Weed control efficacy of herbicide mixtures and digital assessment methods during pre-planting burndown in soybean

Abstract

Farmers use herbicide mixtures to increase the efficacy and spectrum of weed control during the pre-planting burndown stage of soybean. The efficacy of herbicides should be rigorously assessed, and digital tools have great potential for application during pre-planting weed management. This study aimed to evaluate the efficacy of weed control using herbicide mixtures and to test digital tools that analyze spectral responses as a means to assess efficacy. Two field experiments were set up, and the herbicides tested included diquat + carfentrazone-ethyl, glyphosate + 2,4-D, glyphosate + ammonium glufosinate, glyphosate + carfentrazone-ethyl, and glyphosate + clethodim. The main weed species identified were Urochioa piantaginea, Commeiina benghaiensis, Nicandra physaiodes, Ipomoea triloba, Portuiaca oieracea, Cyperus rotundus, Eieusine indica, and Aiternanthera teneiia. Efficacy for each species was assessed using a 0–100% scale for visual injury (VI). Each plot was evaluated using five methods: VI, Visible Atmospherically Resistant Index (VARI), Canopeo App, manual Greenseeker, and evaluation of images on a computer screen. The alternative methods were compared to the VI scores using Pearson correlation analysis. It was observed that, in general, the herbicides were less effective in controlling E. indica, C. benghaiensis, I. triioba, and A. teneiia. Control scores greater than 80% were recorded for U. piantaginea, N. physaiodes, and P. oieracea. Alternative methods can be used as additives or substitutes for herbicide efficacy assessments. Greenseeker generally showed a strong correlation with VI, and the efficacy of fast-acting products was more easily detected using alternative methods.

Keywords:
ARP; Canopeo; Greenseeker; VARI; NDVI

1. Introduction

Weed control, particularly through herbicides, has been a topic of significant debate and research in recent years (Ofosu et al., 2023; Nath et al., 2024). This is especially true in countries with large agricultural sectors and high herbicide consumption, such as Brazil. As the world’s largest producer and exporter of soybeans, Brazil ranks among the world’s highest herbicide consumers (Gaboardi et al., 2023; Klein, Luna, 2023).

In the main soybean-producing regions of Brazil, herbicides are applied in post-emergence conditions before sowing. Tank mixtures are widely used due to the challenges posed by weed diversity, including the presence of tolerant and resistant species (Heap, 2024). Commonly used herbicides include 2,4-D, ammonium glufosinate, carfentrazone-ethyl, clethodim, diquat, and glyphosate (Merotto Jr. et al., 2022). Effective pre-planting weed control supports improved performance of post-planting herbicide applications. Additionally, it is well established that applying herbicides at advanced growth stages in soybean can lead to both quantitative and qualitative yield losses (Cantu et al., 2021; Ceretta et al., 2023; Cassol et al., 2024).

Due to the increasing difficulty of weed control, herbicide use has intensified among farmers. This increased usage generates pressure for developing resistant biotypes, poses risks to human health and the environment, and increases overall production costs (Hasanuzzaman et al., 2020). Additionally, managing the effectiveness of applications, particularly mixtures, has become challenging for field specialists. Common errors include ineffective control, crop damage, and herbicide drift (Barbieri et al., 2022).

Hard-to-control weeds such as Eieusine indica, Ipomoea triioba, and Cyperus rotundus exhibit varying sensitivity to herbicides depending on their growth stage and the environmental conditions in which they occur (Peerzada, 2017; Lopez Ovejero et al., 2019; Zhang et al., 2021). Additionally, herbicide mixtures differ in cost, and farmers often opt for products that offer a lower cost per hectare (Merotto Jr. et al., 2022). These factors highlight the importance of evaluating the efficacy of each mixture for controlling individual species and minimizing the risk of inappropriate product selection.

Currently, efficacy assessments are based on the typical injury that herbicides cause. Evaluators visit the field and assign scores (0 to 100%. 0 absence of symptoms, 100% death of the plant) to measure control effectiveness (European Weed Research Council, 1964; Associación Latinoamericana de Malezas, 1974). Post-emergence herbicides typically injure plants, causing chlorosis followed by necrosis. Depending on the mode of action, symptoms may also include growth suppression and wilting. The severity of these lesions depends on environmental factors, herbicide translocation, and plant-specific characteristics. As a result, field evaluators must account for various interacting variables to assess the effectiveness of herbicide control or crop damage (Székács, 2021).

Digital and remote sensing tools have brought significant innovation to agriculture, especially for herbicide application and field assessments (Talaviya et al., 2020). In Brazil, the development of plant spectral response technologies has gained momentum, especially with the widespread use of drones and sensors (Miller et al., 2024). Errors related to human observation, difficulties evaluating large areas, and misinterpreting plant symptoms hinder effective herbicide management. Even in field research, efficacy assessments are time-consuming, costly, and prone to evaluator bias, often leading to discrepancies. Digital tools and remote sensing have increasingly supported herbicide management, optimizing agricultural processes. Chlorosis in specific plants and crops can be accurately identified using spectral indices derived from sensors operating in the visible (Red-Green-Blue, RGB) and near-infrared spectra (Nkuna et al., 2024).

