The field experiments were conducted at South Block, Agricultural Research Station, Bhavanisagar (11°29’ N, 77°08’ E) (study 1) and Wetland Station, Coimbatore (11°54’ N, 76°56’ E) (study 2), Tamil Nadu Agricultural University, Coimbatore, India. The description of soil and weather parameters for the experimental sites are provided in Tables 1 , 2 .

The experimental field was plowed using a tractor-drawn disc harrow, followed by puddling using a tractor-drawn cage wheel. The puddled field was then leveled using a wooden plank. The rice variety used for the experiment was ADT 43. The rice seeds were soaked in water for a duration of 10 h, followed by incubation in darkness for 24 h to promote seed sprouting. The sprouted seeds were sown on the puddled soil using a drum seeder with a row spacing of 0.2 m. The seeds were sown on May 19, 2022 (Study 1) and May 23, 2022 (Study 2), and a seed rate of 60 kg ha^-1 was used. Water was drained from the field one day after sowing, up to a week to encourage early and uniform germination. After the establishment of seedlings, the water depth was maintained at 5 cm. The recommended dose of fertilizer, comprising 150:50:50 kg ha^-1 of nitrogen, phosphorus, and potassium, was applied in the form of urea, single superphosphate, and muriate of potash. Prior to sowing, 100% of the phosphorus and 25% each of the nitrogen and potassium were applied as a basal application, while the remaining 75% of the nitrogen and potassium were applied in three equal splits at 21 DAS, panicle initiation, and flowering.

A hexacopter UAV (Kisan drone V.2, Garuda Aerospace Pvt Ltd, Chennai, India) equipped with various spraying components ( Figure 1A ) was utilized for the study. The UAV sprayer was equipped with 180 kV Brushless Direct Current (BLDC) motors, 0.57 m foldable propeller, two Lipo batteries, a flight controller, receiver, GPS unit, and a pesticide tank with a capacity of 10 L and four nozzles with a spacing of 0.70 m. The BLDC motor pump was used to pressurize the spray liquid with the pressure of 3.4 kg cm^-2. The flight planner was employed to control the flying height and speed, while the flow rate of the UAV sprayer was adjusted to achieve the desired application flow rate, ranging from 100% to 75%. Prior to the flight, pre-flight calibration of the UAV spraying system was conducted. In contrast, the battery-operated knapsack manual sprayer utilized in the study consisted of a 16 L tank to hold the spray solution, a pump, a filter, a flow control valve, a delivery hose, and a spray gun with nozzle with the pressure of 2.1 kg cm^-2. Detailed information on different components of both the knapsack manual spraying (KMS) (KK-BBS-199, KisanKraft Limited, India as shown in Figure 1B ) and UAV spraying systems are provided in Table 3 .

The application rate was determined following the standard set by the American Society of Agricultural and Biological Engineers standard (ASABE) (S386.2, 2018). The travel speed, effective spray width, and discharge rate values were measured, and the application rate was calculated using the following equation:

Where,

R= Application rate, L ha^-1

Q= Output rate, L min^-1

K= Constant, 600

S= Travel speed, km h^-1

W= Effective spray width, m

A systemic early post-emergence herbicide bispyribac-sodium (Nominee Gold, 10% SC, PI Industries Limited, Udaipur, India) was applied at a rate of 35 g a.i ha^-1. The herbicide was applied as an early post-emergence treatment at 2-to 3-leaf stage of the rice plant using both KMS and UAV ( Figure 2A ). The treatments comprised spraying with the UAV at the rate of 30 and 45 L ha^-1, along with KMS at a rate of 500 L ha^-1, with a weedy check serving as a control. The randomized complete block design was used to arrange the experimental units with a plot size of 40 × 16 m and three replications for each treatment. The operational parameters for the UAV and KMS are presented in Table 4 . During the herbicide application, various meteorological parameters such as temperature, relative humidity, and rainfall were obtained from the meteorological observatory ( Table 5 ). A portable hand-held anemometer (LUTRON AM 4202, Vane type, range: 0.4 to 30.0 m s^-1) was used to measure the continuous wind velocity during the herbicide application (Yang et al., 2018).

Water-sensitive papers (WSP) with the dimension of 2.6 cm x 7.6 cm (Spray Check, SC-20301, USA) were clipped to an iron platform (0.04 m × 0.30 m) to ensure that they remained flat during the experiment ( Figure 2B ) and placed above to the crop canopy within 2 cm. A total of eight samplers were placed randomly within the experimental plots. The first sampler was deployed at a distance of 4 m from the border row. Subsequent samplers were placed randomly in different lines within the row ( Figure 3 ) and samplers were collected and replaced between the flights.

Following the herbicide application, the WSPs were left to dry completely. Once dried, they were carefully placed in zip-tie bags and transported to the laboratory. Afterwards, DropletScan software (USDA, Wooster, OH, USA) was used to measure droplet diameter, droplet density, and coverage area on the WSPs (Whitney and Gardisser, 2003).

