Analyzing Soybean Growth Patterns in Open-Field Smart Agriculture under Different Irrigation and Cultivation Methods Using Drone-Based Vegetation Indices

Abstract

Faced with aging populations, declining resources, and limited agricultural productivity, rural areas in South Korea require innovative solutions. This study investigated the potential of drone-based vegetation indices (VIs) to analyze soybean growth patterns in open-field smart agriculture in Goesan-gun, Chungbuk Province, South Korea. We monitored multi-seasonal normalized difference vegetation index (NDVI) and the normalized difference red edge (NDRE) data for three soybean lots with different irrigation methods (subsurface drainage, conventional, subsurface drip irrigation) using drone remote sensing. Combining NDVI (photosynthetically active biomass, PAB) and NDRE (chlorophyll) offered a comprehensive analysis of soybean growth, capturing both overall health and stress responses. Our analysis revealed distinct growth patterns for each lot. LotA(subsurface drainage) displayed early vigor and efficient resource utilization (peaking at NDVI 0.971 and NDRE 0.686), likely due to the drainage system. Lot B (conventional cultivation) showed slower growth and potential limitations (peaking at NDVI 0.963 and NDRE 0.681), suggesting resource constraints or stress. Lot C (subsurface drip irrigation) exhibited rapid initial growth but faced later resource limitations(peaking at NDVI 0.970 and NDRE 0.695). By monitoring NDVI and NDRE variations, farmers can gain valuable insights to optimize resource allocation (reducing costs and environmental impact), improve crop yield and quality (maximizing yield potential), and address rural challenges in South Korea. This study demonstrates the promise of drone-based VIs for revitalizing open-field agriculture, boosting farm income, and attracting young talent, ultimately contributing to a more sustainable and prosperous future for rural communities. Further research integrating additional data and investigating physiological mechanisms can lead to even more effective management strategies and a deeper understanding of VI variations for optimized crop performance.

1. Introduction

Rural areas worldwide face numerous challenges threatening theirsustainability and prosperity.In SouthKorea,these challenges are particularly acute, with faster rural population aging and decline compared to other countries. Aging populations lead to labor shortages and social isolation (Jensen et al., 2020; Kinsella Kevin, 2001). Additionally, traditional agricultural techniques struggle to compete in global markets, causing economic hardship inrural areas(Mendelsohn, 2009;Misra andGhosh, 2024).Limited accesstowater, land, and technology further hindersrural areas, making them vulnerable to low agricultural productivity and challenges adapting to climate (Falkenmark, 2013). Furthermore, intensive agricultural practices often lead to soil loss, water pollution, and biodiversity loss (Xie et al., 2019). Amidst these challenges, smart and precision agriculture offers a promising solution. These technologies leverage data and technology to optimize resource allocation, improve efficiency, and enhance environmental sustainability (Borra-Serrano et al., 2020).

One key tool in precision and smart agriculture is the vegetation index (VI). Calculated from multi-temporal images acquired by satellites, drones, and otherremote sensing platforms, vegetation indices (VIs) quantitatively measure the greenness of plants (Jensen, 2009). Information derived from VIs offers several benefits to farmers. Firstly, it enables them to enhance productivity while minimizing inputs. By leveraging VIs to optimize the utilization of fertilizers and water, farmers can achieve higher yieldswith reduced resource usage (Guindo et al., 2021; Kumar et al., 2018). Secondly, VIs aid in environmental protection and resource conservation. Through the implementation of intelligent practices guided by VIs, farmers can decrease water, fertilizer, and pesticide usage, thus contributing to environmental preservation and resource conservation (Bhakta et al., 2019; Koutsos and Menexes, 2019).

Furthermore, the adoption of VIs promotes sustainable agriculture. Through the monitoring of soil health and the optimization of resource allocation, smart agriculture practices lead to more sustainable farming techniques and a diminished environmental footprint (Shah and Wu, 2019). Lastly, the integration of data and technology in smart agriculture makes it an appealing career choice for young individuals seeking innovative fields. This focus on modernization attracts young talent to agriculture, fostering a new generation of professionals in the industry (Huambachano et al., 2022; Sharma et al., 2020).

Recent advancements in drone and sensor technology have increased resolution and spectral sensitivity, allowing for more detailed and accurate VI data collection (Ni et al., 2017; Xue and Su, 2017). This study employed both the normalized difference vegetation index (NDVI) and the normalized difference red edge (NDRE) due to their complementary strengths in analyzing soybean growth patterns. NDVI’sstrong correlationwith photosynthetically active biomass (PAB) provides a general overview of plant health and productivity (Rouse et al., 1973; Tucker, 1979).