The Visible Atmospheric Resistance Index (VARI) is a spectral indicator used to assess plant canopy fraction while minimizing sensitivity to atmospheric effects. This index relies on RGB sensors to detect plant stress. Similarly, the Normalized Difference Vegetation Index (NDVI) is used to detect plant stress but is based on responses from the red and near-infrared bands. Spectral indices can be determined through digital image processing using data captured by cameras mounted on satellites, drones, or even common handheld devices. Additionally, handheld digital sensors provide specific and reliable spectral responses for accurate and efficient data collection (Gitelson et al., 2002; Miller et al., 2024).

Canopeo (www.canopeoapp.com) determines green cover based on red/green and blue/green ratios. The system is available as a smartphone app and assesses vegetation through photos or videos. It was developed with the Matlab programming language (Mathworks, Inc., Natick, MA) and used for vegetation assessments (Patrignani, Ochsner, 2015). Similarly, the portable Greenseeker (NTech Industries, Trimble, Sunnyvale, CA, USA) is based on terrestrial remote sensing. The device uses radiation emission diodes and reads reflectance through a microprocessor, precisely indicating the NDVI of the vegetation cover. There is a direct correlation between the intensity of the green, the vigor of the plants, and the score generated on the device’s screen (Zsebő et al., 2024).

Based on the aforementioned context, this study aimed to evaluate the efficacy of herbicides applied in tank mixtures for weed control in pre-planting soybean. Additionally, we aimed to assess the correlation between the conventional visual analysis method (Visual Index - VI) and spectral-based weed analysis following digital image processing.

2. Material and Methods

2.1 Field experiments

Two field experiments were conducted in Uberlândia, MG, Brazil. The first experiment was initiated in November 2023, and the second in February 2024. Different areas containing weeds at the pre-flowering stage were selected. In the first experiment, the weed species identified were Urochloa plantaginea, Commelina benghalensis, Nicandra physalodes, Ipomoea triloba, and Portulaca oleracea. In the second experiment, the species identified were Cyperus rotundus, C. benghalensis, Eleusine indica, Alternanthera tenella, and P. oleracea (additional characteristics of the areas are provided in the supplementary material).

The experimental area was divided into four blocks, each containing six plots measuring 5 × 3 m (15 m²). Treatments were randomly distributed within the blocks, and the following tank mixtures were tested: diquat + carfentrazone-ethyl (300 + 30 g/ha); glyphosate + 2,4-D (1,175.5 + 1,005 g a.e/ha); glyphosate + glufosinate ammonium (1,175.5 g a.e/ha + 400 g/ha); glyphosate + carfentrazone-ethyl (1,175.5 g a.e/ha + 30 g/ha), and glyphosate + clethodim (1,175.5 g a.e/ha + 192 g/ha). Commercial products: Reglone®, Aurora® 400 EC, Roundup Transorb®, Aminol® 806, Finale® e Select 240 EC. A control plot with no herbicide application was included, which served as the untreated control.

We mixed the herbicides at the time of application. A knapsack sprayer was used for herbicide application, maintaining a constant pressure of 2.02 kgf cm⁻² (198 kPa), CO₂-powered and equipped with pressure gauges. The 3.0-m spray bar featured four Teejet (Cotia, Brazil) AI 11002 (air induction flat-fan) nozzles, providing a 3.0-m application width. The spray volume was 150 L ha⁻¹. Environmental conditions during application were as follows: in Experiment 1, relative humidity was 71%, temperature was 25.5 ºC, and wind speed was 3.5 km/h; in Experiment 2, relative humidity was 74.5%, temperature was 26.0 ºC, and wind speed was 4.5 km/h.

2.2 Evaluations and statistical analysis

In the first experiment, evaluations were conducted at 3, 7, 14, 21, 28, and 35 days after herbicide application (DAA). In the second experiment, evaluations occurred at 3, 7, 14, 28, and 35 DAA. Herbicide mixture efficacy was assessed based on the conventional Visual Index (VI) methodology proposed by Associación Latinoamericana de Malezas (1974). The control efficacy was categorized as follows (%): absent (0–40), regular (41–60), good (61–80), very good (81–90), and excellent (91–100). Three evaluators assessed two central meters (2.0 m × 1.0 m) of each plot. Evaluations were performed individually for each species in every plot (Table 1) between 10:00 a.m. and 2:00 p.m. Simultaneously, assessments were performed using Canopeo and Greenseeker, and drone imagery was captured to calculate the VARI.

Table 1
Pearson correlation coefficient (P value = 0.05%) between visual control rates (Associación Latinoamericana de Malezas, 1974) and other methods for evaluating herbicide mixture efficacy

Herbicide mixtures	VARIa		Canopeob		Greenseekerc		C. screend
	2023	2024	2023	2024	2023	2024	2023	2024
Diquat + carfentrazone-ethyl	−0.76	−0.89	−0.93	−0.62	−0.92	−0.96	0.94	0.97
Glyphosate + 2,4-D	−0.02	0.08	−0.66	−0.85	−0.73	−0.81	0.65	0.50
Glyphopsate + ammonium glufosinate	−0.85	−0.22	−0.79	−0.63	−0.41	−0.33	0.49	0.42
Glyphosate + carfentrazone-ethyl	0.01	−0.49	−0.75	−0.66	−0.89	−0.87	0.87	0.89
Glyphosate + clethodim	−0.36	−0.88	−0.74	−0.90	−0.89	−0.98	0.50	0.61

• a: Visible Atmospherically Resistant Index; b: canopeoapp.com. c: Portable Greenseeker. d: Evaluation of untreated photos on a computer screen in the office.

