Agroecological Alternatives for Substitution of Glyphosate in Orange Plantations (Citrus sinensis) Using GIS and UAVs

Abstract

Field mapping is one of the most important aspects of precision agriculture, and community drones will be able to empower young rural entrepreneurs who will be the generational replacement of a new agrosocial paradigm. This research presents an agroecological participatory innovation methodology that utilizes precision technology through geographic information systems and unmanned aerial vehicles to evaluate the integrated ecological management of weeds for glyphosate substitution in a transitional area of Citrus sinensis in San Luis Potosí, Mexico. Modeling methods and spatial analyses supported by intelligent georeference protocols were used to determine the number of weeds with tolerance and glyphosate resistance. Four control flights were conducted to monitor seven treatments. Glyphosate-resistant weeds were represented with the highest number of individuals and frequency in all experimental treatments. Although the treatment with maize stubble showed a slightly better result than the use of Mucuna pruriens mulch, which prevents the emergence of glyphosate resistant weeds before emergence, the second treatment is considered better in terms of the cost–benefit ratio, not only because of significantly lower cost but also because of the additional benefits it offers. Geospatial technologies will determine the nature of citrus and fruit tree agroecological treatments and highlight areas of the plot with binomial soil and plant nutrient deficiencies and pest and disease infestations, which will improve the timely application of bio-inputs through the development of accurate maps of agroecological transitions.

1. Introduction

From a phytosanitary point of view, weeds (endemic, invasive, and glyphosate-resistant) represent strong competition for conventional cash and subsistence crops, as they cause considerable losses in the form of yield reductions of up to 50% and a reduction in crop efficiency of around 20% due to competition for light, water, and nutrients [1]. Based on this vision, the typology of “aggressive weeds” was created. These are plants that have developed resistance to herbicides such that agrochemicals are less effective and are used in larger quantities, and combinations of several herbicides are used [2]. In Mexico, 36% of cases of herbicide-resistant weeds are concentrated, and 1.7% of these are resistant to glyphosate [3]. This is due to the fact that the country has increased the preferential use of glyphosate by 1500% since 1996 due to the cultivation of genetically modified corn, cotton, and soybeans [4]. In Mexican agriculture, both in conventional and subsistence farming, there is also evidence of the increased use of this pesticide in corn-growing units (35%), followed by citrus (14%), pasture (12%), sorghum (11%), and avocado (3%) [5].

Due to this situation, a policy for sustainable agroecosystems was introduced in Mexico with the Presidential Decree of 31 December 2020 [6], which allows the funding of scientific projects to promote agroecological practices (agroecological lighthouses) that are alternatives to the use of glyphosate and represent a paradigm shift: from conventional management based on agrochemicals to the integrated agroecological management of weeds [7].

Integrated agroecological weed management (known in Spanish as MEIA) includes a range of agroecological methods and treatments, such as preventive, cultural, physical, mechanical, and biological control practices, as well as the use of mulches, natural herbicides, natural bio-herbicides, and bio-herbicides [8].

Alternatives for glyphosate substitution require an agroecological transition and a “mapping of change” towards a comprehensive release and empowerment of producers in rural and biocultural contexts, and technologies are a kind of double-faced Janus, as they not only open up a world of possibilities and freedoms but are also part of a world of injustice.

In the last decade, however, geoinformation science and agroecology have developed common areas and mutual cooperative interests and interactions to achieve technological sovereignty. The challenges posed by Geographic Information Technology (GIT) in agroecology and organic agriculture relative to mapping the diversity of sustainability visions and the biological and cultural diversity of our rural communities are encompassed under the approach of “technology for all”: a dynamic combination of available tools adapted to specific places and cultures that goes beyond the universal closed menu of technological offerings that standardize conventional large-scale farms [9].

In the discourses on agricultural sustainability, the role of technology is overestimated, and the multi-layered articulating role of existing agroecological systems is forgotten when they are analyzed only in terms of yields. GIT provides farmers and citizens with tools for the pursuit of the rights of nature and food sovereignty, revealing the “power of maps” at the transformative level of the agroecological landscape.

