vineyard. A) Linear plot of mean deposition, and log-linear plots of B) mean deposition; C) transect comparison; and D) wind direction comparison. Plots A and B are complementary plots of the same data for visual enhancement.
Based on a three-way ANOVA (Table 6), drift was significantly reduced over the entire downwind distance (p < 0.05), but further pairwise comparisons revealed the reduction to be significant up to only the first 30 m. Downwind drift values were significantly different among the 17 treatment test runs (p < 0.05). However, no significant difference in drift was established among the transects (p > 0.05).
Output from a three-way analysis of variance on spray drift data that were collected from artificial sampling structures (flat plastic cards [C], artificial foliage [AF], and horizontal string [HS]), used to evaluate the effects of a spray test run, sampling transect, and downwind distance. DF, degrees of freedom; SS, sum of squares; MS, mean sum of squares; F, F-statistic.
Artificial foliage (AF) samplers
Spray drift data created from the AF samplers are summarized (Figure 6). As with the downwind drift data from the C samplers, drift from the AF samplers varied significantly across the test runs (p < 0.05) and declined significantly downwind (p < 0.05), but no significant difference was established among the transects (p > 0.05) (Table 6). Fitting Figure 6A, 6B, and 6D with logarithmic relationships (curves not shown), a summary of estimated mean termination downwind distances where AF drift completely decayed, and corresponding final drift values, is shown (Table 7).
Spray drift deposit collected on artificial foliage samplers with downwind distance. A) Linear plot of mean deposition, and log-linear plots of B) mean deposition; C) transect comparison; and D) wind direction comparison. Plots A and B are complementary plots of the same data for visual enhancement.
Summary of estimated drift termination distance and final drift amount, collected from artificial sampling structure data (artificial foliage [AF] and horizontal string [HS]) and based on logarithmic curve fitting.
Horizontal string (HS) samplers
Spray drift data captured with the HS samplers are shown (Figure 7). Similar to the C and AF drift data, a significant difference was established for HS drift among the test runs (p < 0.05) and over downwind distance (p < 0.05), but not among transects (p > 0.05), based on a three-way ANOVA (Table 6). A summary of the downwind distance at which HS drift completely decayed, and the final drift values based on logarithmic curve fitting (not shown) of the data in Figure 7A, 7B, and 7D, is shown (Table 7).
Spray drift deposit collected on horizontal string samplers with downwind distance. A) Linear plot of mean deposition, and log-linear plots of B) mean deposition; C) transect comparison; and D) wind direction comparison. Plots A and B are complementary plots of the same data for visual enhancement.
Sampler efficiency
The collection efficiency of the horizontal samplers (C, AF, and HS) was evaluated at the common downwind distances, as shown in Figure 8. Similar to previous findings (Thistle et al. 2009, Rathnayake et al. 2021), the results portray that drift deposition measured by the AF and HS samplers was significantly greater (p < 0.05) than drift deposition measured by the C samplers.
Spray drift deposition at common downwind distances, collected from artificial samplers (flat plastic cards [C], artificial foliage [AF], and horizontal string [HS]). Lowercase letters are mean separation letters. Multiple one-way analyses of variance (ANOVA) were run; subscripted numbers correspond to the particular instance of the ANOVA that was run. SEM, standard error of the mean with reference to sampling transects.
Airborne drift effect on sedimentation
Results from a multiple correlation assessment evaluating how the initial amount of airborne drift (as measured near the edge of the vineyard) affects the amount of downwind drift (sedimentation) are summarized (Figure 9). Airborne drift data were averages of the values from the ~8-m and ~23-m downwind locations and sedimentation data were the averages of C, AF, and HS drift values. Drift values at all downwind distances generally increased as airborne drift values increased (Figure 9).
Effect of initial spray airborne drift on downwind drift deposition at common downwind distances.
Meteorological influence
The effects of the meteorological variables on downwind drift were established as correlations of meteorological variables from the 1.8-m AGL sensor at the Met 2 station against drift at the 15-, 30-, 61-, 122-, and 183-m downwind sampling stations common to the C, AF, and HS samplers (Figure 10). For all the meteorological variables, the trends vary among the sampling locations.
Sensitivity of spray drift deposition to increasing values of meteorological variables.
Using MLR, overall mean drift deposition, which was the dependent variable, can be predicted from a linear combination of the six independent meteorological variables (p < 0.05). Based on the summary output of the analysis, solar radiation, wind speed, and relative humidity appear to account for the ability to predict overall mean drift deposition (p < 0.05) (Table 8).