The Canopeo app was installed on an Apple iPhone 8 with a 12 MP camera. The device was perpendicular to the ground and positioned 1.0 m above the plot. Two images were captured per plot within the central 2.0 m², following the protocol described by Patrignani and Ochsner (2015) and the developer’s website (canopeoapp.com). The portable Greenseeker was also positioned 1.0 m above the plot to determine NDVI values. Two evaluations were conducted taken from the central 2.0 m² of each plot according to the guidelines from NTech Industries (2008) and as described in the manufacturer’s documentation (https://www.nue.okstate.edu).

To determine the VARI, the images were collected with a remotely piloted aircraft (ARP - DJI Air2S®) with a 20-megapixel camera, which flew over the area at an altitude of 25.0 m. The images were loaded into the Qgis program (Qgis Project, 2018), and the VARI was generated using the raster calculator. The bands with reflectance in blue (p467), green (p559), and red (p640) were considered using the formula: (Green - Red) / (Green + Red - Blue) (Gitelson et al., 2002). The spectral information was extracted in vector (point) format, considering a 2.0 m² border. Finally, the images captured with the ARP were evaluated directly on the computer monitor (C. screen). For that, the evaluators assigned control scores (0 to 100%) according to the VI without treating the images.

Data were first tested for variance homogeneity and residual normality. Subsequently, variance and regression analyses were performed. Herbicide efficacy was evaluated by presenting treatment results obtained from regression analysis. VI data were compared with Canopeo, Greenseeker, VARI, and Computer Screen (C. screen) assessments using Pearson’s correlation. VARI and Greenseeker values were standardized to a 0–100 scale for comparison. A 5% probability of error was adopted for all statistical analyses. Data analysis and result visualization were performed using Excel, RStudio, and SigmaPlot.

3. Results

3.1 Weed control efficacy of herbicides tank mixture

Regression models were fitted to analyze the relationship between VI and evaluation days for each weed species. Glyphosate-based herbicide mixtures demonstrated similar efficacy in controlling U. plantaginea, with the maximum control observed around 21 days after application. Diquat combined with carfentrazone-ethyl showed strong efficacy against U. plantaginea, but this effect was limited to the first evaluation week (Figure 1A). For E. indica, glyphosate combined with clethodim and glyphosate combined with 2,4-D provided comparable control levels; however, the glyphosate+clethodim mixture achieved superior results, with control values exceeding 80%. In contrast, diquat+carfentrazone-ethyl and glyphosate+glufosinate ammonium mixtures failed to provide satisfactory control of E. indica (Figure 1B).

Figure 1
Weed control notes (Associación Latinoamericana de Malezas, 1974) after application of herbicide mixtures

In 2023 and 2024, the initial control levels of C. benghalensis were high following treatment with diquat + carfentrazone and glyphosate + carfentrazone-ethyl; however, plants exhibited rapid regrowth thereafter. In 2023, rapid regrowth was particularly notable after treatment with diquat + carfentrazone-ethyl. Other herbicide mixtures showed lower efficacy, with control ratings below 60% (Figures 2A and 2B).

Figure 2
Weed control notes (Associación Latinoamericana de Malezas, 1974) after application of herbicide mixtures

The herbicides demonstrated excellent control of N. physalodes, with control ratings approaching 90%. No statistically significant differences were observed among the treatments with glyphosate + carfentrazone-ethyl, diquat + carfentrazone-ethyl, and glyphosate + ammonium glufosinate (Figure 3A). For C. rotundus, only glyphosate + clethodim achieved over 80% control, outperforming the other treatments. Glyphosate + 2,4-D and glyphosate + carfentrazone-ethyl achieved comparable control, with ratings near 70% at 21 DAA. A rapid resurgence of C. rotundus was observed following treatment with diquat + carfentrazone-ethyl (Figure 3B).

Figure 3
Weed control notes (Associación Latinoamericana de Malezas, 1974) after application of herbicide mixtures

Glyphosate+clethodim, glyphosate+2,4-D, and glyphosate+ammonium glufosinate demonstrated comparable efficacy in controlling P. oleracea in 2023, with control scores of approximately 75% from the fourth evaluation period (Figure 4A). In 2024, similar results were observed for the same herbicides, with control scores nearing 80% at 28 DAA (Figure 4B). P. oleracea showed rapid recovery after treatment with diquat+carfentrazone-ethyl in 2024. Furthermore, the data did not conform to a regression model for glyphosate + carfentrazone applications (Figure 4).

Figure 4
Weed control notes (Associación Latinoamericana de Malezas, 1974) after application of herbicide mixtures

None of the treatments effectively controlled I. triloba, particularly after the third evaluation period. Diquat+carfentrazone-ethyl, glyphosate+ammonium glufosinate, and glyphosate+carfentrazone-ethyl initially provided very good control; however, plant recovery was observed in the following weeks. The best results were achieved with glyphosate+carfentrazone-ethyl, with control scores declining from 80% to 60% over time. For the glyphosate+2,4-D treatment, the regression model did not fit the data for controlling I. triloba (Figure 5A).