Since the pre-digital era, geography and cartography have had a long and established epistemological development and empirical experience relative to the key role of maps in transforming the world through the empowerment of weak and marginalized actors in urban and rural contexts.

The potential of precision technologies, such as GIT, UAVs, and smartphones (GLONASS as geo-referencing technologies) [10], will provide answers in the context of community participatory technologies. Drones, also known as unmanned aerial vehicles (UAVs), remotely piloted aircraft systems (RPAS), or unmanned aerial systems (UASs), are currently widely used in the civilian sector.

The main feature of drones or UAVs is that their operator or pilot is not on board but in a remote-control station on the ground. The technological innovations associated with the development of drones have enabled new civil and commercial applications, such as their use by communities in different parts of the world.

The use of drones has boomed since 2014, particularly in Latin America. The reason for this is the need to document and report territorial impacts to improve community protection and the lack of detailed, accurate, and up-to-date maps for informed decision-making [11].

Mexico is one of the pioneer countries of the use of drones in communities [12]. Several communities have already demonstrated the benefits of this technology, and it is likely that more municipalities and institutions will promote the use of this technology in the near future. Whether for environmental monitoring, defense, or land management, drones in the hands of communities offer an opportunity to democratize remote sensing and the use of airspace.

The main activities involved in the management of an organic orange orchard include the following tasks: plantation establishment, weed control, pruning, tree and soil nutrition improvement measures, pest and disease control, buffer barriers, other cultivation practices, and special features required by organic regulations [13]. However, the use of high-precision monitoring information is not practiced and can potentiate agroecological activities. The development of high-precision spectral maps should identify the best treatments to reduce weed infestations in citrus plants that show an increase in tree spectral response (vegetative vigor) and can be potentiated in the region [14].

This research presents a participatory agroecological innovation methodology that utilizes GIT and UAV precision technology, with the objective of spatiotemporal monitoring and the spectral mapping of seven agroecological treatments for glyphosate substitution in a plot of Valencia orange (Citrus sinensis) plants in the Rioverde watershed in San Luis Potosí, Mexico. The objective of this research is to apply MEIA and evaluate the efficacy of pre-emergence bioherbicide applications for weed control in citrus plants without the use of synthetic herbicides; modeling and spatial analysis methods based on tree/quadrant monitoring are carried out, using intelligent georeference logs (IGL-Survey 123) and the spectral response of tree vigor associated with each agroecological practice. Four monitoring flights were conducted, accompanied by the producer and owner of the land.

To summarize, the innovation of this proposal includes the following measures:

• The creation of agroecological alternatives to replace toxic agrochemicals such as glyphosate and Paraquat, as part of actions that the Mexican federal government is implementing for responsible production and consumption.
• Digitalization of peasant agriculture with the adoption of digital technologies by farmers and rural communities to improve management, handling, and quality of life and modernize the sector: This includes the use of tools such as UAVs, artificial intelligence (AI), and big data (BD) to optimize production, improve efficiency, and increase sustainability.
• Quantifying the impact of actions carried out, in this case agroecological treatments, in a fast and accurate way through the use of unmanned aerial vehicles and geographic information systems.

The article is structured as follows: It begins with the geographical location where the field phase was conducted, followed by a description of the characteristics of the experiment and analysis performed. Then, the details of data collection and image acquisition are described, as well as the analysis and discussion of the results obtained. Finally, conclusions and recommendations were drawn from the analysis of the work.

2. Materials and Methods

2.1. Location and Experiment Setup

The present study was conducted in a citrus orchard in the central zone of San Luis Potosí, Mexico, in a Valencia orange orchard in the municipality of Ciudad Fernandez (Rioverde watershed). The plot was identified as “Tepetate” and is under permanent irrigation, as can be seen in Figure 1 and Figure 2.

The central zone of the state is one of the four geographical regions into which the state is divided. To the south is Rioverde Basin, the citrus-growing region with the highest production value in the state, made up of municipalities of Ciudad Fernandez, Rioverde, Lagunilla, and San Ciro de Acosta, which cover 4741 ha (14.9% of the state) and account for 59.4% of the state’s production value. This agricultural basin has the best yield per hectare and the best average price in the country, surpassing the national and state averages.