Output of multiple linear regression based on ATMOS 41 all-in-one meteorological sensor installed at a height of 1.8 m (6ft) at the Met 2 meteorological station, to evaluate the effect of meteorological conditions on overall mean drift deposition. VIF, variance inflation factor.
Discussion
The wind data show that overall wind direction for the treatment test runs varied during the experiment despite striving to start from southwards, based on monitoring at the Met 2 meteorological station. This variation in meteorological condition is consistent with observations made during similar drift studies in apples (Rathnayake et al. 2021), almond (Larbi et al. 2025), and citrus (Larbi et al. 2022). Off-site spray drift data measured downwind using the different spray deposit samplers generally show that drift deposit progressively declined with downwind distance under varying prevailing meteorological conditions during the test runs. This decline can be chiefly attributed to continued spray dispersion over the downwind distance. Some differences in the drift data were observed among the different samplers, possibly an indication of the differences observed in their collection efficiencies.
The bottom section of the VS samplers captured the amount of drift emerging directly from the far side of the canopy, having missed mainly the canopy and trunk. The middle section captured spray emerging through the canopy immediately above it. The upper VS section captured drift reaching further beyond the top of the canopy. A qualitative assessment of drone footage of the application indicates that much of the spray emerged through the top of the narrow canopy over the entire treatment path while spraying.
The significant reduction in the amount of airborne drift captured by the VS samplers as well as the reduction in the downwind drift deposition captured by the C, AF, and HS samplers was due to dispersion (together with droplet evaporation) and deposition of spray material during downwind transport. The differences in downwind drift deposit values from all sampler types among the treatment test runs were most plausibly due to differences in prevailing meteorological conditions during the different test runs, as previously mentioned. The similarity in drift between the four transects implies that the trend was repeatable among the subsamples, as designed.
Analysis of the influence of meteorological condition on drift denotes that not all meteorological variables appeared to be necessary to explain the overall mean drift deposition that occurred in this study. It is possible that the MLR model (referring to a statistical model used for data analysis, not the mechanistic model under development) used for the analysis was underspecified. However, the results were also partly due to the presence of multicollinearity among the independent variables, mainly caused by the variables with the highest values of variance inflation factor (VIF), as indicated in Table 8 (i.e., air temperature), which can be excluded from the MLR model to improve the relationship. Both air temperature and relative humidity showed high VIF values in similar studies conducted in citrus and almond (Larbi et al. 2025).
The focus of this study on conventional radial airblast sprayers is due in part to airblast sprayers being the most used equipment for vineyard pesticide applications, despite research indicating superior performance of other sprayer types such as tower and over-the-row sprayers. A major observation is that most growers own or manage more than one field, including both vineyards and orchards. Typically, they are limited to using the same spray equipment, for which conventional radial airblast sprayers present the most versatility. For instance, tower sprayers are limited in the range of target height the spray can reach, and over-the-row sprayers are limited in the range of canopy width or row spacing for which the sprayer can be used. These limitations, among other considerations, contribute to their low adoption rate among California growers.
Drift potential of airblast sprayers is generally well known to be influenced by several factors including application parameters, crop-field characteristics, and meteorological condition. However, because the focus of this study was mainly on the effect of meteorological variables at various downwind distances from the treated vineyard, neither application settings nor crop-field variables were varied. Nevertheless, these are appropriately considered as input variables in the related model development, and drift estimates are expected to vary in proportional measures. For instance, while the study was conducted for table grapes, the data collected can be generally applied to grapes by defining the vine canopy profile and foliage density, and the vineyard properties such as inter- and intrarow spacing for model validation purposes. These important variables are accounted for in the related model development and varying them should affect the drift outcome in proportional measures. As such, although they are not considered as factors for explaining drift in this study despite having been measured, they will be utilized for comparing drift in other crops (e.g., apple, citrus, and almond), as mentioned earlier.
Daytime airblast spraying operations considering the meteorological conditions, as was done in this study, were consistent with practices common to the production area of the San Joaquin Valley, although nighttime applications are also generally common in western United States viticulture. The main differences between daytime and nighttime conditions are decreased wind speed and the absence of sunlight during the night. Thus, after properly isolating solar radiation, the drift data from this study can be applied to nighttime applications of similar conditions. Therefore, because meteorological variables are included as input variables in the related model development, predictions should account for whatever conditions prevail during the simulated spray application.