However, NDVI can saturate in dense canopies,where NDRE’s sensitivity to chlorophyll variationsthrough the “red-edge” band becomes valuable (Barnes et al., 2000). This sensitivity allows NDRE to detect stress, nutrient deficiencies, and subtle growth changes masked by NDVIsaturation (Gitelson et al., 2003). This combined approach provides a more comprehensive understanding of soybean growth dynamics.

Artificial intelligence (AI) algorithms are used to analyze VI data, providing actionable insights to farmers, such as pest and disease detection, yield prediction, and optimal irrigation schedules (Ramos et al., 2020; Taşan et al., 2022). South Korea’s efforts in open-field smart agriculture: The Korean Ministry of Agriculture, Food and Rural Affairs has been promoting the pilot project since 2018 to expand smart agriculture to open fields. In 2021, Goesan-gun, Chungbuk Province, was selected as a digital complex forsoybean cultivation. This project explores ways to apply and expand smart agriculture in open fields. Chungbuk actively seeks to revitalize open-field agriculture through scientific advancements. By incorporating sensors, drones, and AI, they aim to address rural challenges, improve farm income, and distribute these technologies to farmers. While the technology applied to open-field agriculture is still in its early stages, securing reliable technology and user-friendly data is crucial for its advancement.

Soybean production plays a vital role in global food security and economic stability.InKorea, optimizing soybean cultivation requires addressing issues like drainage in rice fields. To achieve this, a deep understanding of soybean growth processes is necessary. Drone remote sensing technology provides a powerful tool to analyze these dynamics by capturing VI data throughout the growing season. However, compared to other crops like rice and cabbage, research on soybean VI in Korea is limited, particularly in terms of multi-temporal studies.

This study aims to address these challenges and contribute to the development of open-field smart agriculture by (1) analyzing the multi-seasonal changes in VIs (NDVI and NDRE) data for soybeans grown in three different lot conditions (subsurface drainage, conventional cultivation, subsurface drip irrigation) using drone remote sensing and (2) exploring an appropriate growth model forsoybeans during the growing season based on the VI data.

2. Materials and Methods

2.1. Study Site

This study was conducted in Tapchon-ri, Buljeong-myeon, Goesan-gun, Chungbuk, South Korea (longitude 127°49′13″E, latitude 36°53′24″N) (Fig. 1). This location was chosen because it operates asthe GoesanOpen FieldAgricultureDigitalComplex, a pilot project for open-field agricultural digitalization by the Rural Development Administration (RDA, 2023). As shown in Fig. 1, the study area comprises three test plots, forming the Goesan Open Field Agriculture Digital Complex. These plots were categorized as Lots A, B, and C based on their type of farmland and irrigation drainage system (Table 1). The average yield for each plot was calculated per 10a (1000 m2) by dividing the total yield by the area.

Table 1. Relationship between irrigation drainage type and yield of each test lot

Fig. 1. Location of study area and classification of test bed lots. Here, A, B, and C represent the plot characteristics presented in Table 1.

Thisspecific locationwas chosen due to itssuitability for drone and multi-spectral sensor image acquisition. Additionally, the study aimed to contribute to the development of vegetation index utilization methods and the revitalization of open-field smart agriculture.

2.2. Drone Image Acquisition

Thisstudy employeda rotary-wingdrone,Inspire 2 (DJI, Shenzhen, China), equippedwith a multispectralsensor, RedEdge MX Dual (MicaSense, Seattle, WA, USA), to capture images of the study area (Table 2). Multispectral images were acquired nine times between June 17, 2022, and October 21, 2022, covering the entire soybean growing season until harvest. Additionally, six field surveyswere conducted throughout the study period. The drone imagery was acquired at an altitude of 30 meters with both longitudinal and horizontal overlapssetto 75%to ensure complete coverage and high-quality image mosaics.

Table 2. Specifications and image acquisition conditions of the drone and sensor

2.3. Leveraging Short-Term and Multi-Temporal Data for Soybean Growth Analysis

Soybeans are fundamental to global food security and economic stability. Optimizing their managementwithin digital agricultural complexes hinges on understanding their growth dynamics, requiring both detailed and broad data analysis.