This is due to the maintenance of a phytosanitary status with the low prevalence of native fruit flies since 2010 to date, achieved through the MOSCAFRUT plant protection campaign and the weekly application of sterile flies [15,16].

The producer made 10% of his land available for the establishment of alternatives (Figure 2). This research focused on the establishment of 7 agroecological alternatives in the period from June to November 2023, with the care and training of growers and continuous precision sampling. The treatments were carried out on 7 rows of orange trees aged between 5 and 8 years old, covering an area of 1840 m². All trees tested positive for HLB, the main plant disease in the area [17]. Studies have shown that both the NDVI and the occurrence of HLB are influenced by climatic conditions, especially by vegetative stress processes [18].

The use of legumes in citrus cultivation helps with weed control, nitrogen fixation, and soil disinfection. Crotalaria, for example, helps reduce fungal problems, and Mucuna helps control nematode problems and promotes better nutrient transport. They keep the soil covered to prevent erosion, improve soil structure, and promote soil loosening (organic tillage) and can potentially serve as feed for livestock [13,19].

In treatment A1, the orchard was cleared with a brush cutter.

In treatments A2, A3, and A4, the soil was prepared with a tractor and a rotary tiller for sowing legumes to create a living cover: Mucuna pruriens was sown in A2, Canavalia ensiformis in A3, and Crotalaria juncea in A4. The seeds were sown in June 2023 by broadcasting at a density of 40 kg ha⁻¹.

In treatment A5, a dead cover with maize residues was applied. For establishment, 31,000 kg ha⁻¹ of maize stubble was sown in June 2023.

Treatment A6 used Herbitech™ by Biotech Mexico (Uruapan, Mexico), a bioherbicide and a systemic contact herbicide with a broad spectrum of activity for post-emergence weed control. It works quickly and effectively to control unwanted plants and grasses in crops. Its formulation consists of 100% natural ingredients such as mullein, coconut oil, pine resin, Puccinia fungus, and papain.

The bioherbicide (15 mL L⁻¹) was applied using a mechanical backpack sprayer pressurized with CO₂ and equipped with a four-tip bar pattern, model Teejet XR 110.02, with a pressure of 178 kPa and a distance of 0.5 m between tips. The walking speed during spraying was 1 m s⁻¹; a volume of 200 L ha⁻¹ was used.

Treatment A7 was used as a control, as no agronomic measures were carried out (

Table 1. Experimental treatments established in the Tepetate plot in 2023–2024.

Treatment	Mechanical Implementation and Live–Dead Covers	Dosage and Number of Plants
A1	Brush cutter	7
A2	Mucuna pruriens	7
A3	Canavalia ensiformis	7
A4	Crotalaria juncea	7
A5	Dead corn cover	7
A6	Bioherbicide	10 mL/L
A7	Control witness	7

).

For the detection of glyphosate traces in weeds in the field, 21 samples were taken and analyzed by Raman spectroscopy with a XploRA Plus™ system by Horiba Ltd. (Kioto, Japan) [20]. The soil type identified in the plot is Phaeozem [21], and the results of soil analysis showed that the plot has high content of organic matter (oxygen (O), calcium (Ca), copper (Cu), and sulfur (S)) and low content of iron (Fe), zinc (Zn), and manganese (Mn). In addition to glyphosate and Paraquat, 6 highly hazardous pesticides (HHPs) were used for weed control on the trial site [17].

The thirteen different weed species in

Table 2. Weeds identified in the Tepetate plot.