Figure 5
Weed control notes (Associación Latinoamericana de Malezas, 1974) after application of herbicide mixtures

Glyphosate+clethodim and glyphosate+2,4-D provided comparable control of A. tenella, with control scores approaching 75% by the third week of evaluation. Diquat+carfentrazone-ethyl and glyphosate+ammonium glufosinate initially achieved control scores near 80%, but plant recovery occurred rapidly in the subsequent weeks (Figure 6B). No statistically significant control of A. tenella was observed with glyphosate+carfentrazone-ethyl in 2024 (Figure 5B).

Figure 6
Evaluation of the efficacy of weed control by glyphosate+clethodim using five methods

3.2 Digital tools and visual efficacy

Regarding the responses of the evaluation methods to the treatments, it was observed that glyphosate+clethodim showed the highest VI values at approximately 20 DAA in 2023. In 2024, control was highest during the final evaluation period. The VI results were similar to those of C. screen in 2023 until the third evaluation period. Canopeo and Greenseeker showed comparable results in 2023. Overall, the curves for the evaluation methods exhibited similar trends. However, for glyphosate+clethodim, Canopeo and Greenseeker displayed opposite trends when compared to VI and C. screen scores. VARI values also showed opposite trends to C. screen values in 2023, but this relationship varied depending on the evaluation period (Figure 6).

For glyphosate+carfentrazone-ethyl, VI was consistent across the evaluation period in 2023 (average control = 73%). However, in 2024, maximum control efficacy was achieved at approximately 23 DAA, with a score close to 75%. C. screen and VARI exhibited opposite trends during part of the evaluation period, a pattern also observed for VI and Greenseeker (Figure 7).

Figure 7
Evaluation of the efficacy of weed control by glyphosate+carfentrazone-ethyl using five methods

For glyphosate+2,4-D, VI remained consistent throughout the evaluation period in 2023 (average control = 64%). In contrast, in 2024, control peaked at the end of the evaluation period, reaching close to 80%. Canopeo and Greenseeker scores were similar in both 2023 and 2024. In 2023, Canopeo and Greenseeker displayed an inverse trend compared to C. screen. Lastly, VARI also exhibited an inverse trend compared to C. screen in 2023, although this relationship varied depending on the evaluation period (Figure 8).

Figure 8
Evaluation of the efficacy of weed control by glyphosate+2,4-D using five methods

The control scores (VI and C. screen) for diquat+carfentrazone-ethyl declined over time, starting at around 80% at 3 DAA and decreasing to 20% by 35 DAA. Canopeo and Greenseeker exhibited an inverse relationship with VI, with scores starting close to 20% at 3 DAA and rising to 90% at 35 DAA in 2023. VARI scores increased progressively, showing an inverse trend relative to VI (with a lower slope angle). In 2023, VARI values rose from 3 to 80 over the evaluation period. Finally, in 2024, Greenseeker values increased, reaching 90 at 28 DAA (Figure 9).

Figure 9
Evaluation of the efficacy of weed control by diquat+carfentrazone-ethyl using five methods

For the efficacy of glyphosate+ammonium glufosinate, the VI data showed trends similar to C. screen in 2023. The VARI and Canopeo values remained stable for the first 15 days, then increased in the final two evaluations. In 2024, the VI values followed the same pattern as in 2023, with slight variation, and the control rate remained close to 55. Canopeo values in 2024 mirrored those of 2023. In contrast, the VARI values exhibited increased gradually from 14.2 to 23.3. No significant variation was observed in the Greenseeker data for both years, with values consistently around 40 (Figure 10).

Figure 10
Evaluation of the efficacy of weed control by glyphosate+ammonium glufosinate using five methods

When comparing alternative methods for evaluating herbicide efficacy with VI, a very strong (≥ 90) or strong (70–89) correlation with VI was found in 55% of evaluations. For VARI, Pearson’s correlation was classified as strong in 30% of the evaluations. For Canopeo, the correlation was very strong in 30% of the evaluations, strong in 30%, and moderate in 40%. For Greenseeker, the correlation was very strong in 30% and strong in 50%. For the monitor evaluation (C. screen), the correlation was very strong in 20%, strong in 20%, and moderate in 40% (Table 1).

Considering the two years of experimentation (2023 and 2024) and comparing all methods with respect to VI, the correlation for diquat + carfentrazone-ethyl was very strong in 62.5% of the analyses and strong in 25%. For glyphosate+2,4-D, the correlation was strong for Canopeo in 2024 and for Greenseeker. For glyphosate+ammonium glufosinate, the correlation was strong in 2023 for both VARI and Canopeo. For glyphosate+carfentrazone-ethyl, the correlation was strong in 62.5% of the evaluations, mainly for Greenseeker and the monitor evaluations (C. screen). Finally, for glyphosate+clethodim, the correlations were 25% very strong and 37.5% strong (Table 1).

4. Discussion

4.1 Weed control efficacy of herbicides tank mixture

Glyphosate is the main herbicide sold worldwide, providing excellent control of most weeds, especially herbaceous plants or those from the Poaceae family. All mixtures containing glyphosate were highly effective in controlling U. plantaginea and N. physalodes, particularly three weeks after application, due to its systemic action and the high biomass of the plants at the time of treatment. In the experimental area, U. plantaginea had the largest biomass in 2023, and N. physalodes is a tall plant with large leaves. As a result, these plants were more exposed to the spray, which enhanced the efficiency of the glyphosate-based mixtures (Wynn, Webb, 2022).

