Monitoring and risk analysis of residual pesticides drifted by unmanned aerial spraying

This study aimed to investigate the residual characteristics of pesticides drifted by unmanned aerial spray according to buffer strip, windbreak, and morphological characteristics of non-target crops, suggest prevention for drift reduction, and finally conduct a risk analysis on pesticides exceeding the maximum residue limit (MRL) or uniform level (0.01 mg/kg) of the positive list system (PLS). Non-target crops were collected around the aerial sprayed area (paddy rice) in Boryeong, Seocheon, and Pyeongtaek after UAV spray. When pesticides were detected in more than three samples, Duncan’s multiple range test was performed. In cases where pesticides were detected in only two samples, an independent sample t-test was conducted (p < 0.05). The drift rate of pesticides tends to decrease by up to 100% as the buffer distance from aerial sprayed area increases or when a windbreak, such as maize, is present between two locations. Thus, the reduction of drifted pesticides could be effective if both factors were applied near the UAV spray area. Moreover, the residue of drifted pesticides was found to be the highest in leafy vegetables such as perilla leaves or leaf and stem vegetables such as Welsh onion, followed by fruit vegetables and cucurbits, owing to the morphological characteristics of crops. Therefore, selecting pulse or cereal such as soybean or maize as a farm product near the UAV spray area can be considered to minimize the drift. For pesticides that exceed the MRL or PLS uniform level, %acceptable dietary intake is 0–0.81% with no risk. Additionally, employing pesticides approved for both paddy rice and farm products in UAV spraying can effectively minimize instances where MRL or PLS are exceeded. Therefore, this study aims to provide farmers with effective guidelines for mitigating drift. Furthermore, we strive to promote stable and uninterrupted food production while facilitating the utilization of agricultural technologies such as UAV spraying to address labor shortages and ensure sustainable food security.