Positive Weeds	Negative Weeds
1. Acalypha setosa	11. Malva silvestris
2. Amaranthus palmeri	12. Euphorbia prostrata
3. Argemone munita	13. Solanum elaeagnifolium
4. Bidens odorata
5. Boerhavia erecta
6. Parthenium hysterophorus
7. Stevia berlandieri
8. Taraxacum officinale
9. Trifolium repens
10. Elusine indica

were identified using the SENASICA [15] and IGL-Survey 123 sampling method [22]. In this agroecological transition plot, 4 glyphosate-resistant and 6 glyphosate-tolerant weeds were found using Raman spectroscopy [20]. The methodology for agroecological field treatments was determined using the MEIA model [9,10]. In monitoring and quantifying the results, the precision agriculture (agromatics) methodology was applied, which consists of taking advantage of technology; observing, measuring, and analyzing temporal and spatial variability; and using all these data to indicate the agroecological treatments that provide the best results in weed control, both in agricultural terms and in terms of cost.

2.2. General Workflow

Figure 3 shows the methodological proposal developed for this study, which consists of three procedures: (a) collection of georeferenced ID-survey information on weeds through the sampling design and the resulting spatial modeling; (b) spectral survey (reflectance per tree) through radiometric campaigns and the creation of spectral libraries, resulting from the processing of the UAV images (mosaic and creation of the NDVI algorithm) and the application of statistical analysis on the obtained spectral results; and (c) the correlation of both spatial modeling and spectral results was calculated by linear regression, and the significance level was determined by the agroecological treatment (Figure 3).

2.3. Data Collection

Geographers, geomaticians, and agroecologists who practice precision agriculture often use different tools and data collection methods to analyze different variables and find the best course of action for both the producer and experimental plot. This is carried out to contribute to an agroecological transition that is integrated into the most sustainable biocultural region while improving production and investment costs. Data collection is at the heart of precision agriculture and farmer empowerment, which translates into georeferenced weed monitoring by using treatments.

2.3.1. Ground Truth

To determine the effectiveness of alternatives, 5 control evaluations were conducted at 30, 60, 90, 120, and 150 DDE (days after establishment) to estimate the percentage of weeds in each of the alternatives. During the evaluations, the number of individuals of emerged weeds per m² was counted through a metallic quadrant (0.5 m × 0.5 m), counting the weeds that were within the quadrant [15]. Figure 3 also shows how the Survey 123 application was used to collect data. A total of 20,500 field data points were collected during sampling. With this in mind, field data collection was supplemented by digital and cartographic methods. All sampling data were imported into a GIS and then stored in a geodatabase. For this purpose, data import, analysis, management, and mapping were carried out using the ArcGIS 10.2.

2.3.2. Image Collection

Monitoring with infrared cameras (UAV) is fundamental as they determine the spectral response of the tree to any agroecological treatment. A spectral comparison was made with the trees that showed greater vegetative vigor.

For this purpose, the vegetative vigor (NDVI) for each of the orange alternatives was determined using the Mavic 2 drone (SZ DJI Technology Co., Ltd., Shenzhen, China) with a visible 4K camera to obtain the georeferenced orthomosaic of the corresponding plot and the Mapir (MAPIR, Inc., San Diego, CA, USA) infrared camera to obtain the vegetation index measurements, such as NDVI, which includes data acquisition, information extraction, and field application [23].

Table 3. UAV missions in the orange plantation plots.

Flight Mission	Sensor	Flight Date (Acquisition)	Flight Height (m)	N of Images	Surface (m2)	Pixel (GSD) (cm)
1	Mapir Survey3, 1″ CMOS (visible)	4 July 2023	25	120	1840	1
2		18 August 2023		146	1840	1
3		20 October 2023		90	1840	1.17

shows the main characteristics of the 3 flight missions conducted according to the phenological stage of the citrus and weeds present. In addition, the requirements for the operation of a pilot-operated aircraft system (RPAS) in Mexican airspace and the formats for UAV reports established in the official Mexican standard NOM-107-SCT3-2019 [24] and in the official protocols of LANGIF-UASLP were met. In 2023, severe drought caused sprouting to decrease (also in risk systems), and in July, the orange trees were in the post-harvest and weed growth phase (M1—weed growth). In August, the orange trees began to sprout with increasing rainfall (M2—growth and ripening); in September, the orange trees began to flower, and the weeds began to grow strongly (M3—seed production and vegetative development).

