Introduction

In recent decades, food sustainability and consumer protection gained importance and relevance, as evidenced by the growing consumer interest in vegetable production and distribution (Melović et al., 2020; Rahman et al., 2021). Given the escalating global demand for food and considering the significant impact of biotic factors on annual yield losses (FAO, 2017), plant protection products (PPP) persist in safeguarding crops from insects, pests, and fungi in agricultural scenarios (Lykogianni et al., 2021). PPP distribution is a high-risk operation, which poses substantial risks and results in residues in food, feed, water bodies, and non-target organisms, with profound implications for consumer safety and the environment (Akpan et al., 2023; Chavarri et al., 2004; Syafrudin et al., 2021). Commonly executed through vehicles equipped with spraying systems following predetermined paths in the field, PPP applications pose challenges in terms of time, labor, and health risks for operators (Dhananjayan & Ravichandran, 2018). As farmers predominantly favor traditional ground spraying methods, evaluating application efficiency becomes crucial to minimize human and environmental hazards (Dhananjayan et al., 2020; Kestur et al., 2020; Van Den Berg et al., 2020). Although Aerial spraying performed through airplanes and helicopters is considered an economical and fast method for PPP spraying able to cover large fields without physical impact on crops or soil structure, it is prone to drawbacks such as poor distribution accuracy (Quantick, 1979), leading to high product overdose and losses, requiring high flight altitudes and speeds and large areas for take-off and landing, and it is not suitable for small-scale farms, forcing farmers to adopt ground-based technologies (Baio et al., 2019; Grella et al., 2017; Woods et al., 2001). Addressing these concerns has spurred policy initiatives, legislative regulations, and an increasing demand for environmentally friendly management methods. Notably, these concerns led the European Parliament to prohibit aerial PPP distribution in Union countries (European Parliament 2009) except in specific limited situations where conventional operations are not feasible.

Unmanned Aerial Systems (UAS) are remotely piloted aircraft platforms suitable for remote sensing data missions across multiple scenarios and scientific disciplines (Tsouros et al., 2019). UASs have become crucial tools for agriculture thanks to their ability to provide high spatial resolution data at a high temporal frequency (Mogili & Deepak, 2018). Several UAS applications can help improve knowledge about agricultural complex scenarios, sustainability, and operational efficiency in agriculture (Radoglou-Grammatikis et al., 2020), like water management (Maimaitiyiming et al., 2020; Pádua et al., 2019; Park et al., 2017; Zhang et al., 2019), fertilizers management zone delineation (Kayad et al., 2021; Seo et al., 2019), and weed detection (De Castro et al., 2018; Peña et al., 2013). Although UASs are highly automatable and have achieved a high build quality, many safety issues exist, limiting their autonomous application in agriculture. Environmental conditions, large data volumes, flight legislations, the mandatory presence of a pilot during the flight, the ground control system vulnerability to interference, and the limited flight autonomy represent the most common UASs limitations (Kakaletsis et al., 2022). Despite that, because of UAS’s ability to access terrains hardly reachable by terrestrial machinery and operate under multiple conditions, their integration into diversified farming practices represents a step forward to more efficient and sustainable agriculture (Sarri et al., 2019). In recent years, multi-rotor Unmanned Aerial Spraying Systems (UASSs) have gained attention due to their ability to follow complex patterns, fly at low altitudes, adapt to hilly terrain, take off and land vertically, perform low-volume and site-specific PPP applications, limiting operator’s health risks with a low operational cost, preserving the crops and soils structure form physical damage (Qin et al., 2016). When working with UASS there are some aspects to consider which may affect the result of the treatment including spraying and flight settings (pump pressure, flight height and speed, etc.), payload capacity, weather conditions (specifically, wind direction and velocity), flight range, and environmental dispersion (drift effect and ground leaking) (H. Chen et al., 2021; Hunter et al., 2020; G. Wang et al., 2020). Among all the requested technical improvements, nozzles represent a crucial component in reducing droplet dispersion and improving target coverage. Major companies are beginning to develop UASS nozzle types for aerial spraying, but their characteristics are not yet standardized. Further research is needed to find the best solution to make treatments as effective as possible (Lou et al., 2018; Qin et al., 2016), including automatization strategies based on Artificial Intelligence (Sassu et al., 2023).

After Japan pioneered the application of helicopter UASS for crop protection in the 1980s, in the last decade, South Korea and China have gradually adopted UASS, witnessing a surge in aerial spraying applications contributing to the emergence of several manufacturers. Due to a service operations profitability decline related to market expansion, UASS’s manufacturers and service providers are shifting their focus from herbaceous crops to fruit trees with higher economic value (C. Wang et al., 2022). Vineyards, exposed to several diseases recurring during certain periods of the year and under specific climatic conditions and growing stages, require prompt and successful treatments (Pertot et al., 2017). Except for monitoring purposes (Sassu et al., 2021), the application of UASS in vineyards is not yet well investigated despite being a viable alternative to conventional application methods.