Pesticides application is considered a necessary procedure to protect agricultural products from harmful insects and diseases 1 , and total pesticide use has increased by approximately 50% in the 2020s compared to the 1990s 2 . However, concerns have been raised regarding the excessive use of pesticides and the risks they pose to both human health and the environment 3 . Furthermore, some countries are attempting to reduce the use of pesticides to achieve sustainable intensification (SI) in food production to meet the needs of a growing global population 4 .
However, in response to these concerns, the Agricultural Chemical Regulation Law 5 and risk assessment 6 have been implemented for the safe use of pesticides, as has been done in other developed countries 7 . Additionally, SI could be made feasible via technology, such as Internet of Things (IoT) 8 , big data 9 , artificial intelligence (AI) 10 , and unmanned aerial vehicles (UAVs) 11 in agriculture. In particular, UAVs could prove to be an effective alternative solution to address labor shortages in agricultural work by enabling crop monitoring and pesticide spraying 12,13 , as the population of farmers has decreased while their average age has increased in some countries [14][15][16] .
Nevertheless, upon spraying pesticides with UAVs, the airborne pesticides could drift to non-target areas through the air 17 , resulting in unintentionally contaminating humans, plants, animals, and the environment 18 . To reduce pesticide drift, some factors have been studied, including (1) meteorological conditions 19 such as wind direction and speed 20 , humidity, and temperature 21 ; (2) UAV spray conditions such as spray pressure 22 , flight height 23 , and flight speed 24 ; (3) UAV components such as rotor 25 and nozzle 26,27 ; and (4) physical properties of spray solutions according to adjuvant 28 and formulation 29  www.nature.com/scientificreports/ However, in the context of aerial spraying conducted over paddy rice fields in diverse topographies during a specific period in Korea, the factors contributing to drift, including meteorological conditions, UAV spray conditions, UAV components, types of UAVs (multicopter or helicopter), and physical properties of the spray solution 30 , were not identified. Furthermore, it is important to note that drifted pesticides have the potential to impact individuals residing in proximity to farming areas 31 and pose a risk of drifting onto non-target crops, considering that UAVs primarily apply pesticides onto paddy fields 32 . In such a case, the residue of pesticides in these crops would exceed the maximum residue limits (MRL) or positive list system (PLS) uniform standard (0.01 mg/kg), which could pose a risk to human health if such contaminated crops are ingested.
However, buffer strips 33 , windbreaks 34 , and morphological characteristics 35 can potentially impact the residue of drifted pesticides and should be investigated following UAV spraying. Furthermore, although drift mitigation measures have shown some reduction in pesticide contamination over time, significant risks to human health and the environment still persist 36 . Therefore, it is necessary to monitor drifted pesticides considering these three factors and conduct risk analysis after UAV pesticide application to ensure effective prevention. Hence, this study aims to achieve the following objectives: (1) investigate the residual characteristics of pesticides that have drifted onto non-target crops surrounding the aerial sprayed area (paddy rice), taking into account three factors, that is, buffer strip, windbreak, and morphological characteristics of non-target crops; (2) implement preventive measures to reduce pesticide drift by monitoring various non-target crops near paddy rice in three regions (Boryeong, Seocheon, and Pyeongtaek) in the Republic of Korea; and (3) conduct a risk analysis to assess the risk posed by drifted pesticides in non-target crops, utilizing the residue levels of pesticides exceeding the standard MRL or PLS (0.01 mg/kg).
Approval statement. The samples were collected for monitoring pesticide drift by UAV spray with the approval of the farmers. Additionally, all methods, from collection of samples to analysis of residual pesticides, were performed in accordance with the relevant guidelines and regulations of the National Institute of Agricultural Sciences (NAS) and Ministry of Government Legislation (MOLEG).
Stock and working solution. To prepare 1,000 mg/L of stock solutions of pesticides, each RM was weighted with precision balance, considering the purities of the RM. Thus, 20.26 mg of etofenprox, 20.20 mg of hexaconazole, and 20.30 mg of propiconazole were dissolved in 20 mL of methanol. Each stock solution was diluted with acetonitrile to concentrations ranging from 0.005 to 100 mg/L. These concentrations were used to plot the regression curve and conduct the recovery test.  38 . The samples (10 g) were placed in a 50-mL conical centrifuge tube (FalcornTM, US) and shaken with acetonitrile (10 mL) for 5 min at 1300 rpm. The sample was then shaken again under the same conditions with magnesium sulfate (4 g), sodium chloride (1 g), trisodium citrate dihydrate (1 g), and disodium hydrogencitrate sesquihydrate (0.5 g) (QuEChERS EN extraction packet). The mixture was centrifuged for 5 min at 3500 rpm, and the supernatant (1 mL) was added in a dispersive Solid-Phase Extraction (d-SPE) tube containing magnesium sulfate (150 mg) and primary secondary amine (PSA) (25 mg) for clean-up. The tube was then vortexed for 30 s. The purified solution was filtered using a syringe filter (PTFE, 13 mm, 0.22 µm) after centrifuging for 5 min at 12,000 rpm. The supernatant was then mixed in a 50:50 (v/v) proportion with acetonitrile to create a matrixmatched sample, which was analyzed with LC-MS/MS according to Supplementary Table S2 39 . For soybean, the sample was analyzed directly after matrix matching without performing the purification procedure.
Verification of analysis method. The limit of quantitation (LOQ) was determined by setting a signalto-noise ratio of over 10 in a matrix-matched standard considering the PLS uniform standard (0.01 mg/kg) 40 . A regression curve for calibration was plotted by analyzing more than five matrix-matched standards and comparing the concentration and intensity of the peaks to evaluate the correlation coefficient (r 2 ) according to SANTE/11312/2021 41 . The recovery test was conducted with apple, wakegi onion, perilla leaves, and soybean as representative crops from commodity groups of collected crops 42 . The accuracy and repetition of the recovery test were evaluated using three fortification levels of LOQ, 10 LOQ, and 50 LOQ with recovery (%) and relative standard deviation (RSD) according to performance criteria for analysis of pesticide 43 .
Decision on drifted pesticides. To understand the drift of pesticides through UAV spraying, a wealth of information is required, encompassing meteorological conditions, UAV types, spray conditions, UAV components, and physical properties of the spray solution that can influence drift. However, in this study, pesticides were simultaneously sprayed onto paddy rice in various topographies using two types of UAVs within a specific timeframe. Consequently, it was not feasible to capture all the details during each spraying event. Additionally, the presence of residual pesticides in non-target crops may not be solely attributed to UAV drift but could also be attributed to farmer practices. Hence, we explore other factors that could influence drift and are applicable for investigation even after UAV spraying. The residue of drifted pesticides is influenced by plant morphology 44 , buffer strips 45 , and windbreaks 46 . Therefore, we examined these factors when collecting non-target crops. To understand the characteristics of residual pesticides, we analyze the pesticide residues based on three factors: buffer strip, crop morphology, and windbreak. Furthermore, we investigate whether the aerially sprayed pesticides are commonly detected and registered in the harvested crops.

Risk analysis.
A risk analysis was conducted using estimated daily intake (EDI) and % acceptable daily intake (%ADI) in cases where the residue of pesticides in crop around aerial sprayed area exceed MRL or PLS uniform standard (0.01 mg/kg), specifically for crops with an established ADI 47 . To calculate EDI and %ADI, food consumption (g/day) was determined from the "National Food & Nutrition Statistics" 48 . Moreover, the average weight of an adult in Korea, which is 66.55 kg, was established based on the "National Health Screening Statistical Yearbook" (Eqs. 1 and 2) 49 .