The first two missions dealt with the first 7 trees of each row, and mission 3 covered the entire row of orange trees depending on the grower’s interest.

Similarly, the relationship between weed cover and NDVI, derived from all survey data for each of the treatments and the corresponding flights of the Mapir (spectral) camera UAV, was modeled in a regression-adjusted manner (R²).

2.4. Data Analysis

It must be said that large amounts of fertilizers, pesticides, and herbicides were used on this plot until the arrival of the work team. To determine the performance and outcome of agroecological treatments in eliminating glyphosate-tolerant and -resistant weeds, ID-survey in the Origin Lab program 2018 was used. One-way analysis of variance (ANOVA) was used to determine the effects of treatments applied using the Tukey test with XLSTAT 2025.1. To determine the linear dependence between the results of two technologies, Pearson’s correlation was applied according to the following formula:

$$r={\displaystyle \frac{\sum (X_i-\overline{X})\left(Y_i-\overline{Y}\right)}{\sqrt{\sum {\left(X_i-\overline{X}\right)}^2\sum {\left(Y_i\overline{Y}\right)}^2}}}$$

where X_i and Y_i represent the values of the variables, and $\overline{X}$ and $\overline{Y}$ represent the mean values of X and Y, respectively. This provided the opportunity to correlate the percentage of weed removal and the NDVI values, which quantifies the linear relationship between two continuous variables.

Pearson’s correlation coefficient is expressed as −1 to 1. In this sense, a value of 1 indicates a perfect positive correlation, i.e., as one variable increases, the other increases proportionally. A value of −1 indicates a negative correlation, i.e., an increase in one variable is accompanied by a decrease in the other, and a value of 0 means that there is no linear relationship between the variables [25].

3. Results and Discussion

The weed survey identified 13 weed species in the 7 alternatives, of which 2 were the most abundant on the farm and were cataloged as resistant to glyphosate [3], namely Parthenium hysterophorus and Eleusine indica.

Table 4. Glyphosate-resistant and -tolerant weeds by the number of individuals and frequency in all experimental treatments.

Scientific Name		Parthenium hysterophorus				Eleusine indica (plants/m2)				Trifolium repens
		bs	30 DAE	60 DAE	90 DAE	bs	30 DAE	60 DAE	90 DAE	bs	30 DAE	60 DAE	90 DAE
Treatments	A1—Brush cutter	230	200	187	142	300	288	250	190	176	131	248	42
	A2—Mucuna pruriens	70	62	50	10	170	157	75	25	149	103	76	50
	A3—Canavalia ensiformis	98	87	50	30	140	128	120	110	136	83	40	35
	A4—Crotalaria juncea	95	89	50	25	150	121	70	69	100	84	89	12
	A5—Dead corn cover	65	55	27	9	130	114	36	10	129	121	12
	A6—Bioherbicide	125	110	50	25	300	283	69	59	50	48	46
	A7—Control witness	112	120	128	134	250	280	340	370	74	65	68	79

bs = Before establishment; DAE = days after establishment.

summarizes how the use of GIS also enabled the creation of maps showing the spatial distribution of weeds present on the farm.

Of the seven experimental treatments, dead cover (A5) was the one that reduced pre-emergence weed emergence by 90%, followed by Mucuna pruriens (A2) at 75% from the last observation date. In treatment A7 with 120% weed growth, which was a control and in which no maintenance or bioherbicide application was carried out, the number of individual specimens per m² remained at a higher level from the first to the third evaluation (Figure 4).

Mucuna pruriens (A2) was the one that adapted the best to the site and quickly contributed to nitrogen fixation. Figure 5 shows that the brush cutter (A1) and the bioherbicide (A6), together with the control (A7), did not contribute to the increase in vegetative vigor (nutrient and moisture content of the orange trees). One-way analysis of variance showed that the value of vegetative vigor was significantly dependent on the agroecological treatment of the orange trees.

In mission 1 (one month after planting), most trees had an unchanged vigor level (NDVI). After two months, treatments A2, A3, and A5 were the first to show a significant spectral response.