In order to characterize the droplet distribution between the various collection points, the coefficient of variation (CV) for droplet deposition at each collection point was computed. Smaller CV values result in more uniform droplet deposition and better efficacy (Zhang et al., 2020).

Weed density (number of weeds m^-2) and dry weight (g m^-2) of the weeds were recorded in each experimental plot at 40 DAS. Weed density was observed in four quadrants (0.5 × 0.5 m) placed in each plot. To determine weed dry weight, the weeds were removed at ground level and allowed to dry in the shade for six days. Subsequently, they were oven-dried at a temperature of 78 ± 2°C until a constant weight was attained (Jabran et al., 2012). Weed control efficiency was calculated as per the procedures given by Mani et al. (1973) and expressed in percentage.

Due to the high variability observed in the data recorded from the WSP for spray deposition, droplet density, and area coverage, transformations were applied to stabilize the wide variation and meet the assumption of normality. A variable refers to traits, either quantitative or qualitative, that vary across observations within the same treatment. The data expressed as a percentage of area coverage was subjected to arcsine transformation ( arcsin x / 100 ) . The other parameters obtained from the WSPs data were transformed using the log transformation log(x+1). Additionally, the weed density and weed dry weight data were transformed using square root transformation   ( x + 0.5 ) ,   to stabilize the wide variation and meet the normality assumptions. The transformed data were used to perform the Shapiro-wilk normality test and Bartlett test for analyzing homogeneity of variance. Then, the data were analyzed by using one-way ANOVA and multiple comparisons were performed using Tukey’s Honest significant difference test at 5% probability using IBM SPSS statistics software version 26.0 (Steel et al., 1997).

The major weed flora found in the experimental fields comprised of Leptochloa chinensis, Echinochloa colona, Cyperus difformis, Cyperus iria, Marsilea quadrifolia, Bergia capensis, and Ludwigia parviflora. In addition, Sphenoclea zeylanica and Monochoria vaginalis were also observed in study 1 experimental plot. The application of herbicide using both UAV and KMS sprayers resulted in significantly lower total weed density and weed dry weight compared to the weedy check (p<0.001 and p=0.000, respectively). The Shapiro-Wilk normality test confirmed that the data sets were normally distributed, while the Bartlett test demonstrated that the variance was homogeneous across all data sets without any bias. The results for normality and homogeneity of variance are presented in Supplementary Table 1 . The total weed density recorded in the test plot were 75.5 ± 13.36, 63.5 ± 19.80, 47.5 ± 18.38, and 220.5 ± 27.04 numbers m^-2 in study 1 and 83.5 ± 16.26, 88.0 ± 17.97, 62.5 ± 27.58, and 201.5 ± 31.82 numbers m^-2 in study 2 for UAV sprayer at 30, 45 L ha^-1, KMS treatments, and weedy check, respectively ( Figure 7A ). Numerically lower weed dry weight was recorded in the herbicide-treated plots using the KMS (15.17 ± 2.95 and 14.70 ± 7.32 for study 1 and study 2, respectively) ( Figure 7B ), and it was on par with UAV treatments. A higher WCE of 87.4% in study 1 and 84.7% in study 2 were recorded in the KMS. The WCE values for the UAV treatments at spray volume for 30 and 45 L ha^-1 were 85.7% and 79.7%, 86.1% and 77.0%, for study 1 and study 2, respectively.

Our study evaluated the effective spray volume for herbicidal weed management in DSR using UAV applicator. We analyzed the droplet spray parameters for UAV with two tested volumes in comparison with KMS applicator. Increasing carrier volume from 30 to 45 L ha^-1 with UAV increased the VMD value only by 4.6% in study 1 and 4.3% in study 2. Creech et al. (2015a) concluded that increasing the carrier volume from 47 to 187 L ha^-1 increased the VMD value only by 5% at constant height (1.4 m), which indicates that the droplet size of the herbicide was not dependent on carrier volume. Similarly, there was no significant difference in the spray area coverage between the UAV treatments with the tested spray volumes. This can be attributed to the fact that increasing spray volume under the same spraying system does not have much influence on the spray area coverage, especially when the weather condition is the least concern particularly environmental wind. Wind velocity changes during flight passes were relatively narrow (0.62 to 0.78 m s^-1 and 0.81 to 0.95 m s^-1 for study 1 and study 2, respectively), indicating minimal impact on deposition. Similar results were reported by Martin et al. (2020) in an experiment on UAV application of spray mixture on palmer amaranth and morning glory, where a spray volume of 18.7 and 37.4 L ha^-1 yielded similar area coverage under the same spraying system.