Short-term data, captured frequently, reveals rapid changes in plant health,stressresponses, and environmental impacts(Bai et al., 2016). It provides real-time insights into photosynthesis, transpiration, and canopy temperature, offering a window into plant function and stress reactions (Ni et al., 2017). However, its dynamic nature often demands sophisticated analysis pipelines for real-time decision-making (Borra-Serrano et al., 2020).

Multi-temporal data, acquired overweeks or months, captures crucial growth stages like emergence, flowering, and pod development (Falco et al., 2021). This broader view enables the identification of long-term trends and correlations, allowing analysis of resource use efficiency, yield determinants, and responses to accumulated environmental factors. By averaging outshort-term fluctuations, multi-temporal data can be analyzed and modeled to provide stronger, more stable signals (BorraSerrano et al., 2020).

Each data type has its strengths and limitations. To effectively characterize soybean growth,we must utilize them appropriately. This study aims to achieve this by integrating and analyzing multi-temporal drone images. This approach leverages the complementary advantages of both data types, enabling betterinformed decision-making and improved resource management throughout the entire soybean production cycle.

2.4. Vegetation Indices

Among various vegetation indices used in drone-based remote sensing, two stand out for monitoring soybean growth of the NDVI and the NDRE (Table 3).

NDVIshowssimple to calculate,requiring only red and nearinfrared bands, making it readily available and interpretable. Extensive research provides a strong foundation for understanding its application. Effective for monitoring overall vegetation health, biomass, and land cover changes. Limited sensitivity to chlorophyll content in dense vegetation and susceptible to soil background effects.

NDRE represents red edge bands that enhance sensitivity to chlorophyll content, even in dense vegetation. May be useful for differentiating between plant species due to its sensitivity to chlorophyll variations. Requiresspecialized sensorswith red edge bands, limiting its broader application. Sensitive to atmospheric conditions,requiring careful consideration during data analysis.

Table 3. Two vegetation indices applied in this study

This study selected both NDVI and NDRE to leverage their complementary strengths. NDVI provides a stable and practical foundation for general monitoring, while NDRE offers the potential for detailed analysis and identification of specific lot conditions based on chlorophyll changes. This combination aims to balance existing research and future possibilities.

2.5. Soybean Growth Stages and Modeling

Soybeans exhibit a characteristic sigmoidal growth curve, with distinct growth stages defined by the rate of dry matter accumulation (Monzon et al., 2021). Understanding these stages is crucial for optimizing yield and resource management. Traditionally, soybean growth is categorized into eight stages, ranging from V1 (vegetative stage 1) to R8 (maturity) (Fehr and Caviness, 1977).

This study focuses on five key stages: (a) Reproductive organ formation stage (V1–V6): Thisinitialstage is marked by slowand steady growth as the plant establishes its foundation. Leaves unfold, but dry matter accumulation remains minimal (Malek et al., 2012). (b) Rapid reproductive organ growth stage (V7– R3): Photosynthesis accelerates, leading to rapid dry matter accumulation and vigorous growth. Leaves stretch towards sunlight, stems elongate, and branches proliferate (Malek et al., 2012). (c) Reproductive organ development stage (R4–R6): Flowers bloom, pods form, and small seeds begin to swell. This stage witnesses peak soybean growth rate (SGR) and maximum dry matter accumulation, with resources prioritized for reproduction (Monzon et al., 2021). (d) Seed charging stage (R7):Allresources are directed towardsseed development. Seeds expand, attracting dry matter, while SGR gradually declines due to shifting resource allocation priorities. (e) Senescence and maturity stage (R8): Leavessenesce, gracefully releasing nutrients to supportthe finalstages ofseed development. Growth stagnates, and the plant reaches its full potential yield.

Several soybean growth models exist, with the logistic model gaining widespread popularity due to its simplicity and effectiveness(Wardhani andKusumastuti, 2014). This model employs a sigmoid function (Eq. 1) with three parameters to predict soybean growth: carrying capacity (representing the maximum achievable biomass), initial dry matter weight, and relative growth rate.

\(\begin{align}\operatorname{sigmoid}\left(x, G_{\max }, G_{0}, k\right)=\frac{G_{\max }}{1+\left(\frac{G_{\max }}{G_{0}}-1\right) \times e^{-k x}}\end{align}\)       (1)

where x is the variable value (time), G_max is the maximum growth value (asymptote of the sigmoid function), G₀ is the initial growth value, k is a parameter controlling the slope of the function.