Diquat+carfentrazone ethyl provided poor or inconsistent control of all plants (except N. physalodes), particularly in evaluations conducted further from the application date. Both herbicides have non-systemic action and require extensive plant coverage to be effective. Consequently, the suboptimal control observed in the present experiment is likely attributable to the reduced spray volume employed (150 L ha⁻¹) and the advanced weed growth stage at the time of application (pre-flowering).

Additionally, we used an air-induction spray nozzle, which generates larger droplets and consequently results in less target coverage. Taller plants with greater biomass also created an umbrella effect, shading shorter plants (e.g., P. oleraceae, C. rotundus, and C. benghalensis). Finally, vegetatively propagating plants (e.g., U. plantaginea and C. rotundus) are less affected by non-systemic herbicides. These factors combined lead to the lower efficacy of diquat+carfentrazone-ethyl (Székács, 2021).

Glyphosate-resistant and -tolerant biotypes of E. indica are present in Brazil (Heap, 2024). This vigorous-growing plant tolerates most herbicides when applied late, making it one of the most significant weeds in agriculture (Zhang et al., 2021). A mixture of glyphosate with an ACCase-inhibiting herbicide (such as clethodim) is recommended to improve control effectiveness. Similarly, combining glyphosate with 2,4-D enhances control efficacy (Agostinetto et al., 2022). Notably, the mixtures of glyphosate with clethodim and 2,4-D significantly improved control of E. indica, particularly from the third evaluation period onwards.

The control scores for C. benghalensis, which were classified as good, fair, or even absent (Associación Latinoamericana de Malezas, 1974), especially in the early days after application, can be explained by the plant’s tolerance to glyphosate (Bottcher, 2022). Additionally, C. benghalensis exhibits prostrate growth, thrives in shaded areas, and has a high capacity for vegetative reproduction (Isaac et al., 2013). As a result, the plant was minimally exposed to the herbicide spray and reproduced vegetatively in the weeks following the application.

C. rotundus, although sensitive to glyphosate, reproduces vigorously vegetatively and is also sensitive to light deprivation. Vegetative reproduction limits the effectiveness of herbicides with low mobility, while the plant’s reliance on light for growth makes it susceptible to shading (Peerzada, 2017). In this experiment, C. rotundus was shaded by other plants, but the herbicide application still provided good to very good control scores. The slight difference in control between glyphosate+2,4-D and glyphosate+carfentrazone-ethyl can be attributed to the lack of translocation of the PROTOX inhibitor, which may interfere with the translocation of glyphosate (Werlang, Silva, 2002).

The herbicides did not provide adequate control of I. triloba, likely due to its tolerance to glyphosate. Additionally, this species thrives under shaded conditions, and at the time of application, it was protected by other plants (Lopez Ovejero et al., 2019). Mixtures of glyphosate with 2,4-D, carfentrazone-ethyl, and ammonium glufosinate are recommended for controlling this species (Joseph et al., 2017; Carneiro et al., 2020). However, proper spray coverage is essential for these mixtures’ effectiveness, especially for non-systemic products.

For A. tenella, control was classified as good only in the first few days after applying glyphosate combined with carfentrazone-ethyl and ammonium glufosinate. Mixtures containing systemic products provided high control scores only in the third evaluation period. The decline in control over time is likely due to the growth habits of the plants. A. tenella is a low-growing plant that was protected by larger plants during spraying.

Chemical weed control in pre-planting soybean fields should be conducted in multiple stages when the vegetation is dense and diverse, especially when using non-residual herbicides. In shaded conditions, plants can benefit from the herbicide’s effects on other species. In such cases, sequential applications before crop installation are recommended (Palharani et al., 2023). Increased light penetration into the plot promotes accelerated growth of more prostrate species when controlling larger plants (such as U. plantaginea and N. physalodes). This study suggests that herbicide applications should be performed sequentially prior to soybean planting. In the first stage, the mixture of glyphosate + 2,4-D effectively controls larger plants. Approximately 15 days later, combinations such as glyphosate + ammonium glufosinate or glyphosate + carfentrazone-ethyl can be applied to control smaller plants that were previously sheltered beneath the canopy of other weeds.

Considering all plants, the best control scores (VI) ranged from approximately 60% to 80%. Mixtures with systemic herbicides showed higher efficacy between 14 to 21 DAA. Non-systemic products provided higher control scores in the first evaluations, but their effectiveness decreased over time, which can be attributed to the faster action of contact herbicides. Mixtures containing glyphosate exhibited similar behavior, likely due to the variety of weeds in the area and the "umbrella effect" created by taller plants (Székács, 2021).

4.2 Digital tools and visual efficacy.

When evaluating the correlation between methods for assessing herbicide efficacy, both Canopeo and Greenseeker showed a high correlation with visual evaluations for nearly all herbicides. Canopeo analyzes all image pixels based on color proportions (red-to-green, blue-to-green) and an excess green index. In other words, this system prioritizes spectral bands between approximately 500 and 570 nm (Patrignani, Ochsner, 2015). Canopeo can assess vegetation cover and light interception and even indirectly estimate biomass and nutrient accumulation (Campana et al., 2023). Similarly, Greenseeker employs radiation emission diodes across the 650 nm (red) to 770 nm (near infrared) bands. The resulting ratio between the near-infrared and red light reflects plant health, where high values reflect vigorous vegetation with extensive soil coverage (Zsebő et al., 2024).