A study by (Sarri et al., 2019) reported the UASS spray application results on a small vineyard (3000 m2) of Sangiovese cv. in the mountains of Massa (Italy) and its comparison with a knapsack sprayer and a sprayer gun. The aerial spraying operations were performed with a commercial six-rotor UASS (Bly-c-agri model by AERMATICA3D, Como, Italy) at 5 m flight height AGL, 0.5 m s−1 speed, perpendicular to the rows along the maximum slope gradient line. The research team specified that a preliminary and unsuccessful test was conducted over the rows with nozzles positioned on the inter-rows, resulting in low leaf coverage and ground dispersion, probably due to the low GNSS accuracy, which did not allow the UASS to keep a straight direction over the row. The research team focused on working capacity (ha h−1), droplet coverage (%), droplets per cm2 (n), and volume median diameter (μm), extracted from WSPs image analysis performed through ImageJ software with the DepositScan plugin and applying the spread factor to get an estimation of the original droplets. Promising results were obtained about work capacity, suggesting that work time was greatly reduced using the UASS compared to the spray gun and knapsack sprayer. The aerial application of the UASS was less effective in creating droplet diffusion and reaching the lower part of the plant than the high-volume applications of the spray gun, which gave better results, but with a high risk of runoff. These differences could also be related to the maximum working pressure of the UASS pump, which was limited to 0.4 MPa. The aerial application resulted in a not oriented droplets diffusion, which led to a not uniformly coverage of the plants. Results showed that the sprayer gun gave the highest droplet coverage, and the UASS gave the lowest (from 0.90–3.83%). For each WSP position, the UASS had the lowest number of droplets per cm2 (ranging from 8.27–32.35) with no differences at the three sampling heights and in the internal part of the plant, the size of the droplets was similar (ranging from 209.24–306.39 µm), due to a high variability. As pointed out by the authors, the low coverage results were caused by the low operating pressure of the pump, the too-high distance between the nozzles and the plants, and the absence of a conveying effect of the air jet generated by the propellers towards the plant canopy. The covered area, droplets per cm−2, and NMD are far from the results obtained in this work, where only a few treatments at the lowest sampling position showed comparable results. The lower flight height above the canopy helped gain better deposition performances because of the closer position of the UASS to the target, pushing the droplets down inside the canopy and reducing the drifted product. The different path followed during the operation perpendicular to the rows could also have affected the deposition results.

Another work by Biglia et al. (2022) aimed to evaluate a UASS system, a six-rotor DJI Matrice 600 Pro equipped with a customized standalone sprayer system, performance in an experimental vineyard (Barbera cv.) at complete growth stage and compare it to an axial fan sprayer coupled with a tractor. Three different flight modes were evaluated for their effect on both canopy deposition and ground losses, combined with two cruise speeds (1 and 3 m s−1) and different nozzle types (conventional and air inclusion) at 3 m height. The results indicated that the band flight mode (along the vine rows) improved spray application performance, increasing the average canopy deposition and reducing the average ground losses, in contrast with those by (Giles & Billing, 2015), who suggested the cross-flight (perpendicular to the rows) to the vine rows orientation as the best flight mode to improve canopy deposition. In agreement with Biglia et al. (2022), further trials are needed to better understand the most suitable spray application rate for UASS applications in vineyards to guarantee adequate treatment efficacy, also according to the different growth stages that are characterized by different canopy density and shape. Differently from what was found by Biglia et al. (2022) the results showed much fewer in-field ground losses compared to the spray deposits measured on the canopies, and these phenomena could be related to the investigated IR UASS flight positions.

The work by Wang et al. (2021) aimed to compare the spray performance of three commercial UASS models equipped with hollow cone nozzles (HCNs) and air-injector flat fan nozzles (AIN) in an artificial vineyard through the analysis of deposition characteristics, drift, and mass balance. Three UASS models were selected: a gasoline-powered helicopter (3WQF120-12, Anyang Quanfeng Aviation Plant Protection Technology Co., Ltd., Henan, China), an electric six-rotor UASS (3WM6E-10, TT Aviation Technology Co., Ltd., Beijing, China, Fig. 2b), and an electric eight-rotor UASS (3WM8A-20, TT Aviation Technology Co., Ltd., Beijing, China). In their trials, the path followed by the UASSs was perpendicular to the rows’ orientation. The 6-rotor UASS showed the best performance, followed by the 8-rotor and the helicopter. The AIN promoted spray deposition and uniformity and reduced drift significantly compared to HCN for all tested UASS, improving the utilization of plant protection products. However, the worst spray performance was achieved with the helicopter due to its stronger rotor’s airflow, suggesting the downwash airflow field is one of the most important factors influencing the spray deposition characteristics and drift.

As a relatively new technology, evaluating their effectiveness and spraying configuration is essential to understand when and how to implement UASS in agricultural PPP distribution scenarios and to compare them to terrestrial machinery performances (Biglia et al., 2022; Morales-Rodríguez et al., 2022; Bloise et al., 2022; Rapareli et al., 2021; Sarri et al., 2019; Severin & Söffker, 2022; Wang et al., 2021). Thus, the authors initial hypothesis of the manuscript suggests that UASS can effectively and safely perform PPP spraying on row crops, particularly vineyards, given appropriate environmental conditions and flight settings.

Specifically, this research aims to verify and characterize a commercial UASS distribution uniformity and dispersion, testing different flying parameter combinations on a vineyard. The spraying performance, evaluated with Water Sensitive Papers (WSPs) placed inside the canopy and on the ground along the inter-row, were obtained through a MATLAB code, able to extrapolate several droplet properties from the WSPs such as the distribution of target and off-target rows droplet deposition for the considered variables, namely the covered area, droplets per cm2, the Number Median Diameter (NMD), and the penetration rate through the canopy.