Statistical analysis. To plot a calibration regression curve and assess the correlation coefficient, Microsoft
Excel (USA) was utilized. The residual characteristics of the collected samples with drifted pesticides were analyzed using Statistical Package for the Social Sciences (SPSS) software (Ver. 26, IBM Corporation, USA). In cases where pesticides were detected in only two samples at a specific site, an independent sample t-test was employed for residue analysis. In cases where pesticides were detected in more than three samples, a one-way analysis of variance (ANOVA) was conducted, followed by Duncan's multiple range test (DMRT) with a significance level of p < 0.05.
In three locations (location nos. 1, 14, and 16), the residues of both etofenprox in peach and azoxystrobin in chili pepper exceeded the MRL (Tables 1, 2 and 3). However, the other pesticides sprayed by UAVs were not detected, and there was no evidence of drifted pesticides, as the residue of two pesticides did not decrease as the distance increased from the sprayed area 32 . Furthermore, azoxystrobin is commonly used in farm products such as chili pepper 50 , and etofenprox is frequently detected in both herbal fruits and stalk and stem vegetables 51,52 . Therefore, it can be concluded that the five cases that exceeded the MRL were not due to the drift of airborne pesticides but rather the presence of pesticides that were already sprayed before the UAV spraying.
In total, the residue of pesticides exceeded the PLS uniform standard in 31 cases across 15 locations (location nos. 2, 6-4, 10,12,21,25,28,29,30,31,32,33,34,35,36). These include one case of etofenprox in white-flowered gourd, five cases of propiconazole (three cases of squash leaves and a case of dureup and white-flowered gourd), 21 cases of ferimzone (four cases of soybean leaves, eight cases of chili pepper, four cases of Welsh onion, two cases of eggplant, and one case each of squash leaves, perilla leaves, and tomato), and four cases of dinotefuran in soybean leaves. It was inferred that the pesticides that exceeded the PLS uniform standards were drifted by UAV spray, as all the cases showed uniform residual tendency of pesticides according to distance from the sprayed area, windbreak, and morphology of crops 32 , and the pesticides sprayed by UAV were commonly detected. It is recommended that pesticides for UAV spray be chosen in the case of both rice and farm products, considering ferimzone, which is only applied to rice, and the 68% of total cases that exceeded PLS.
Drift characteristics according to buffer strip. Airborne pesticides were found to decrease as the distance from the aerial sprayed area increased, with 0-100% drift reduction (location nos. 2, 4, 5, 6-4, 8, 9, 11, 12, 14, 15, 21, 25, 29, 30, 32, 33, 35). This trend is consistent with the results of wind tunnel and field tests, which also showed that the amount of airborne pesticides decreased with increasing distance 22,53 . The samples were collected from distances ranging from 0 to 22 m from the aerial sprayed area, with an average distance of 5.7 m. It appears that crops were grown around aerial sprayed area without a uniform buffer strip. Therefore, it would be appropriate to establish a uniform buffer strip around the aerial sprayed area 34 . However, target pesticides were detected in non-target fruits and leafy vegetables, such as squash leaves, white-flowered gourd leaves, and maize leaves, beyond 5.7 m from the sprayed area (Tables 1, 2 and 3). Moreover, wind speeds have been identified as a factor influencing drift deposition 54 , and it has been advised to establish a buffer strip ranging from 25 to 300 m in the case of herbicide spraying 34 . However, the effectiveness of a buffer strip in reducing pesticide drift can vary, resulting in inconsistent levels of drift reduction ranging from 0 to 100%. This variability is attributed to the influence of other factors, including meteorological conditions, morphological characteristics of non-target crops, and UAV spray conditions. Consequently, relying solely on the implementation of a buffer strip may prove insufficient in preventing drift.
Drift characteristics according to crop morphology. The residue of airborne pesticides in non-target samples varies according to the morphological characteristics of the collected samples (location nos. 6-4, 9, 12, 17, 21, 23, 24, 25, 28, 30, 34, 36). Analysis results show that pesticide residue was higher in leafy or stalk and stem vegetables than fruiting vegetables other than cucurbits, followed by cucurbits (p < 0.05). Soybean and maize appear to be less susceptible to drifted pesticides because of their outer layer, and the residue of pesticides was less than in other crops (Tables 1, 2 and 3). This tendency is similar to earlier reports stating that the residue of carbamates pesticide was the lowest in cereal and pulses 55 and that residues of pesticides were higher in leafy vegetables than in leaves of root and bulb vegetables, including pulses with pods, fruits, pulses and cereal grains, and root and bulb vegetables 44 .
In addition, peach, perilla leaves, and squash leaves with glandular hair tended to have higher residues of pesticides (p < 0.05). The deposition of droplets was affected by components such as microstructure, wax, stomata, and hair on the surface of leaves 56 . The retention of droplets improves in leaves with longer even hairs or rougher surface 57 . Additionally, four types of leaves, namely cocklebur, morning glory, velvet leaf, and coffee senna, showed 99, 77, 65, and 55% of deposition efficiencies, respectively, when sprayed with 140 µm of droplets 35 . Similarly, needle-like leaves had two to four times higher deposition than broad leaves 58 ; thus, it was concluded that the amount of deposit of airborne pesticides could differ according to the morphological characteristics of crops. Consequently, it is recommended to grow crops less susceptible to retention of airborne pesticides around aerial sprayed area to prevent unintentional contamination.
Given the early-or mid-growth stages of the collected samples, it was not an appropriate time to assess the safety of pesticides. Furthermore, the residue of pesticides tends to dissipate or degrade over time due to the growth of agricultural products, which is influenced by meteorological conditions such as radiation, temperature, humidity, and rainfall, as well as the physio-chemical properties of the pesticides 59 . Consequently, it is likely that the residue of drifted pesticides in farm products will decrease by the time of harvest. The following section describes the residual patterns of drifted pesticides, assuming that they were caused by drift.
Drift characteristics according to windbreak. The analysis of residual pesticides indicates that the residues of pesticides were lower when windbreaks were between non-target crop and aerial sprayed area (location nos. 10 Risk analysis. The result of the risk analysis involving 36 cases that exceeded the MRL or PLS uniform standard showed that the %ADI ranged from 0.00 to 0.81% (Table 4); this shows that there was no risk 63 . Considering that processing treatments, including washing, peeling, heat treatments, and drying, can effectively eliminate residual pesticides 64 , the likelihood of consuming these farm products with any associated risk is low. However, it is essential to maintain ongoing monitoring of residual pesticides that drift onto non-target crops through aerial application. This continued monitoring aims to investigate the causes and potential risks posed by drifted pesticides, specifically to implement preventive measures against drift.