Herbicides such as carfentrazone-ethyl, diquat, and ammonium glufosinate have low plant mobility and cause rapid chlorosis and necrosis. In contrast, glyphosate, clethodim, and 2,4-D act more slowly, with chlorosis appearing days or even weeks after application, depending on environmental conditions (Székács, 2021). These rapid changes in plant color are easily detected by systems like Canopeo and Greenseeker, which explains why these two systems often generate similar results when monitoring plant responses to herbicide-induced injury.

VARI uses all visible bands of the electromagnetic spectrum. Since the indicator works with RGB bands (Miura et al., 2001), a strong correlation with the on-site evaluation was observed in 40% of the results, particularly for diquat+carfentrazone-ethyl. Comparing VARI with C. screen, negative correlation trends were observed for all herbicides in 2023 (Figures 7 to 11). It is important to note that the correlation varies depending on the evaluation period.

The evaluation of images captured by the RGB camera mounted on the drone, displayed directly on the computer monitor (C. screen), also showed good correlations with the on-site evaluations for treatments containing carfentrazone-ethyl. There was no cloud cover during the flights; a 20 MP camera was used, and the drone flew at an altitude of 25 m. Good flight conditions and equipment performance resulted in detailed images of the plots, allowing for a reliable evaluation by assessors in front of a computer monitor. Among the herbicides studied, carfentrazone-ethyl and diquat caused the most easily identifiable injuries. However, it is crucial to emphasize that correlations are influenced by the evaluation period (DAA).

Regarding comparing all methods with the conventional evaluation (VI), it is possible that diquat+carfentrazone-ethyl showed better values due to the rapid non-selective effects of the mixture. These products cause plant necrotic lesions within a few days (Székács, 2021). Glyphosate+clethodim and glyphosate+carfentrazone-ethyl also presented many good or very good correlation values, as the predominant vegetation in the experiments was U. plantaginea (in 2023) and C. rotundus (in 2024), as both species are highly responsive to these herbicide combinations.

The best correlation values for Greenseeker may be linked to the spectral range that the system evaluates. In conventional herbicide efficacy assessments, evaluators consider, in addition to chlorosis and necrosis, symptoms such as biomass reduction, leaf, and stem curling, epinasty, and others that may not always indicate chlorosis (Associación Latinoamericana de Malezas, 1974; Székács, 2021). In other words, the spectral responses of plants, specifically the ratio between near-infrared and red-light reflectance, may more accurately reflect herbicide injuries (especially systemic ones) than RGB-based evaluations.

When evaluating the efficacy of a given herbicide, it is essential to understand the specific injuries the product causes. This study observed epinasty (caused by 2,4-D) using VI, but other methods would likely not have detected such injury. Factors such as species variability, application timing, environmental conditions, and application technology directly impact efficacy. Highly efficient products can benefit plants hidden in dense vegetation, while non-systemic products may promote regrowth in many species. In the studies presented here, numerous products and diverse weed communities were evaluated, and it was observed that alternative methods to visual assessment could serve as a substitute or complementary tool, especially for products that act quickly or when the substitute indicator is NDVI (Greenseeker). Ultimately, these are simple methods that are easy to understand and apply.

5. Conclusions

Diquat+carfentrazone-ethyl, glyphosate+2,4-D, glyphosate +ammonium glufosinate, glyphosate+carfentrazone-ethyl, and glyphosate+clethodim provided good levels of control for U. plantaginea, N. physalodes, and P. oleraceae.

However, applying these products did not effectively control E. indica, C. benghalensis, I. triloba, and A. tenella. This outcome may be related to the application technology or the tolerance of these species’ specific biotypes.

A sequential application is recommended before soybean planting for pre-plant burndown in areas with high density, vigor, and late-stage weeds.

The evaluation of herbicide mixture efficacy using VARI, Canopeo, Greenseeker, and images captured by drone-mounted cameras can be used as a substitute or complement to on-site evaluations.