Material and methods

Study site

The experimental field was a 480 m2 vineyard (Savatiano variety—Vitis Vinifera L.) located in Spata, Greece (228,285.983 East – 4,208,689.259 North; WGS84 – UTM 35 N, EPSG: 32,635) at 123 m above the sea level, 20 km East from Athens (Fig. 1). The grapevines were planted with a 1.5 m × 2 m spacing, a NW–SE rows’ orientation, and trained as Vertical Shoot Position (VSP). The survey area, selected for its position, far from the field borders, and its flat location, was a small plot of the experimental vineyard owned by the Agricultural University of Athens. Experiments took place on 21, 22, 28 June and 13 July 2022 (BBCH 77). The weather conditions faced during the experiments were optimal for UASS operations: sunny days, clear sky, and wind absence.

Fig. 1
figure 1

The location of the field in Spata (indicated with a red dots), 20 km east of Athens (Greece) in a graphical representation of Greece’s borders obtained using QGis software, and the view from above of the experimental vineyard obtained from UAS (Phantom 4 Pro, DJI Technology Co., Ltd, Shenzhen, China) flight images and elaborated through Agisoft Metashape Structure from Motion (SfM) software

Full size image

The UASS platform

The UASS used for the trials was a commercial and ready to use Agras T16 (DJI Technology Co., Ltd, Shenzhen, China) equipped with a spraying system with a 16 L thank, eight nozzles (XR11001VS, Teejet Technologies, Glendale Heights, Illinois, USA), an RTK navigation system, a DBF imaging Radar, and FPV camera (Fig. 2). The total weight of the UASS is 18.5 kg excluding the battery and an empty thank, the hovering time can vary from 18 to 10 min depending on the takeoff weight, the wind speed, and the flight’s settings. More UASS specifics are reported in Table 1. Since the trials aimed to verify and characterize UASS’s distribution uniformity and dispersion on a vineyard testing different flying parameter combinations, the air-injection flat fan nozzles were chosen to perform pure water spray distribution over the rows, because of better deposition performance, uniformity, and reduced drift than hollow cone ones (Wang et al., 2021).

Fig. 2
figure 2

The UASS DJI Agras T16 used for the survey equipped with the dedicated spraying system connected to pairs of nozzles positioned under four propellers out of six at the sides of the UASS

Full size image
Table 1 Agras T16 UASS technical features provided by the producing company DJI (DJI Technology Co., Ltd, Shenzhen, China)
Full size table

DJI Agras T16 automatic flights were performed using the SmartFarm app (DJI Technology Co., Ltd, Shenzhen, China), also operated to set the flight and spraying parameters.

Experimental design

The experiment employed a factorial design involving flight height Above Ground Level (AGL) (H), flying speed (S), and the UASS over-row (OR) and inter-row (IR) overflight positions with two levels per factor (Table 2), for a total of eight treatments repeated three times (Table 3). The UASS flew OR and IR (Fig. 3) at 2 m and 2.5 m AGL at 1 m s−1 and 1.5 m s−1.

Table 2 The three factors considered in the trials and the corresponding levels. Each treatment was repeated three times
Full size table
Table 3 The eight treatments included in the experiment and the respective factors’ values
Full size table
Fig. 3
figure 3

The UASS during the spraying tests in OR (a) and IR (b) treatments. The distance of the propellers positioned at opposite sides of the UASS matches the inter-row width, positioning first the UASS’s nozzles on the inter-rows in OR flights (purple rows reaching the ground) and then perpendicular to the target rows in IR flights (purple rows reaching the top part of the canopy). The target rows are highlighted by the 3D blue transparent shapes

Full size image

The OR treatments’ target was the row the UASS was flying over, with the two adjacent rows used to evaluate the off-target dispersion, for a total of three rows. Instead, the IR flights had as target the two rows surrounding the inter-row area overflown by the UASS, with the two external rows used to evaluate the off-target dispersion, for a total of four rows. Figure 4 reports the two flying paths with the sampling points’ position consisting of WSPs (7.6 cm × 2.6 cm; 19.86 cm2, Syngenta Crop Protection AG, Basel, Switzerland) located along three deposition lines spaced 20 m from each other to collect droplets. Test area boundaries were marked visibly in the field using a red tape. A 20 m preparation zone was defined to allow the UASS to reach the aimed speed and pump distribution pressure before entering the test zone.

Fig. 4
figure 4

A graphical representation obtained using the SVG software Inkscape 1.2.2. (Software Freedom Conservancy) of OR (a) and IR (b) treatments seen from above and a detailed disposition of the WSPs inside the canopy and on the inter-row area in each deposition line

Full size image

For each row’s sampling point three WSPs were hooked to the trellis at 0.3 m, 0.6 m, and 1 m AGL to measure the PPP penetration inside the canopy (Fig. 5a and 9a). Four WSPs for the OR treatments and six WSPs for the IR treatments for each deposition line, were placed on the ground in the inter-rows (2 WSPs per inter-row) over wooden dedicated supports to evaluate the off-target PPP ground deposition (Fig. 4 and 5c). All the WSPs were manually positioned by the operators using latex gloves to prevent stains on the WSPs sensitive to water and humidity (Fig. 5a and b).

Fig. 5
figure 5

A sprayed WSP clipped to the trellis system (a) by an operator wearing latex gloves during the WSPs positioning operation (b) and a WSP located in the inter-row on the wood made support to catch the ground off-target droplets

Full size image

The treatments were performed with a wind speed lower than 3 m/s direction N, NW, parallel to the rows, with a 1.3 l min−1 to 1.8 l min−1 average pump flow velocity values.