In mission 3, alternatives A2 (Mucuna pruriens), A3 (Vicia sativa), and A5 (dead vegetation) increased the percentage of vegetative vigor, while alternatives A1 (brush cutter), A6 (bioherbicide), and A7 (control) showed a lower percentage of vigor under the same irrigation conditions.

The agroecological treatment with the highest correlation was A5: dead cover with R² = 0.83, followed by living cover A2: Mucuna pruriens. This was followed by the intermediate results of covers A3 and A4. A7 was the control with an R² = 0.189, which clearly shows that agroecological treatment very quickly leads to an improvement in the nutrient content of orange trees.

It was also observed that the root system of the trees to which the mulch of Mucuna pruriens was added improved, which must have influenced the improvement in the vitality of the trees.

In addition, the cost of each treatment was calculated in Mexican pesos and then converted to US dollars at the time of calculation (18 March 2025) using the fixed exchange rate (used to settle US-dollar-denominated debts) [26] (

Table 5. Effectiveness and costs of the treatments used.

Treatment	Mechanical Implement and Live–Dead Covers	R2 Between Weed Removal and NDVI	Cost of Treatment USD/ha
A1	Brush cutter	0.331	217
A2	Mucuna pruriens	0.823	2666
A3	Canavalia ensiformis	0.525	2666
A4	Crotalaria juncea	0.590	2897
A5	Dead corn cover	0.836	3409
A6	Bioherbicide	0.410	119
A7	Control witness	0.189	0

).

As can be seen, treatment A2, in which a live plant cover of Mucuna pruriens was added, is similarly effective in increasing the vegetative vigor of the orange trees on which it was applied. In addition, this legume was the one that best adapted to the soil and increased nitrogen fixation in the soil, which must be related to an improvement in the root system. On the other hand, the cost of treatment A2 is significantly lower (USD 2666) than that of treatment A5 (USD 3409), but it also has the additional benefits of the first treatment.

4. Conclusions and Recommendations

Field mapping is one of the most important aspects of precision agriculture, and in the future, community drones may enable young rural entrepreneurs, who could be the sons of producers (the generational shift of the new agro-social paradigm), to demonstrate the social, nutritional, and environmental benefits of agroecological practice. Thanks to drone technology, it will be easier to identify problem areas and create a complete and accurate map of agroecological transitions [27]. Dead cover (A5) was the one that prevented the emergence of weeds the most before sowing [28]. This treatment is expensive in the region, as it was made from corn residues and amounted to USD 3409 ([26]: Table 5). Therefore, it would be necessary to look for other types of mulches with organic residues from local crops or plantations, not only to reduce costs but also to create a sustainable economy, such as cactus stubble. In the citrus-growing area of Huasteca of San Luis Potosí, residues from sugar cane plantations could be used [29]. The next recommended treatment for the Rioverde citrus-growing zone would be to cover Mucuna pruriens, although the cost is relatively high (USD 2666). It will be necessary to sensitize growers to care for it, and it is advisable to use it as livestock feed. The bioherbicide did not achieve significant results in the area, as more than 67% of the weeds were tolerant or resistant to glyphosate [30], so its effectiveness was very low, although it is less expensive (USD 119). Brushcutting is a good field tool for the maintenance and pruning of legumes, but it should not be the only treatment. The mapping of weeds can be a useful tool for localized control, making in-field removal more streamlined, effective, and economical. Similarly, live cover crops (legumes) have dual benefits as they control weeds pre-emergence and are highly likely to fix nitrogen [31], improve secondary roots and absorptive hairs, and have the power to expose the phloem, which is the main problem caused by HLB (this is still being investigated by the working group). The use of legumes plays a triple role in diversifying the citrus agroecological landscape: It improves tree nutrition, produces natural enemies against insect pests, and serves as livestock feed [25]. The use of UAVs and multispectral cameras enables a precise and quantifiable assessment of alternatives for glyphosate replacement based on spectral information, such as the NDVI for this orange-growing area [32]. The mapping of weeds can be a useful tool for localized control, making their elimination on farms rational, effective, and economical.