References

• Agostinetto D, Pigatto CS, Zandoná RR, Roberto A, Silva BMD, Andres A. Interaction of clethodim with glyphosate and/or 2, 4-D at different doses and spray volumes in the control of glypho-sate-resistant ryegrass. Adv Weed Sci. 2022;40:1-7. Available from: https://doi.org/10.51694/AdvWeedSci/2022;40:00022
  » https://doi.org/10.51694/AdvWeedSci/2022;40:00022
• Associación Latinoamericana de Malezas – ALAM. Recomendaciones sobre unificación de los sistemas de evaluación en ensayos de control de malezas. Assoc Latinoam Malezas. 1974;1:35-8.
• Barbieri GF, Young BG, Dayan FE, Streibig JC, Takano H, Merotto AJ et al. Herbicide mixtures: interactions and modeling. Adv Weed Sci. 2022;40:1-25. Available from: https://doi.org/10.51694/AdvWeed-Sci/2022;40:seventy-five011
  » https://doi.org/10.51694/AdvWeed-Sci/2022;40:seventy-five011
• Bottcher AA, Albrecht AJP, Albrecht LP, Kashivaqui ESF, Cassol M, Souza CNZD et al. Herbicide efficacy in the fall management of Rich¤rdi¤ brosiliensis, Commelino benghaiensis, Conyza sumatrensis, and Digitaria insularis. Biosci J. 2022;38:1-9. Available from: https://doi.org/10.14393/BJ-v38n0a2022-53544
  » https://doi.org/10.14393/BJ-v38n0a2022-53544
• Campana M, Valle TAD, Fernandes LS, Pereira FRDS, Garcia TM, Osório JAC et al. Canopeo and GreenSeeker applications as tools to support tropical pasture management. Cienc Rural. 2022;53(6):1-10. Available from: https://doi.org/10.1590/0103-8478cr20220167
  » https://doi.org/10.1590/0103-8478cr20220167
• Cantu RM, Albrecht LP, Albrecht AJ, Silva AF, Danilussi MT, Lorenzetti JB. Herbicide alternative for Conyza sumatrensis control in pre-planting in no-till soybeans. Adv Weed Sci. 2021;39:1-9. Available from: https://doi.org/10.51694/AdvWeed-Sci/2021;39:000012
  » https://doi.org/10.51694/AdvWeed-Sci/2021;39:000012
• Cassol M, Albrecht AJ, Albrecht LP, Silva AFM, Larini WF, Galvão FGC. Mixtures of herbicides for the control of multiple resistant Cony-za sumatrensis in soybean pre-sowing burndown. Rev Bras Cienc Agr. 2024;19(3):1-8. Available from: https://doi.org/10.5039/agraria.v19i3a3719
  » https://doi.org/10.5039/agraria.v19i3a3719
• Carneiro GDOP, Souz MF, Linas HA, Chagas, PS, Silva TS, Silva Teofilo TM, Pavao QS, Grangeiro LC, Silva DV. Herbicide mixtures affect adsorption processes in soils under sugarcane cultivation. Geoderma. 2020;379:1-18, Available from: 10.1016/j.geoderma.2020.114626
• Ceretta JM, Albrecht AJ, Albrecht LP, Silva AF, Yokoyama AS. Can pre-and/or post-emergent herbicide application affect soybean seed quality? Rev Caatinga. 2023;36(4):740-7. Available from: https://doi.org/10.1590/1983-21252023v36n401rc
  » https://doi.org/10.1590/1983-21252023v36n401rc
• European Weed Research Council - EWRC. Report of the 3rd, and 4th meetings of EWRC: Comitte of methods in weed research. Weed Res. 1964;4.
• Gaboardi SC, Candiotto LZP, Panis C. Agribusiness in Brazil and its dependence on the use of pesticides. Hyg Environ Health Adv. 2023;8:1-11. Available from: https://doi.org/10.1016/j.heha.2023.100080
  » https://doi.org/10.1016/j.heha.2023.100080
• Gitelson AA, Kaufman YJ, Stark R. Novel algorithms for remote estimation of vegetation fraction. Rem Sens Environ. 2002;80:76-87. Available from: https://doi.org/10.1016/S0034-4257(01)00289-9
  » https://doi.org/10.1016/S0034-4257(01)00289-9
• Hasanuzzaman M, Mohsin SM, Bhuyan MHMB, Bhuiyan TF, Anee TI, Masud AAC et al. Phytotoxicity, environmental and health hazards of herbicides: challenges and ways forward. In: Prasad MNV, editor. Agrochemicals detection, treatment and remediation: pesticides and chemical fertilizers. Oxford: Butterworth-Heinemann; 2020. p. 55-99 Available from: https://doi.org/10.1016/B978-0-08-103017-2.00003-9
  » https://doi.org/10.1016/B978-0-08-103017-2.00003-9
• Heap I. The international herbicide-resistant weed database. Weedscience. 2024. Available from: https://www.weedscience.org
  » https://www.weedscience.org
• Isaac WA, Gao Z, Li M. Managing commelina species: prospects and limitations. In: Price AJ, Kelton JA, editor. Herbicides: current research and case studies in use. London: IntechOpen; 2013[access 03, 21,2025]. Available from: https://doi.org/10.5772/55842
  » https://doi.org/10.5772/55842
• Joseph DD, Marshall MW, Sanders CH. Palmer amaranth (Amaranthus palmeri S. Wats.) and pitted morningglory (Ipomoea lacunosa L.) control in dicamba tolerant soybean (Glycine max L.). Am J Plant Sci. 2017;8(13):3429-42. Available from: https://doi.org/10.4236/ajps.2017.813230
  » https://doi.org/10.4236/ajps.2017.813230
• Klein HS, Luna FV. Cattle. In: Klein HS, Luna FV, editors. Brazilian crops in the global market: the emergence of Brazil as a world agribusiness exporter since 1950. Cham: Springer Nature; 2023. p. 295-323.