Weather condition during the test

Weather conditions were monitored by a weather station located Sud east at the border of the field 15 m off the flight path, close to the piloting zone, at 1.5 m AGL. The weather requirements (suggested by the ISO 23117-1:2023) necessary to perform the aerial distribution were:

  • Ambient dry temperature between 10 °C and 35 °C.

  • Relative humidity under 70%.

  • Mean wind speed under 2.0 m s−1.

Test procedure

The following test procedure was adopted:

  • The flight centerline path, the test area, depositions lines, and the preparation zone were delineated using a red tape.

  • The WSPs were placed by fixing them at three heights on the trellis (see Sect. “Experimental design” for further details) using clothespins and on the inter-row at the ground level over the wood made supports before each treatment flight.

  • Each treatment was repeated 3 times on the three deposition lines along the UASS’s route.

  • At each flight end, the WSPs were collected using latex gloves and tweezers, and individually placed into numbered and sealed plastic bag to avoid any contamination and humidity infiltration (Fig. 6).

  • Each WSP was then digitalized using an office scanner and analyzed using a script to extrapolate spray features, as explained in the following section.

Fig. 6
figure 6

WSPs placed separately into sealed plastic bag to avoid any contamination and humidity infiltration (a), and an operator during the collection process (b)

Full size image

Automatic droplets detector development

The first step to perform the automatic droplets detection was scanning the WSPs with 1200 dpi maximum resolution office scanner (Epson L3250). All the WSPs hooked to the trellis system belonging to the same deposition line, were scanned in the same image (9 WSPs for OR and 12 for IR operations). The same procedure was applied to the WSPs located on the ground support (4 for OR and 6 for IR operations), as reported in Fig. 7.

Fig. 7
figure 7

A full WSPs deposition line divided in canopy and ground samples collected after OR (a) and IR (b) flights. The progressive number represents the row’s number during the experiment, while the H, M, L, and X letters respectively represent the high, medium, low position on the canopy, and the samples located over the ground in the inter-rows

Full size image

The WSPs used for the test were then analyzed through a MATLAB R2021b code, able to analyze the WSPs images. Further details about the automatic droplet detector development are described in (Ghiani et al., 2020). WSP’s droplets covered area (cm2), droplets per cm2, and NMD (the diameter corresponding to the 50th percentile of droplet diameter distribution, 50% of droplets diameter is below NMD, and 50% is above NMD) were used to evaluate the UASS distribution performance.

The script, in addition to the variables considered for the survey, can provide the following WSP features:

  • Droplets number (n).

  • Volume median Diameter (VMD). The diameter corresponds to the 50th percentile of droplet volume distribution (50% of droplet volume is below VMD, and 50% is above VMD). The VMD10 and VMD90 were also calculated.

  • The distribution uniformity index CH obtained by dividing VMD/NMD.

  • The average diameter (µm) derived from the average of the droplet diameters in each WSP.

  • Diameter STD. The standard deviation related to the mean diameter.

All the previous features related to the stain size (diameter and volume) were measured considering the USDA-ARS System droplets spread factor (Eq. 1) reported by Ahmad et al. (2020) and Hoffmann and Hewitt (2005).

$$\text{D}=(0.53549306 \times \text{ d}) - (0.000084839 \times {\text{d}}^{2})$$
(1)

where “D” represents the droplet diameter (μm) and “d” the stain diameter on the WSP (μm).

Statistic analysis for spray performance evaluation

Spray features and UASS performance were evaluated using the MATLAB-based WSPs analysis and then processed using Python (Python Software Foundation) scripts through Pandas, Statsmodel, and Matplotlib packages, respectively for data frame manipulation, statistical analysis, and plotting. The statistical analysis was prompted by the works of Biglia et al. (2022); Chen et al. (2020); Sarri et al. (2019); Wang et al., (2021, 2022).

Droplet deposition distribution in target rows and the drift effect evaluation

The covered area, the droplets number per cm2, and the NMD variables were plotted separately for each sampling height (high, medium, and low) by averaging the three repetitions (corresponding to each deposition line) of each treatment and separating the target and off-target rows results. The same procedure was applied to the ground samples by averaging all the WSPs results of the same treatment. A three-way analysis of variance (ANOVA) test was performed to detect the effect of flight speed, flight height, and UAS flight position factors (Table 2) and their combination on each of the three variables, differentiating target rows, off-target rows, and ground loss. The significance of statistical computations for ANOVA and Tukey test was considered for P < 0.05 unless otherwise stated. Data distribution and homoscedasticity were respectively tested through Shapiro–Wilk and Levene’s test. If the ANOVA performed on the fitted model showed a significant factors effect, the Tukey’s test for multiple comparisons was performed to compare mean values of parameters between treatments, and results showed inside the respective bar plots.

Droplet deposition penetration measurement

The droplets deposition penetration gives information regarding the efficiency of the treatment to reach the lower part of the row where the protection products are more needed.

In order to characterize droplets PR at each collection point, the ratio of the lower WSP’s droplets covered area (CAlow) to the sum of the upper (CAup) and mid (CAmid) WSPs’ droplets covered area was used, as explained by the Eq. 2 (Wang et al., 2022):

$$PR=\frac{{CA}_{low}}{{CA}_{up}+{CA}_{mid}}\times 100\%$$
(2)

Results

The results are presented in two main sections: the first reports the distribution of target and off-target rows droplet deposition for the considered variables (covered area, droplets per cm2, and NMD) at each height position of the WSPs, while the second showcases the PR results through the canopy.