Thisstudy analyzesthe characteristics of each soybean growth stage using the obtained NDVI and NDRE data.

2.6. Evaluating Soybean Growth Stages Using Drone Images

This study utilized the process depicted in Fig. 2 to assess soybean growth stages using drone imagery. The research primarily focused on the analysis of vegetation indices, modeling, and visualization.

Fig. 2. Workflow for soybean data acquisition with drone and field survey, including ground control point (GCP) measurement, preprocessing, vegetation index calculation, and logistic growth model fitting.

3. Results and Discussion

3.1. Analysis of NDVI Variation Characteristics across Soybean Lots

All three soybean lots(Fig. 3) mirrored the typical growth curve with NDVI progressively increasing from sowing to peak near the reproductive stage (R4–R6), reflecting maximum leaf area and biomass (Malek et al., 2012; Monzon et al., 2021). This pattern parallels biomass accumulation, accelerating during vegetative growth, peaking with pod set and seed development, and then plateauing as nutrients shift towards seeds (Song et al., 2016). However, subtle differences emerged within this general trend,reflecting the influence of varying irrigation and cultivation methods. Lot A (subsurface drainage) shows early vigor and efficient resource utilization. Lot A initially low NDVI (0.2–0.3) indicated slow leaf emergence, but a rapid ascent to 0.5–0.6 during the V stages suggested efficient leaf expansion and biomass accumulation. A high and sustained NDVI plateau throughout the reproductive phase pointed towards optimal resource allocation to both leaves and reproductive organs, potentially benefiting from efficient drainage and leading to higher yield. This early advantage continued with a gradual and controlled NDVI decline during seed filling, further signifying efficient resource utilization.

Lot B (conventional cultivation) shows slower growth and potential stress. Lot B showed a slower growth trajectory compared to Lot A, highlighting the potential limitations of the conventional cultivation method. While eventually reaching similar NDVI levels in the vegetative stage, the initial rise was less pronounced, suggesting slower establishment or resource limitations.Asharper decline in the reproductive phase indicated potential stress impacting biomass accumulation and seed production, possibly due to competition for resources or lack of efficient drainage.

Lot C (subsurface drip irrigation) shows high early vigor and resource trade-offs. Lot C started slightly higher (0.3–0.4) and rapidly climbed to 0.6–0.7 during the V stages, exceeding the others due to optimal conditions for leaf expansion likely facilitated by controlled water supply. Maintained a relatively high NDVI plateau during the reproductive phase, suggesting continued resource allocation to both leaves and reproduction, potentially leading to higher yield. However, the initial slower rise hinted at potential establishment challenges, and the rapid decline after the peak suggested limited resource availability compared to Lot A, possibly due to higher transpiration.

Fig. 3. Variations in the spatiotemporal distribution of NDVI for each lot obtained using drone-acquired images and boxplots. (a) Lot A, (b) Lot B, (c) Lot C, and (d) NDVI average of 3 lots.

In conclusion, while all three lots followed the typical NDVI trend, their individual trajectories revealed variations in growth patterns, resource allocation, and potential yield, influenced by their specific irrigation and cultivation methods. This analysis highlightsthe need forfurtherinvestigation to refine management strategies for each lot based on their unique needs and potential limitations.

3.2. Analysis of NDRE Variation Characteristics across Soybean Lots

Mirroring the typical soybean growth curve, all three lots exhibited a similar NDRE trajectory (Fig. 4). Starting low at sowing (0.2–0.3), NDRE gradually increased during vegetative growth, peaked near the reproductive stage (R4–R6) reflecting maximum leaf area and biomass, and then slightly declined during seed filling and senescence (R7, R8). However, subtle differences emerged when examining each lot individually, revealing the influence oftheirrespective irrigation and cultivation methods.

Lot A (subsurface drainage) exhibited initially slow growth, withNDRE remaining stable around 0.2–0.3 untilrapid vegetative growth (V4–V6) triggered a sharp rise to 0.6–0.7. This suggests efficient leaf expansion, potentially due to optimal conditions facilitated by the drainage system. During the reproductive phase, NDRE maintained a high plateau (0.6–0.7), indicating efficient resource allocation to both leaves and reproductive organs, potentially leading to higher yield.

Fig. 4. Variations in the spatiotemporal distribution of NDRE for each lot obtained using drone-acquired images and boxplots. (a) Lot A, (b) Lot B, (c) Lot C, and (d) NDRE average of 3 lots.