The number of spray droplets is a critical factor that affects weed control efficacy. The droplet density was not significantly different between UAV treatment with tested spray volumes. Previous studies have shown that the number of droplets tends to increase as the droplet size decreases (Adams and Hall, 1990; Bouse et al., 1992). However, smaller droplets can lead to serious drift problems (Chen et al., 2020). Additionally, increasing the spray volume did not influence the droplet density when the droplet size remained the same (Shan et al., 2021). Another possible reason could be increasing spray volume caused the droplets to overlap and resemble as a single droplet. Increasing the number of droplets can enhance contact with weeds and improve herbicide efficacy, particularly with highly concentrated herbicide formulations. Syngenta Crop Protection AG (Basel, Switzerland) recommended a minimum 30 to 40 droplets cm^-2 for post-emergence herbicide application to achieve effective weed control (Zhu et al., 2011).

The uniformity of droplet distribution can be affected by sprayer operating parameters (Qin et al., 2016) and meteorological conditions (Hussain et al., 2019). In both studies, the fluctuations of wind velocity were relatively narrow (0.62 to 0.78 m s^-1 in study 1 and 0.81 to 0.95 m s^-1 in study 2), resulting in minimal influence on droplet deposition. At the same time, UAV rotor downwash air assisted in the uniform distribution of spray droplets at lower spray volumes (Shengde et al., 2017). The higher uniformity of spray deposition observed in the UAV at 30 L ha^-1, can be attributed to ideal flight parameters such as flight height (1.5 m) and forward speed (5 m s^-1) for the autonomous navigation of the UAV sprayer. However, the efficiency of UAV was not uniform to all models, and it may vary with changes in nozzles, flight parameters, and UAV models. Researchers have reported that an increase in flight altitude and flight speed can result in insufficient and poor uniformity of droplet deposition due to drift and evaporation (Zhang et al., 2020; Hunter et al., 2020). Poor uniformity of the spray distribution can negatively affect the efficacy of herbicide application and lead to poor weed control. Flight operational parameters in this study were consistent with the result of Qin et al. (2016). In contrast, the poor uniformity recorded in the KMS treatment may be attributed to the operator’s unsteady walking speed and arm movement as well as instances of missing and overlapping operations (Wang et al., 2019). The larger droplet size and higher spray volume used in the KMS treatment resulted in higher droplet deposition on the WSP samplers. Conventional sprayers operating at low pressures produce larger droplets, which tend to deposit more quickly on target weed plants. This is due to the increased gravitational pull on larger droplets compared to smaller ones, helping to reduce both droplet drift and evaporation losses (Chen et al., 2020).

The greater weed control was achieved at a spray volume of 500 L ha^-1 using the KMS, which was comparable to the UAV treatments. The efficacy of herbicides was not influenced by carrier volumes, particularly for UAVs. This can be attributed to uniform droplet deposition on weed plants and the major role of UAV rotor downwash in penetrating droplets deep into dense crop canopy systems. Martin et al. (2020) reported that UAV-based spraying with a spray volume of 37.4 L ha^-1 resulted in a 4-fold increase in droplet deposition on the abaxial surface of weed foliage compared to a conventional sprayer with a spray volume of 140 L ha^-1. The application of systemic herbicide (bispyribac-sodium) using UAV and KMS methods significantly reduced weed density compared to untreated check, which confirms that the efficacy of herbicide was not affected by the application methods. The application of bispyribac-sodium resulted in effective weed control in all treated plots as it inhibited the biosynthesis of the enzyme acetolactate synthase, thereby arresting weed growth. Similarly, Jeevan et al. (2023b) also stated that there was no significant difference between the efficacy of systemic herbicide in UAV and conventional applications.

In both studies, the application of herbicide resulted in a significant reduction in weed dry weight under KMS, which was comparable with UAV treatments. Similar results were reported by Pranaswi et al. (2024) and confirmed that the application of systemic herbicide metribuzin on wheat through UAV resulted in similar weed control as compared to backpack application (Pranaswi et al., 2024). Similarly, Kumar et al. (2022) also reported no significant difference in weed control efficacy between low and high volume application of systemic herbicides applied through UAV and knapsack sprayers in wheat. In both studies, WCE under UAV treatments was similar to KMS, with a narrow variation of 1.7% to 5.1% WCE recorded between the KMS and UAV at 30 L ha^-1. This confirms that the application method and carrier volume did not affect herbicide efficacy. Chen et al. (2019) reported that herbicide mixtures (isoproturon + clodinafop-propargyl + mesosulfuron) applied through UAV on wheat resulted in similar WCE compared to a conventional knapsack sprayer. However, the effects of spray volume on herbicide efficacy can vary depending on the physicochemical properties of the herbicides, such as solubility, droplet distribution, retention capacity, and chemical stability. For instance, contact herbicides require more carrier volume than systemic herbicides for UAV application to achieve desirable droplet distribution and retention on target weed (Shan et al., 2021). Creech et al. (2015b) reported that the carrier volume requirement to improve efficacy for different herbicides (glyphosate, glufosinate, lactofen, fluazifop-P, and 2,4-D) was not uniform in field and greenhouse experiments, indicating the need to consider specific application environments when assessing the effects of spray volume on herbicidal efficacy.