Covered area

Covered area results tent to decrease from the higher to the lower canopy section, with treatments A and B gaining the largest covered area. Treatment H got a similar result in the lower position, impacting on the PR discussed below in Sect. 3.6 and plotted in Fig. 16.

The higher covered area in treatments A and B could be related to the position of the UASS perpendicular to the row and the higher flight altitude AGL, which led a larger volume of product to reach the upper-mid part of the canopy. Treatments C and D, instead, due to the lower flight altitude and the location of the nozzles perpendicular to the inter-rows (Fig. 3a), gained a lower covered area, leading a conspicuous volume of product to hit the off-target rows. In fact, despite the UASS frame was positioned over the target row, due to the large diameter of the UASS and the derived location of the nozzles over the inter-rows, treatments C and D flight settings caused a lower covered area on the target row compared to A and B.

Treatment H results at low sampling height (Fig. 8c) could derive from the combined action of flight height and speed. The first one helped reduce the off-target dispersion (larger in treatments E and F, specifically at low positions), while the second one, thanks to the higher spinning rotation of the propellers, caused a bigger canopy opening, allowing the droplets to reach the lower position. It is essential to emphasize the importance of finding the proper flight setting that allows a homogeneous distribution of product throughout the canopy height, with a specific focus on the mid-low parts, known to be the main canopy’s infection sites, which require more than others to be treated and limiting an over-coverage of the highest part.

Fig. 8
figure 8

The bar charts containing the mean covered area results and Tukey’s test group division for target and off-target rows divided for each treatment at high (a), medium (b), and low (c) WSPs heights

Full size image

The results obtained in the IR treatments (E, F, G, and H) could be attributed to the same reason described above, where, this time, the UASS propellers, and consequentially the nozzles located below, were positioned over the two target rows (Fig. 9). This condition provoked a generally smaller covered area on the two target rows at the three sampling heights, a large covering at the lower sampling height (probably derived by the action of the propellers that opened the canopy), and a lower dispersion on off-target rows compared to low flight height over-row treatments (C and D). These differences were also emphasized by the longer distance of the off-target rows from the UASS in IR treatments compared to the OR ones (Fig. 4) and the different generated turbulence between the rows.

Fig. 9
figure 9

A graphic representation obtained using the SVG software Inkscape 1.2.2. of the influence of flight height AGL and position on the spraying performance. The nozzles’ position over the row in IR treatments (a) limited the product off-target dispersion and the propellers facilitated droplets penetration through the canopy represented by the purple arrows. In OR treatments the position of the UASS over the row provoked a larger ground dispersion in the inter-rows as highlighted by the vertical arrows

Full size image

In Fig. 10, treatments B and H WSPs, related only to the target rows, are reported. Treatment B showed an over-coverage, higher than the 30% as described by Y. Chen et al. (2013), in the high-mid section of the canopy (Fig. 10a), which could result in an inefficient PPP application, a waste of product and could cause the dripping on the lower leaves and on the ground. On the contrary, the H treatment showed a smaller coverage at the upper part of the canopy but an ideal one at the lower part. H treatment also showed a drift effect probably caused by wind gusts, which pushed the droplets on the left-side target row during the spraying operation (Fig. 10b). This effect could compromise treatment efficacy since the mid-low part of the canopy was reached by a few droplets. Similar behavior was detected in all treatments and cannot be completely avoided. To limit the wind effect and the influence of canopy architecture on the results, the research team decided to average the three repetitions, as described in materials and methods section.

Fig. 10
figure 10

The high, medium, and low sampling positions WSPs for treatment B (a) and treatment H (b) on target rows, and the corresponding droplets diameter cumulative curves and bar plots for each WSP, three for treatment B (c) and 6 for treatment H (d), crossed by a red horizontal line in correspondence of the NMD value, and dashed vertical lines to help detecting the minimum and maximum NDM of each treatment, with the minimum one of H treatment, obtained by the lowest WSP, colored in red to highlight the opposite behavior respect to treatment B gained by the highest WSP. Treatment B, obtained an over-coverage in the high-mid heights of the canopy, opposed to the H treatment, object of an uneven distribution due to wind influence. Same results are confirmed by the bar plots of treatment B (e) and H (f)

Full size image

According to the three-way ANOVA (Table 4), the factors flight height and UASS flight position had a significant influence on the area covered in the upper part of the canopy, also showing a significant interaction effect. The two factors probably caused treatments A and B to get a larger WSPs’ covered area, specifically at the higher sampling position. Since flying speed was not considered a significant predictor of the covered area, spraying operations could be performed at a higher speed to cover larger surfaces in less time.

Table 4 Covered area 3-way ANOVA results divided by high, medium, and low WSP’s canopy height for target and off-target rows
Full size table

The 3-ways ANOVA table also suggested that the two mentioned factors, singularly and combined, influenced the off-target results at the lowest off-target sampling position. As shown in Fig. 8c, the treatments C, D, E, and F reported a higher off-target dispersion. For the OR treatments (C and D), this effect was probably caused by the position of the nozzles on the two inter-rows alongside the target row, which helped disperse the product on the inter-rows and the lower part of the off-target rows. The E and F treatments provoked a smaller off-target dispersion probably because of the different nozzles’ position on the target rows (perpendicular to them) and the greater distance from the off-target rows. The higher position of treatments E and F compared to G and H probably provoked a drift effect due to the wind effect, which pushed the droplets far from the monitored off-target rows.