• Lopez Ovejero RF, Picoli GJ, Takano HK, Palhano M, Westra P. Residual herbicides in Roundup Ready soybean: a case study in multiple years and locations with Ipomoea triloba. Cienc Agrotec. 2019;43:1-10. Available from: https://doi.org/10.1590/1413-7054201943000319
  » https://doi.org/10.1590/1413-7054201943000319
• Merotto Jr A, Gazziero DL, Oliveira MC, Scursoni J, Garcia MA, Figueroa R et al. Herbicide use history and perspective in South America. Adv Weed Sci. 2022;40(spe 1):1-18. Available from: https://doi.org/10.51694/AdvWeedSci/2022;40:seventy-five010
  » https://doi.org/10.51694/AdvWeedSci/2022;40:seventy-five010
• Miller JO, Mondal P, Sarupria M. Sensor-based measurements of NDVI in small grain and corn fields by tractor, drone, and satellite platforms. Crop Environ. 2024;3(1):33-42. Available from: https://doi.org/10.1016/j.crope.2023.11.001
  » https://doi.org/10.1016/j.crope.2023.11.001
• Miura T, Huete AR, Yoshioka H, Holben BN. An error and sensitivity analysis of atmospheric resistant vegetation indices derived from dark target-based atmospheric correction. Rem Sens Environ. 2001;78(3):284-98. Available from: https://doi.org/10.1016/S0034-4257(01)00223-1
  » https://doi.org/10.1016/S0034-4257(01)00223-1
• Nath CP, Singh RG, Choudhary VK, Datta D, Nandan R, Singh SS. Challenges and alternatives of herbicide-based weed management. Agronomy. 2024;14(1):1-21. Available from: https://doi.org/10.3390/agronomy14010126
  » https://doi.org/10.3390/agronomy14010126
• Nkuna BL, Chirima JG, Newete SW, Nyamugama A, Van der Walt AJ. Developing models to detect maize diseases using spectral vegetation indices derived from spectral signatures. Egypt J Rem Sens Space Sci. 2024;27(3):597-603. Available from: https://doi.org/10.1016/j.ejrs.2024.07.005
  » https://doi.org/10.1016/j.ejrs.2024.07.005
• Ofosu R, Agyemang ED, Márton A, Pásztor G, Taller J, Kazinczi G. Herbicide Resistance: managing weeds in a changing world. Agronomy. 2023;13(6):1-16. Available from: https://doi.org/10.3390/agrono-my13061595
  » https://doi.org/10.3390/agrono-my13061595
• Palharani W, Mauad M, Silva PV, Medeiros ES, Schedenffeldt BF, Bicalho Medeiros CC. Management of Digitaria insularis in soybean pre-sowing desiccation by ACCase alternative herbicides and its impact on soybean carryover. J Environ Sci Health Part B. 2023;58(2):110-9. Available from: https://doi.org/10.1080/03601234.2023.2172279
  » https://doi.org/10.1080/03601234.2023.2172279
• Patrignani A, Ochsner TE. Canopeo: a powerful new tool for measuring fractional green canopy cover. Agron J. 2015;107(6):2312-20. Available from: https://doi.org/10.2134/agronj15.0150.
  » https://doi.org/10.2134/agronj15.0150.
• Peerzada AM. Biology, agricultural impact, and management of Cyperus rotundus L.: the world’s most tenacious weed. Acta Physiol Plant. 2017;39:270. Available from: https://doi.org/10.1007/s11738-017-2574-7
  » https://doi.org/10.1007/s11738-017-2574-7
• Székács A. Herbicide mode of action. In: Mesnage R, Zaller J, editor. Herbicides: chemistry, efficacy, toxicology, and environmental. Amsterdam: Elseiver; 2021. p. 41-86.
• Talaviya T, Shah D, Patel N, Yagnik M, Shah M. Implementation of artificial intelligence in agriculture for optimisation of irrigation and application of pesticides and herbicides. Art Intellig Agric. 2020;4:58-73. Available from: https://doi.org/10.1016/j.aiia.2020.04.002
  » https://doi.org/10.1016/j.aiia.2020.04.002
• Werlang RC, Silva AA. Interação de glyphosate com carfentra-zone-ethyl. Planta Daninha, 2002;20(1):93-102. Available from: https://doi.org/10.1590/S0100-83582002000100013
  » https://doi.org/10.1590/S0100-83582002000100013
• Wynn S, Webb E. Impact assessment of the loss of glyphosate within the EU: a literature review. Environ Sci Eur. 2022;34(1):1-10. Available from: https://doi.org/10.1186/s12302-022-00667-3
  » https://doi.org/10.1186/s12302-022-00667-3
• Zhang C, Yu CJ, Yu Q, Guo WL, Zhang TJ et al. Evolution of multiple target-site resistance mechanisms in individual plants of glyphosate-resistant Eleusine indica from China. Pest Manag Sci. 2021;77(10):4810-7. Available from: https://doi.org/10.1002/ps.6527
  » https://doi.org/10.1002/ps.6527
• Zsebõ S, Bede L, Kukorelli G, Kulmány IM, Milics G, Stencinger D et al. Yield Prediction using NDVI values from GreenSeeker and MicaSense cameras at different stages of winter wheat phenology. Drones. 2024;8(3):1-20. Available from: https://doi.org/10.3390/drones8030088
  » https://doi.org/10.3390/drones8030088

Edited by

• Editor in Chief:
  Carlos Eduardo Schaedler

• Associate Editor:
  Silvia Fogliatto

Publication Dates

• Publication in this collection
  15 Sept 2025
• Date of issue
  2025

History

• Received
  11 Feb 2025
• Accepted
  23 May 2025