Lot B (conventional cultivation) started similarly with a low NDRE and gradual increase, but the rise was less pronounced, reaching only 0.4–0.5 by V6, suggesting slower initial growth. Thisslowness persisted during both rapid vegetative growth and the reproductive phase, with NDRE remaining around 0.4–0.5. Thisimplieslimitationsin resource acquisition orstressimpacting leaf expansion and potentially affecting reproductive development and seed filling.

Lot C (subsurface drip irrigation) displayed better initial establishment and early growth, starting with a slightly higher NDRE (0.3–0.4). During rapid vegetative growth, NDRE rapidly and sustainably increased to 0.6–0.7, similar to Lot A, again indicating optimal conditions. The sustained high NDRE plateau (0.6–0.7) through the reproductive phase suggests continued resource allocation towards both leaves and reproduction, potentially leading to higher yield.

This analysis underscores the subtle yet significant influence of irrigation and cultivation methods on soybean growth, as captured by NDRE variations across different stages. By understanding these influences, we can develop targeted management strategiesto optimize both yield and resource use for each specific crop lot.

3.3. Modeling NDVI and NDRE Variation Curvesin Soybeans

The distinct growth patterns observed in the NDVI and NDRE curves for the three soybean lots call for a modeling approach that captures their unique features (Fig. 5).

Thisstudy aimed to select the best model forfitting the NDVI and NDRE growth curves of soybean lots under different irrigation methods. We evaluated four candidate models.

(a) Logistic growth model: This classic model captures overall trendswith interpretable parameterslike carrying capacity, growth rate, and inflection point. It works well until the post-growth maintenance phase (V1–R6; Fig. 6b), but its applicability sharply declines at the R8 stage due to a significant decrease (Fig. 6a).

(b) Gompertz growth model: Similarto the logistic model, but with a more pronounced initial lag phase, potentially betterfitting Lot C’sslowerinitialrise. It includes an additional parameterfor a refined transition between growth phases.

(c) Richards’ growth model: Thisflexible model adaptsto both sigmoid and asymptotic growth patterns, suitable for capturing diverse features across all lots (initial lag, rapid rise, plateaus/ declines). It incorporates parameters for carrying capacity, growth rate, and a shape parameter controlling curve curvature.

(d) Double logistic model: This approach uses two sequential logistic models to capture separate vegetative and reproductive growth phases. It might be useful for analyzing specific features like Lot B’s reproductive decline and Lot A’s sustained NDRE plateau.

Incorporating seasonal variations in NDVI and NDRE with sinusoidal terms can improve model accuracy (Fig. 7 dashed line). Including relevant factors like temperature, precipitation, or solar radiation can further refine the model and explain growth dynamics. Model selection and parameter estimation should be accompanied by statistical tests for model fit and parameter significance.

Fig. 5. NDVI and NDRE variation curves of soybeans for each lot. (a) The NDVI curves from V1 to R8 stages for each lot and (b) the NDRE curves fromV1 to R8 stages for each lot.

Fig. 6. Importance and problems of selecting the user’s application period, etc.: the need to select a different model depending on the selection of each variable. (a) The NDVI curves fromV1 to R8 stages and (b) the NDVI curves fromV1 to R4–R6 stages.

Fig. 7. Comparison of results of applying NDVI and sigmoid curve over time for soybeans in each lot. (a) The NDVI and logistic growth model curve of Lot A, (b) Lot B, and (c) Lot C.

The optimal model depends on our specific study goals and desired level of complexity (Fig. 7).While the logistic (Fig. 7) and Gompertz models offer a good starting point for general trends, the Richards’ and double logistic models provide more flexibility for detailed analysis. Ultimately, the chosen model should be validated against independent data to ensure accuracy and generalizability

3.4. NDVI and NDRECurve Difference and Application

Analyzing NDVI and NDRE curvesforthree soybean lots(yields: Lot A - 320 kg/10 a, Lot B - 340 kg/10 a, Lot C - 360 kg/10 a) under different irrigation and cultivation methods unveiled distinct growth patterns (Fig. 8).

Lot A (subsurface drainage) shows initial sluggishness transitioned to rapid leaf expansion and biomass accumulation (NDVI: 0.6–0.7) during V4–V6, suggesting efficient resource utilization aided by the drainage system. This advantage persisted, with a sustained high NDRE plateau during the reproductive phase, potentially indicating optimal resource allocation towards both leaves and reproductive organs.