Averaged ground dispersion results reported in Fig. 11a showed a general difference between OR treatments, characterized by a higher covered area (most of all treatment B), and IR ones. Such a result suggests a better performance for IR treatments, which showed a lower ground dispersion of product, especially for treatments E and F, and a higher one for treatments G and H. This behavior could be related to the higher flight performed in treatments E and F, which involved an increased droplets’ exposition to the wind action (especially the smaller ones), which influenced distribution results and provoked value fluctuation.

Fig. 11
figure 11

Covered area results with Tukey’s test grouping divided by treatments for ground WSPs (a) and a specific focus on IR treatments’ ground dispersion (b), larger in central inter-row of treatments G and H

Full size image

The three-way ANOVA shows how the speed and UASS flight position (singularly and combined) influenced the ground results (Table 5). As described above, the position of the nozzles over the target rows in the IR flying treatments helped reach good covering results and limit the quantity of product that reached the ground, differently from the other treatments, characterized by the nozzles spraying perpendicular to the inter-row. The flight speed and the consequent higher propellers’ speed rotation could have provoked an increased downward flow that pushed the droplets onto the ground, as described by Wen et al. (2019), visible in treatments B and D.

Table 5 Covered area 3-way ANOVA results for ground WSPs
Full size table

In the OR treatments, the treated inter-rows were only two, where the UASS sprayed identically through each nozzle positioned above (Fig. 4a). Differently, in IR flights, the sprayed inter-rows were three (Fig. 4b): a central inter-row sprayed by both UASS’s side nozzles and two side inter-rows, respectively sprayed from half part of the nozzles. Figure 11b shows the different covered area results obtained in the central inter-row (over which the UASS was flying) and the side inter-rows. While treatments E and F showed almost homogeneous results throughout the inter-rows, treatments G and H showed higher coverage percentages on the central inter-row. Such result determines the influence of the flight height on the downward direction of the droplets in IR flights, limiting the wind effect on the spray jet, protected by the target rows and how, in general, the central row receives an increased volume of PPP.

Droplets per cm2

As highlighted in the previous chapter for the covered area, the higher WSP position shows no statistical differences between the treatments related to the number of droplets per cm2 for both target and off-target rows (Fig. 12a). Noticeable differences occur in the mid position, where the treatments A and B differ for the droplet number. This behavior disappears in the lowest position where treatment B decreases consistently and treatment H increases. This specific result adds information to the ones displayed in Fig. 8c, where treatment H gained a similar coverage to treatment B, showing how, even if the two treatments reached a comparable coverage percentage, treatment B showed a lower droplet number than treatment H. Therefore, the droplets in treatment H are smaller than in treatment B, as demonstrated in Fig. 14c. This result is crucial because it emphasizes the importance of choosing the proper flight set up before the operation starts to obtain a homogeneous treatment through the canopy and limit excessive leave coverage.

Fig. 12
figure 12

The bar charts containing the mean droplets per cm2 results and Tukey’s test group division for target and off-target rows divided for each treatment at high (a), medium (b), and low (c) WSPs heights

Full size image

The three factors had no influences on the target rows’ results (Table 6), except the UASS position factor at the mid-height. The F treatment reported different results from other treatments, showing a reduced droplet number per cm2 and a limited covered area (Fig. 12b).

Table 6 Droplets per cm2 3-way ANOVA results divided by high, medium, and low WSP’s canopy height for target and off-target rows
Full size table

Interesting results are reported for the off-target rows, where the ANOVA table (Table 6) showed a significative effect of the height and the combined effect of flight height and UASS position on the number of droplets per cm2. Treatments A and B appear different from other treatments, and the reason could be mainly associated with the flight height, which exposed more of the droplets to the wind action and drifted them further over the placed off-target WSPs. The fact that treatments E and F, performed at higher heights, did not show any difference with treatments G and H could be related to the influence of the rows, which acted as barriers, and the distance of WSPs positioned further than OR treatments.

Figure 13a shows a similar trend to that shown in Fig. 11a for the covered area variable, providing further information on ground dispersion, where OR treatments achieved a higher number of drops per cm2 due to the UASS flight position (Table 7). Larger covered areas are represented by a smaller droplets number, confirming the hypothesis that larger covers result from larger droplet diameters. In addition, Fig. 13b shows a similar trend to Fig. 11b, where the results of the central interleaves for treatments G and H show a discrepancy with the lateral interleaves.

Fig. 13
figure 13

droplets per cm2 results with Tukey’s test grouping divided by treatments for ground WSPs (a) and a specific focus on IR treatments’ ground dispersion (b), larger in central inter-row of treatments G and H

Full size image
Table 7 Droplets per cm2 3-way ANOVA results for ground WSPs
Full size table

Number mean diameter (NMD)

The results displayed in Fig. 14 show almost an identical pattern through the canopy with no treatment difference, a part for the lower position (Fig. 14c) where treatment B shows a bigger NMD compared to treatment F (which showed instead a comparable covered area in Fig. 8c), proving that the first has a larger covered area derived from bigger droplets’ diameter.