Conclusion
Humans and the environment can potentially be exposed to pesticides by drift from vineyard spray applications. Measurements from table grape airblast spray application under varied meteorological conditions demonstrated material drift up to nearly 170 m downwind from the treated vineyard. Airborne drift flux varied significantly over sampling height (p < 0.05), and the amount significantly reduced over downwind distance (p < 0.05), from ~8.0 to ~23.0 m. Due to dispersion and deposition of spray material during downwind transport, drift deposition significantly diminished with downwind distance (p < 0.05) until termination but was mainly significant within up to the first 30 m. Drift deposition terminated at a mean distance of 106.4 m for AF samplers and 164.4 m for HS samplers, based on logarithmic curve fitting. Notably, AF and HS samplers both produced greater drift deposition values than C samplers at common downwind distances. As these drift termination estimates are limited to the specific characteristics of the vineyard sprayed and the meteorological conditions that persisted, they are not intended to be used for generalized predictions outside of the range of explanatory variables. Corresponding variability in the drift data was significantly affected by variation in meteorological condition among spray test runs (p < 0.05), but due to multicollinearity among the six variables, only solar radiation (p < 0.05), wind speed (p < 0.05), and relative humidity (p < 0.05) appeared to be relevant in explaining the variability. Overall, winds blowing upwind or across the downwind direction, as opposed to blowing in the downwind direction, resulted in less drift and shorter drift termination distances. The generated data set represents a rich resource to support the validation of a mechanistic model being developed to estimate drift from vineyard and orchard airblast applications, accounting for canopy characteristics and meteorological parameters. The future model will account for drift from all included rows upwind of the treated vineyard and thus, future work will evaluate how moving the sprayer upwind influences the amount of drift measured at various downwind locations. It should therefore be noted that until completion of the future mechanistic model, the data provided in this paper are not intended to be directly compared with existing regulatory models.
Supplemental Data
The following supplemental materials are available for this article in the Supplemental tab above:
Supplemental Figure 1 Views of the field set-up, showing installed horizontal sampling structures and vertical string sampling structures.
Supplemental Figure 2 Built sampling structures for different sampler combinations. A) Holds flat card, artificial foliage, and horizontal string; B) holds flat card and artificial foliage; C) holds flat card or artificial foliage.
Supplemental Figure 3 Temporal variation of meteorological variables during spray experiments.
Supplemental Figure 4 Compass roses of all wind speeds and directions at various sensor heights above ground level (AGL) at the Met 2 meteorological station measured during the spray experiment.
Supplemental Figure 5 Compass roses of all wind speeds and directions during treatment test runs at the Met 1 (A; 2 m sensor height above ground level [AGL]) and Met 2 (B; 1.8 m sensor height AGL) meteorological stations.
Data Availability
The data underlying this study are available on request from the corresponding author.
Footnotes
This research was funded by the California Table Grape Commission (Y20-4996), Washington State Wine Commission (Y20-5159), and E. & J. Gallo Winery (Y20-5161), Almond Board of California/UC ANR (University of California Division of Agriculture and Natural Resources) Spray Technology Grants, Project No. 19-6065, Almond Board of California (Water14.Larbi), Citrus Research Board (5400-161). Additional funding was provided by the UC ANR. Access to research site was made possible by a cooperating grower/landowner. Special acknowledgments to the following people for providing field and/or lab assistance: Christian Basulto (Staff Research Associate (SRA), Agricultural Application Engineering Laboratory (AgAppE Lab), Kearney Agricultural Research and Extension Center (KARE Center)); Dr. Franz Niederholzer (University of California Cooperative Extension (UCCE), Colusa and Sutter/Yuba Counties); Dr. Mae Culumber (UCCE Farm Advisor, Fresno County); Dr. Gabriel Torres and Dr. Greg Douhan (former UCCE Farm Advisors, Tulare County); Daniel Cabrera, Sharon Asakawa, Ruben Chavez, and David Rodriguez Herrera (UC ANR/AgAppE Lab); Daniel Syverson, (UC ANR), German Zuniga-Ramirez (UC Davis Digital Ag Lab); Parry Klassen, Courtney Jallo, Maureen Thompson, Ezra Klassen, Marcia Klassen (Coalition for Urban and Rural Environmental Stewardship). The mention of trade names and commercial products is solely for providing specific information and does not imply any recommendation by the authors or by the University of California.
Larbi PA, Zhuang G, Thistle HW and Willett MJ. 2025. Downwind drift from grape airblast spray applications: Field evaluation to support mechanistic model development. Am J Enol Vitic 76:0760025. 10.5344/ajev.2025.25005
By downloading and/or receiving this article, you agree to the Disclaimer of Warranties and Liability. If you do not agree to the Disclaimers, do not download and/or accept this article.
- Received January 2025.
- Accepted July 2025.
- Published online October 2025
This is an open access article distributed under the CC BY 4.0 license.
References
Vol 76 Issue 2
Jump to section
Related Articles
Cited By...
More from this TOC section
Similar Articles