Lot B (conventional cultivation) shows slower initial growth andconsistently lowerNDVI comparedtoLotAhinting atpotential limitations.AsharperNDREdeclineduring the reproductivephase further suggested resource limitations or stress impacting leaf expansion and chlorophyll content.

Fig. 8. Comparison of changes in growth characteristics of NDVI and NDRE according to soybean growth in each parcel. (a) The NDVI and NDRE of Lot A, (b) Lot B, and (c) Lot C.

Lot C (subsurface drip irrigation) boasts the highest initial NDVI and rapid rise during V4–V6, Lot C showcased optimal conditionsfor early leaf expansion, likely facilitated by a controlled water supply. However, a slightly slower initial rise and rapid NDREdecline afterpeak suggestpotential establishment challenges and limitations in resource availability compared to Lot A.

Factors driving individual curves: These unique curvesreflect the interplay of several factors. (a) Subsurface drainage in Lot A might promote early root development and efficient resource use, while conventional cultivation in Lot B could limit initial growth. Subsurface drip irrigation in Lot C might facilitate rapid initial growth but potentially lead to higher water loss later. (b) Differences in soil fertility, drainage, and water holding capacity across the lots could also contribute to the observed variations. (c) If different varieties are used, their inherent growth patterns and nitrogen utilization could influence the curves.

Harnessing NDVI and NDRE for improved management: By monitoring NDVI and NDRE changes, farmers can gain valuable insights for (a) track changes to identify potential issues like nutrient deficiencies, water stress, or disease outbreaks, (b) the use of NDVI and NDRE data to guide resource allocation, ensuring efficient use and preventing over-application. For example, Lot B might benefit from earlier orincreased irrigation, while Lot C might require adjustments to drip irrigation timing or nutrient concentrations, (c) by understanding the factors influencing individual curves, farmers can adjust management practices for each lot to maximize yield and quality. Lot A’s early vigor and efficient resource utilization suggest the potential for high yield,while Lot B mightrequire additionalsupport to reach its full potential. Lot C’s strong early growth and high NDRE during the reproductive phase also suggest potential for good yield, but its resource limitations need to be addressed.

While this study provides valuable insights into the potential of NDVI and NDRE for improving soybean management, further investigation is needed. Integrating additional data like soil maps, weather data, and detailed irrigation records could enable the development of site-specific management strategies for precision agriculture. Additionally, research on the specific physiological mechanisms underlying the observed NDVI and NDRE variations could provide a deeper understanding and potentially lead to targeted interventions for optimizing crop performance.

4. Conclusions

Thisstudy explored the potential of drone-based VIs(NDVI and NDRE) to analyze soybean growth patterns in open-field smart agriculture. By monitoring multi-seasonal NDVI and NDRE data for three soybean lots with different irrigation methods (subsurface drainage, conventional cultivation, subsurface drip irrigation),we identified distinct growth dynamicsinfluenced by each approach.

Our analysis revealed variations in resource utilization, leaf expansion, and chlorophyll content across the lots, reflecting the impact of irrigation and cultivation practices. Subsurface drainage in LotApromoted early root development and efficient resource use, leading to rapid leaf expansion and a sustained high NDVI plateau during the reproductive phase. In contrast, conventional cultivation in Lot B limited initial growth and showed a sharper NDRE decline, suggesting potential resource limitations orstress. Subsurface drip irrigation in LotCfacilitated rapid initial growth but hinted at establishment challenges and limitations in resource availability compared to Lot A.

By monitoring NDVI and NDRE changes, farmers can gain valuable insights into crop health, allowing them to optimize irrigation and fertilization, ultimately improving yield and quality. For instance, Lot B could benefit from earlier or increased irrigation,while LotC mightrequire adjustmentsto drip irrigation timing or nutrient concentrations.

Further research integrating additional data like soil maps, weather data, and irrigation records holds promise for developing site-specific management strategies for open-field smart agriculture.Additionally,investigatingthephysiologicalmechanisms underlying NDVI and NDRE variations could provide a deeper understanding of targeted interventions, further optimizing crop performance.

Overall, this study demonstrates the significant potential of drone-based vegetation indices for revitalizing open-field agriculture, boosting farm income, and attracting young talent. By unlocking the potential of precision agriculture, we can contribute to a more sustainable and prosperous future for rural communities.