Fig. 14
figure 14

The bar charts containing the mean NMD results and Tukey’s test group division for target and off-target rows divided for each treatment at high (a), medium (b), and low (c) WSPs heights

Full size image

Treatments A and B showed a similar NMD to other treatments in high and mid positions, where they reached a significantly bigger covered area (Fig. 8a and b). After a visual analysis of the WSPs, treatments A and B droplets showed a general larger diameter than the other treatments (an example in Fig. 10), especially if high-position treatment B is compared to high-position treatment F (characterized by a reduced covered area). Such a result could be influenced by two factors: the first is related to a high number of thin droplets in treatment B that shift the NMD to a smaller value (Fig. 10c and d), and the second to how the image analysis script works. Specifically, Fig. 10c and e show that high-mid position WSPs are covered with droplets smaller than 50 µm, so extremely thin, and very few big and merged droplets. Figure 10d and f, on the contrary, show how the number of thinner droplets increase moving from the higher to the lower positions (red dashed vertical line) in a smaller interval (except for the WSP 3 L results, probably influenced by the drift effect). Starting with the assumption that most non-regular shape droplets (not spherical) could derive from two or more smaller merged droplets, it automatically performs the division, provoking the NMD reduction. Bar plots reported in Fig. 10e and f show how in general treatment H got a bigger number of smaller droplets, specifically in the lower layer, than treatment B. ANOVA test (Table 8) confirmed the influence of flight height and position on off-target results, specifically on lower WSP position, already defined as the priority target of most PPP spray operations.

Table 8 NMD 3-way ANOVA results divided by high, medium, and low WSP’s canopy height for target and off-target rows
Full size table

Figure 15a did not show significant differences for ground WSPs results, except for treatment E, which showed almost a half NMD value than F treatments. Based on ANOVA results (Table 9), this difference could be attributed to the higher UASS flight speed and propellers rotation velocity that provoked a thinner production of droplets.

Fig. 15
figure 15

NMD results with Tukey’s test grouping divided by treatments for ground WSPs (a) and a specific focus on IR treatments’ ground dispersion (b), larger in central inter-row of treatments F, G and H

Full size image
Table 9 NMD 3-way ANOVA results of ground WSPs
Full size table

Differently from the covered area and droplets per cm2 variables, NMD did not show any difference between central and side inter-rows for treatments E, F, G, and H (Fig. 15b), showing how such variable is uniformly distributed and is not affected by the UASS position during the spraying operations.

Penetration rate

The PR is highly connected to the covered area variable, as shown in Formula 2. For this reason, high PR was obtained because of the ratio between lower sampling point’s results characterized by an expanded covered area value and a small high and mid-sampling points sum. The ANOVA test found no significant effect of flight settings on PR (Table 10), although different treatment performance can be identified.

Table 10 PR target rows 3-way ANOVA
Full size table

For instance, even if treatment B showed an increased coverage in all three sampling heights, it showed a lower PR compared to treatment H, which instead gained a reduced coverage in the higher parts of the canopy and a large coverage in the lower parts (Fig. 10). The same interpretation could be transferred to off-target results in Fig. 16, which showed reduced PR than target rows. It is essential to stress that the mid-low part of the canopy is the one that, more than others, requires adequate PPP coverage, and for this reason, treatments with higher PR are preferred. Also, for off-target rows, WSPs in lower positions were hit more than the higher ones but, if visually inspected, the results derived from almost empty WSPs, meaning that few droplets reached the off-target rows’ lower part and even less in the upper parts, with a crucial product saving.

Fig. 16
figure 16

The mean PR results bar chart for target and off-target rows divided by treatments

Full size image

Discussion

Field trials insights

Flight position was the factor that, more than others, influenced the results, especially covered area, and droplets per cm2 variables for target, off-target, and ground WSPs. The treatments that gained a higher covered area were treatments A, B, and H. The first two showed an apparent homogeneous distribution of the canopy but the upper WSPs were over-covered, a sign of a less efficient application (Fig. 10) than that achieved by the H-treatment. This behavior was in agreement with the experiments of Wen et al. (2019), where lower altitude flights with a speed of around 2 m/s were also identified as optimal for increased foliage coverage and drift reduction. Off-target rows showed similar results for these three treatments, while ground WSPs revealed a larger dispersion of PPP for OR treatments in terms of covered area and droplets per cm2.

Even if the ANOVA tables mainly showed a significant impact of flight position and height on the high and mid-levels, it confirms the importance of the position of the UASS to limit spray over-dosage on leaves and how flight height also played a crucial role in OR treatments, causing a higher coverage where the UASS flew higher (treatments A and B), and also in IR treatments where a lower height helped increasing the spray coverage through the canopy height.

Flight position and height appeared strictly connected because in OR treatments the bigger height helped improve the covered area of treatments A and B, while in IR flight position, the lower height helped reduce the dispersion of product and increase the covered area and the PR (as in treatment H). As a result of the nadiral nozzles’ position over the row, the turbulence effect generated by the propellers helped to open the canopy to facilitate the PPP penetration, specifically for treatment H, where the rotary speed of the propellers was higher (Martinez-Guanter et al., 2020).

As described above, although with a reduced results impact, also flight position and speed were strictly connected. The flight speed and the consequent higher propellers’ speed rotation could have provoked an increased downward flow (Bloise et al., 2022; Wang et al., 2021). For OR treatments, the higher speed pushed the droplets on the ground (Biglia et al., 2022), increasing the dispersion (this effect is most visible in treatments B and D), while IR flights generated an efficient product penetration inside the canopy.

All treatments showed similar results in terms of product dispersion, but OR treatments C and D showed an increased influence in the lower positions, probably due to the lower flight height and position, which, more than other treatments, facilitated ground and side-row waste. In addition, the higher flight altitude of treatments A and B could have provoked droplets to drift further over the sampled rows, providing only a partial representation of the total dispersion, as suggested by Wang et al. (2019).