Conclusion
To address the labor shortage in domestic agriculture, the use of pesticides sprayed by UAVs has become necessary. However, this practice poses the risk of pesticide drift onto non-target crops and potential harm to humans. Consequently, monitoring the drift of pesticides and conducting risk analysis on non-target crops is crucial. The analysis results indicate that pesticide drift can be minimized by increasing the distance between non-target crops and the UAV spray area or by implementing windbreaks, such as planting maize, between them. Moreover, the Table 1. Residual concentration of pesticides in the crops surrounding paddy rice sprayed with UAV in Boryeong. a It was considered that the pesticide was already sprayed before UAV spraying took place. b Location was surrounded by UAV sprayed area. c It was grown in vinyl house with a side window opened by 0.8 m. d It was grown in vinyl house. *There were significant differences at the p < 0.05 level by t-test. No significant difference exists within the same location if the same small letter appears in the same column (p < 0.05). www.nature.com/scientificreports/ residue of drifted pesticides tends to be lower in pulses or cereal grains, such as soybeans or maize, compared to leafy or stalk vegetables such as Welsh onions or crops with granular hairs, such as perilla leaves. Additionally, using pesticides registered for both paddy rice and other farm products in UAV spraying helps prevent unintentional pesticide contamination. Therefore, it is anticipated that these guidelines will assist farmers in avoiding violations of PLS or MRL, thereby enabling stable and continuous food production in agricultural fields and positively impacting food self-sufficiency rates. Furthermore, we envision that agricultural technologies such as UAV spraying can be utilized not only to address labor shortages but also to enhance sustainable food security.

Location no Crop
In future research, we plan to investigate specific prevention measures for drift using a geographic information system to understand how terrain factors can potentially influence drift resulting from UAV spraying. Additionally, we will consider factors such as buffer strips, windbreaks, crop morphology, and the choice of pesticides for UAV spraying to assess their effectiveness in reducing drift. Table 2. Residual concentration of pesticides in the crops surrounding paddy rice sprayed with UAV in Seocheon. a It was considered that the pesticide was already sprayed before UAV spraying took place. b It was grown in vinyl house with a side window opened by 1 m. c It was grown in vinyl house. *There were significant differences at p < 0.05 level by t-test. No significant difference exists within the same location if the same small letter appears in the same column (p < 0.05).  Table 3. Residual concentration of pesticides in the crops surrounding paddy rice sprayed with UAV in Pyeongtaek. a It was considered that the pesticide was already sprayed before UAV spraying took place. b Location was an interjection of two sides of UAV sprayed area. *There were significant differences at p < 0.05 level by t-test. No significant difference exists within the same location if the same small letter appears in the same column (p < 0.05).

Data availability
The datasets used and analyzed during the current study can be made available from the corresponding author upon reasonable request.