It is essential to highlight that the treatments that reached a higher PR are preferable (A, B, and H) since most current PPP treatments applied to grapevine have the lower part of the canopy, where the grapes are located, as a target. Based on the results, treatment H seems to be the most appropriate one since it provided good PR results, it did not provoke an over-dosage effect in the upper part of the canopy, the ground dispersion is lower than treatments A and B, and it allows to treat two rows simultaneously and at higher speed, with crucial performance benefits.

The script proposed in this work to analyze the spraying deposition on WSPs played a key role in obtaining reliable results and boosting the analysis process. Other WSP image analysis software was considered during the experiment design process (Ferguson et al., 2016; Sies et al., 2017; Zhu et al., 2011), nevertheless, the research team opted for a proprietary one to better adapt it to UASS spraying survey scenarios. Even if only covered area, droplets per cm2, and NMD were considered in the present work, the script allows the obtainment of multiple variables, as described in the materials and methods section, including the spread factor. Despite the information provided by the proposed script, the research team decided to focus on those that were frequently reported in other UASS studies and, more concisely, would have helped define the spraying performance with a reduced measurement error like the Volume Mean Diameter (VMD) that provided questionable results. As shown below in Fig. 17, treatment B gained an excessive VMD at the high and medium canopy sections compared to other treatments, especially if compared to treatment A, which showed almost identical results for other variables. This behavior could be related to the excessive coverage of treatment B WSPs at those heights, which did not allow the software to provide a reliable estimation. Such deductions are supported by the lower canopy section results (Fig. 17c), where all treatments, characterized by WSPs containing fewer and scattered droplets, showed similar results, confirming the limit of the proposed script to evaluate VMD variable for over-covered WSPs.

Fig. 17
figure 17

The bar charts containing the mean VMD results for target and off-target rows divided for each treatment at high (a), medium (b), and low (c) WSPs heights

Full size image

Future trials will be designed to compare the image analysis results (especially the spread factor ones) with more sophisticated laser-based droplet analysis. However, it is undeniable how the WSPs analysis methodology represents a more practical, customizable, and effective way to measure UASS PPP deposition on leaves deployable by every farmer and research group with limited costs and time.

Final considerations and future perspectives

The main goal of this work was not to find the best flight setting to perform UASS PPP distribution over vineyards but to analyze the effect of flight position, height, and speed factors on the treatment performance, showing how the obtained results are influenced by the UASS model and the vineyard characteristics. These findings must be traced back to the specific diameter of the UASS (which in the presented trials helped reach both rows simultaneously in IR treatments), the number, model, position, and angle orientation of the nozzles under the propellers, generally used for herbaceous crops in this configuration. Specific nozzles for UASS arboreal operation, with a smaller spray angle, differently oriented would have taken to probably different and better results. In the same way, the vineyard’s plant spacing along and between the rows would have affected the results.

Further studies are required in vineyard UASS application to analyze the effect of speed on spraying performance and to learn what is the limit over which the downward movement that improves droplet penetration ends and their dispersion begins. Turbulences represented the unknown factor that could have influenced the results. Appropriate sensors should be placed inside and outside the canopy and on the inter-row to evaluate the effect on the droplets’ penetration and dispersion.

Strictly correlated with the turbulence is the shape and configuration of the UASS. It is essential to define what is the most appropriate configuration (size, diameter, number of propellers and their orientation, etc.) to gain the best performances (Wang et al., 2022) in viticultural scenario. As with every application scenario, UASS PPP application should benefit from specific nozzles designed to reduce product dispersion and modulate droplet spectra depending on the chosen protection product and the environmental requirements. Since the experiments were conducted only on one (OR treatments) and two target rows (IR treatments), it is crucial to evaluate the spray coverage derived from treating an extended surface and give better details about the overlapping results of parallel applications along the rows.

Further research is needed to provide the necessary guidelines for aerial distribution by UASS, to limit the dispersion of pesticides (according to ISO 22866), to evaluate the economic benefit (Morales-Rodríguez et al., 2022) and to boost the aerial spraying sector in agriculture, which is restricted to specific situations where terrestrial treatments are not feasible (European Parliament 2009). These restrictions, limited to the European Union, have repercussions on technological progress and the implementation of this technology in specific scenarios that could benefit from this approach, such as viticulture traditionally located in Europe, and limit the research development.

Conclusions

The work showed the flight position, height, and speed factors effect on PPP aerial distribution from UAS on vineyards, and how the obtained results were influenced by the UASS model and the vineyard characteristics. 2.0 m height IR flight position showed the best performance, with the lowest off-target and ground dispersion, demonstrating how flying over the inter-row, with nozzles right above the canopy, represents the best setting to facilitate droplets penetration inside the canopy, reaching the lowest parts, generally more affected from disease, and reduce off-target dispersion. Flight speed did not significantly affect the performance, apart the cases where a faster propeller rotation helped open the canopy and facilitate the droplet penetration. This result represents an essential logistic information, enabling the performance of shorter flight missions and increasing the general efficiency of the system. The droplet analysis procedure showed high potential and it could help boost the UAS spraying scenarios in vineyards. Further research is required to evaluate the aerial PPP distribution from UASs effectiveness and assess the environmental impact in agricultural scenarios. This knowledge is essential to boost the implementation of such technologies, specifically in those countries where aerial treatments are not still allowed.